diff --git "a/openvino_model.xml" "b/openvino_model.xml"
new file mode 100644--- /dev/null
+++ "b/openvino_model.xml"
@@ -0,0 +1,49982 @@
+<?xml version="1.0"?>
+<net name="Model25" version="11">
+ <layers>
+ <layer id="3" name="input_ids" type="Parameter" version="opset1">
+ <data shape="?,?" element_type="i64" />
+ <output>
+ <port id="0" precision="I64" names="input_ids">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2" name="attention_mask" type="Parameter" version="opset1">
+ <data shape="?,?" element_type="i64" />
+ <output>
+ <port id="0" precision="I64" names="attention_mask">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1" name="position_ids" type="Parameter" version="opset1">
+ <data shape="?,?" element_type="i64" />
+ <output>
+ <port id="0" precision="I64" names="position_ids">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="0" name="beam_idx" type="Parameter" version="opset1">
+ <data shape="?" element_type="i32" />
+ <output>
+ <port id="0" precision="I32" names="beam_idx">
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="4" name="Constant_164391" type="Const" version="opset1">
+ <data element_type="f32" shape="" offset="0" size="4" />
+ <output>
+ <port id="0" precision="FP32" />
+ </output>
+ </layer>
+ <layer id="5" name="ShapeOf_164377" type="ShapeOf" version="opset3">
+ <data output_type="i64" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="6" name="Constant_164379" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="7" name="Constant_164381" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="8" name="Gather_164382" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="9" name="Constant_164384" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="12" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="10" name="Constant_164386" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="11" name="Constant_164388" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="20" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="12" name="Concat_164389" type="Concat" version="opset1">
+ <data axis="0" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="13" name="Broadcast_164392" type="Broadcast" version="opset3">
+ <data mode="numpy" />
+ <input>
+ <port id="0" precision="FP32" />
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="14" name="ReadValue_164025" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.9.valuepresent.9.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.9.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="15" name="Constant_163575" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="16" name="Gather_163576" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="17" name="Constant_157508" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="28" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="18" name="self.model.embed_tokens.weight" type="Const" version="opset1">
+ <data element_type="u8" shape="128256, 2048" offset="8220" size="262668288" />
+ <output>
+ <port id="0" precision="U8">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="19" name="Convert_181615" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U8">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="20" name="self.model.embed_tokens.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u8" shape="128256, 1" offset="262676508" size="128256" />
+ <output>
+ <port id="0" precision="U8">
+ <dim>128256</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="21" name="Convert_181618" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U8">
+ <dim>128256</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>128256</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="22" name="self.model.embed_tokens.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>128256</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="23" name="self.model.embed_tokens.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="128256, 1" offset="262804764" size="256512" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>128256</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="24" name="self.model.embed_tokens.weight/fq_weights_0" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>128256</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="25" name="__module.model.embed_tokens/ov_ext::embedding/Convert" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="26" name="__module.model.embed_tokens/ov_ext::embedding/Convert_1" type="Convert" version="opset1">
+ <data destination_type="i32" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="27" name="__module.model.embed_tokens/ov_ext::embedding/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="" offset="0" size="4" />
+ <output>
+ <port id="0" precision="I32" />
+ </output>
+ </layer>
+ <layer id="28" name="__module.model.embed_tokens/ov_ext::embedding/Gather" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I32" />
+ </input>
+ <output>
+ <port id="3" precision="FP32" names="183,266,277,hidden_states.1,inputs_embeds">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="29" name="Constant_157417" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="263061276" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="30" name="Constant_157416" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="31" name="Constant_157414" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="32" name="__module.model.layers.0.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="267">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="33" name="Constant_128400" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="268">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="34" name="__module.model.layers.0.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="269,variance.1">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="35" name="Constant_157415" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="36" name="__module.model.layers.0.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="270">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="37" name="__module.model.layers.0.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="38" name="__module.model.layers.0.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="271">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="39" name="__module.model.layers.0.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="272,273,hidden_states.3,hidden_states.5">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="40" name="__module.model.layers.0.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="274,276,hidden_states.7">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="41" name="self.model.layers.0.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="263069488" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="42" name="Convert_182114" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="43" name="self.model.layers.0.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="265166640" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="44" name="Convert_182116" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="45" name="self.model.layers.0.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="46" name="self.model.layers.0.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="265199408" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="47" name="self.model.layers.0.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="48" name="Constant_182121" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="49" name="Reshape_182122" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="50" name="__module.model.layers.0.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="51" name="__module.model.layers.0.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="285,query_states.1">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="52" name="Constant_162268" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="53" name="__module.model.layers.0.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="291">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="54" name="__module.model.layers.0.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="55" name="__module.model.layers.0.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="292,q.1">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="56" name="__module.model.rotary_emb/aten::unsqueeze/Unsqueeze_1" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 32, 1" offset="265330544" size="128" />
+ <output>
+ <port id="0" precision="FP32" names="237,238">
+ <dim>1</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="57" name="ShapeOf_158522" type="ShapeOf" version="opset3">
+ <data output_type="i64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="58" name="Constant_158523" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="59" name="Constant_158524" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="60" name="Gather_158525" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>3</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="I64" names="1110,1126,1264,1280,1418,1434,1572,1588,1726,1742,1880,1896,2034,2050,209,2188,2204,2342,2358,239,2496,2512,2650,2666,340,356,494,510,648,664,802,818,956,972">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="61" name="Constant_153633" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="62" name="Constant_153635" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="63" name="__module.model.rotary_emb/prim::ListConstruct" type="Concat" version="opset1">
+ <data axis="0" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="3" precision="I64">
+ <dim>3</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="64" name="__module.model.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="241,246,inv_freq_expanded,inv_freq_expanded.1">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="65" name="144" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64" names="144" />
+ </output>
+ </layer>
+ <layer id="66" name="__module.model.rotary_emb/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="I64">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="I64" names="243,244">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="67" name="__module.model.rotary_emb/aten::to/Convert_2" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="245,247,position_ids_expanded,position_ids_expanded.1">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="68" name="__module.model.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="248">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="69" name="__module.model.rotary_emb/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="3" offset="265330680" size="12" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>3</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="70" name="__module.model.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="249">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="71" name="__module.model.rotary_emb/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="251,emb">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="72" name="__module.model.rotary_emb/aten::cos/Cos" type="Cos" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="252,254,256,259,cos.1,cos.3,cos.5">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="73" name="__module.model.layers.0.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1072,1226,1380,1534,1688,1842,1996,2150,2304,2458,2612,299,456,610,764,918,cos,cos.11,cos.13,cos.15,cos.17,cos.19,cos.21,cos.23,cos.25,cos.27,cos.29,cos.31,cos.33,cos.35,cos.7,cos.9">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="74" name="__module.model.layers.0.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="301">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="75" name="Constant_128797" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="76" name="Constant_128799" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="77" name="Constant_128801" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="78" name="__module.model.layers.0.self_attn/aten::slice/Reshape" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330708" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="79" name="__module.model.layers.0.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="311,x2.1">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="80" name="Constant_157418" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="81" name="__module.model.layers.0.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="312">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="82" name="Constant_128694" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="83" name="Constant_128738" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="84" name="Constant_128740" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="85" name="__module.model.layers.0.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="306">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="86" name="__module.model.layers.0.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="314">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="87" name="__module.model.rotary_emb/aten::sin/Sin" type="Sin" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="253,255,257,260,sin.1,sin.3,sin.5">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="88" name="__module.model.layers.0.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1073,1227,1381,1535,1689,1843,1997,2151,2305,2459,2613,300,457,611,765,919,sin,sin.11,sin.13,sin.15,sin.17,sin.19,sin.21,sin.23,sin.25,sin.27,sin.29,sin.31,sin.33,sin.35,sin.7,sin.9">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="89" name="__module.model.layers.0.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="315">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="90" name="__module.model.layers.0.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="316">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="91" name="ReadValue_163963" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.0.keypresent.0.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.0.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="92" name="Constant_163518" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="93" name="Gather_163519" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="94" name="self.model.layers.0.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="265330720" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="95" name="Convert_182444" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="96" name="self.model.layers.0.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="265855008" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="97" name="Convert_182446" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="98" name="self.model.layers.0.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="99" name="self.model.layers.0.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="265863200" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="100" name="self.model.layers.0.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="101" name="Constant_182451" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="102" name="Reshape_182452" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="103" name="__module.model.layers.0.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="104" name="__module.model.layers.0.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="287,key_states.3">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="105" name="Constant_162269" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="106" name="__module.model.layers.0.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="294">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="107" name="__module.model.layers.0.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="108" name="__module.model.layers.0.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="295,k.1">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="109" name="__module.model.layers.0.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="317">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="110" name="Constant_128940" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="111" name="Constant_128942" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="112" name="Constant_128944" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="113" name="__module.model.layers.0.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="327,x2.3">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="114" name="Constant_157419" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="115" name="__module.model.layers.0.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="328">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="116" name="Constant_128837" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="117" name="Constant_128881" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="118" name="Constant_128883" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="119" name="__module.model.layers.0.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="322">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="120" name="__module.model.layers.0.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="330">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="121" name="__module.model.layers.0.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="331">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="122" name="__module.model.layers.0.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="332,key_states.5">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="123" name="__module.model.layers.0.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.0.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="124" name="146" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="265896016" size="8" />
+ <output>
+ <port id="0" precision="I64" names="146" />
+ </output>
+ </layer>
+ <layer id="125" name="__module.model.layers.0.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="347,348,349">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="126" name="Constant_153641" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="12" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="127" name="Constant_153643" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265896024" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="128" name="ShapeOf_158566" type="ShapeOf" version="opset3">
+ <data output_type="i64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="129" name="Constant_158567" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="265896016" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="130" name="Constant_158568" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="131" name="Gather_158569" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ <port id="1" precision="I64" />
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="I64" names="187" />
+ </output>
+ </layer>
+ <layer id="132" name="Constant_180534" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="133" name="Reshape_158581" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="I64" />
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="134" name="Constant_158535" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="135" name="Constant_158536" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="136" name="Gather_158537" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>3</dim>
+ </port>
+ <port id="1" precision="I64" />
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="I64" names="189,194" />
+ </output>
+ </layer>
+ <layer id="137" name="Constant_180535" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="138" name="Reshape_158583" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="I64" />
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="139" name="Add_158584" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="I64" names="1113,1142,1267,1296,1421,1450,1575,1604,1729,1758,1883,1912,2037,2066,2191,2220,2345,2374,2499,2528,2653,2682,343,372,497,526,651,680,805,834,959,988">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="140" name="Constant_153646" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="20" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="141" name="__module.model.layers.0.self_attn/prim::ListConstruct_2" type="Concat" version="opset1">
+ <data axis="0" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="I64">
+ <dim>5</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="142" name="__module.model.layers.0.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="351,hidden_states.11">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="143" name="Constant_162270" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="144" name="__module.model.layers.0.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="355,key_states.7">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="145" name="ReadValue_163965" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.0.valuepresent.0.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.0.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="146" name="Constant_163521" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="147" name="Gather_163522" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="148" name="self.model.layers.0.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="265896064" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="149" name="Convert_182434" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="150" name="self.model.layers.0.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="266420352" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="151" name="Convert_182436" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="152" name="self.model.layers.0.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="153" name="self.model.layers.0.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="266428544" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="154" name="self.model.layers.0.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="155" name="Constant_182441" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="156" name="Reshape_182442" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="157" name="__module.model.layers.0.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="158" name="__module.model.layers.0.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="289,value_states.1">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="159" name="Constant_162271" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="160" name="__module.model.layers.0.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="297">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="161" name="__module.model.layers.0.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="162" name="__module.model.layers.0.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="298">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="163" name="__module.model.layers.0.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.0.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="164" name="__module.model.layers.0.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="363,364,365">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="165" name="__module.model.layers.0.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="367,hidden_states.15">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="166" name="Constant_162272" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="167" name="__module.model.layers.0.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="371">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="168" name="__module.model/aten::triu/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="" offset="0" size="4" />
+ <output>
+ <port id="0" precision="I32" />
+ </output>
+ </layer>
+ <layer id="169" name="Multiply_155215" type="Const" version="opset1">
+ <data element_type="f32" shape="" offset="266461312" size="4" />
+ <output>
+ <port id="0" precision="FP32" />
+ </output>
+ </layer>
+ <layer id="170" name="Constant_155220" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="266461316" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="171" name="ShapeOf_158542" type="ShapeOf" version="opset3">
+ <data output_type="i64" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="172" name="Constant_158543" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="173" name="Constant_158544" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="174" name="Gather_158545" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ <port id="1" precision="I64" />
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="I64" names="195,213" />
+ </output>
+ </layer>
+ <layer id="175" name="Constant_180536" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="176" name="Reshape_158547" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="I64" />
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="177" name="__module.model/prim::ListConstruct" type="Concat" version="opset1">
+ <data axis="0" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="178" name="Maximum_155221" type="Maximum" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="179" name="__module.model/aten::mul/Multiply" type="Broadcast" version="opset3">
+ <data mode="numpy" />
+ <input>
+ <port id="0" precision="FP32" />
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="198">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="180" name="ShapeOf_158552" type="ShapeOf" version="opset3">
+ <data output_type="i32" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="I32">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="181" name="Constant_158553" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="182" name="Constant_158554" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="183" name="Gather_158555" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="I32">
+ <dim>2</dim>
+ </port>
+ <port id="1" precision="I64" />
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="I32" />
+ </output>
+ </layer>
+ <layer id="184" name="__module.model/aten::triu/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="" offset="266461332" size="4" />
+ <output>
+ <port id="0" precision="I32" />
+ </output>
+ </layer>
+ <layer id="185" name="__module.model/aten::triu/Range" type="Range" version="opset4">
+ <data output_type="i32" />
+ <input>
+ <port id="0" precision="I32" />
+ <port id="1" precision="I32" />
+ <port id="2" precision="I32" />
+ </input>
+ <output>
+ <port id="3" precision="I32">
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="186" name="__module.model/aten::triu/Unsqueeze" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I32" />
+ </input>
+ <output>
+ <port id="2" precision="I32">
+ <dim>1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="187" name="__module.model/aten::triu/Convert" type="Const" version="opset1">
+ <data element_type="i32" shape="" offset="266461332" size="4" />
+ <output>
+ <port id="0" precision="I32" />
+ </output>
+ </layer>
+ <layer id="188" name="Constant_158559" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="189" name="Constant_158560" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="190" name="Gather_158561" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="I32">
+ <dim>2</dim>
+ </port>
+ <port id="1" precision="I64" />
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="I32" />
+ </output>
+ </layer>
+ <layer id="191" name="__module.model/aten::triu/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="I32" />
+ <port id="1" precision="I32" />
+ </input>
+ <output>
+ <port id="2" precision="I32" />
+ </output>
+ </layer>
+ <layer id="192" name="__module.model/aten::triu/Range_1" type="Range" version="opset4">
+ <data output_type="i32" />
+ <input>
+ <port id="0" precision="I32" />
+ <port id="1" precision="I32" />
+ <port id="2" precision="I32" />
+ </input>
+ <output>
+ <port id="3" precision="I32">
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="193" name="__module.model/aten::triu/Unsqueeze_1" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I32" />
+ </input>
+ <output>
+ <port id="2" precision="I32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="194" name="__module.model/aten::triu/GreaterEqual" type="GreaterEqual" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="I32">
+ <dim>1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="BOOL">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="195" name="__module.model/aten::triu/ConvertLike" type="Const" version="opset1">
+ <data element_type="f32" shape="" offset="0" size="4" />
+ <output>
+ <port id="0" precision="FP32" />
+ </output>
+ </layer>
+ <layer id="196" name="__module.model/aten::triu/Select" type="Select" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="BOOL">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="FP32" />
+ </input>
+ <output>
+ <port id="3" precision="FP32" names="199_1">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="197" name="__module.model/aten::arange/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="" offset="0" size="4" />
+ <output>
+ <port id="0" precision="I32" />
+ </output>
+ </layer>
+ <layer id="198" name="__module.model/aten::arange/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="" offset="266461332" size="4" />
+ <output>
+ <port id="0" precision="I32" />
+ </output>
+ </layer>
+ <layer id="199" name="__module.model/aten::arange/Range" type="Range" version="opset4">
+ <data output_type="f32" />
+ <input>
+ <port id="0" precision="I32" />
+ <port id="1" precision="I64" />
+ <port id="2" precision="I32" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="200" name="__module.model/aten::arange/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="i64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="I64" names="200">
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="201" name="__module.model/aten::gt/Convert" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="202" name="__module.model/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="I64" />
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="I64" names="191" />
+ </output>
+ </layer>
+ <layer id="203" name="__module.model/aten::arange/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="" offset="266461332" size="4" />
+ <output>
+ <port id="0" precision="I32" />
+ </output>
+ </layer>
+ <layer id="204" name="__module.model/aten::arange/Range_1" type="Range" version="opset4">
+ <data output_type="f32" />
+ <input>
+ <port id="0" precision="I64" />
+ <port id="1" precision="I64" />
+ <port id="2" precision="I32" />
+ </input>
+ <output>
+ <port id="3" precision="FP32" names="193,cache_position">
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="205" name="Constant_128008" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="266461336" size="16" />
+ <output>
+ <port id="0" precision="I64" names="201">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="206" name="__module.model/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="202">
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="207" name="__module.model/aten::gt/Greater" type="Greater" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="BOOL" names="203">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="208" name="__module.model/aten::mul_/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <input>
+ <port id="0" precision="BOOL">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="209" name="__module.model/aten::mul_/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="199,causal_mask.1">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="210" name="137" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" names="137" />
+ </output>
+ </layer>
+ <layer id="211" name="__module.model/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="205">
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="212" name="__module.model/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="206,207,208">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="213" name="Constant_153669" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="214" name="Constant_153671" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="215" name="Constant_153673" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="216" name="__module.model/prim::ListConstruct_1" type="Concat" version="opset1">
+ <data axis="0" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="217" name="__module.model/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="211,214,215,216,224,225,226,229_1,230_1,231_1,causal_mask.5">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="218" name="Constant_180537" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="219" name="SliceAssign_128225/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="220" name="Constant_144472" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="221" name="SliceAssign_128225/ShapeOf" type="ShapeOf" version="opset3">
+ <data output_type="i64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="222" name="SliceAssign_128225/ReduceProd" type="ReduceProd" version="opset1">
+ <data keep_dims="false" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="I64" />
+ </output>
+ </layer>
+ <layer id="223" name="Constant_144473" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="224" name="SliceAssign_128225/Range" type="Range" version="opset4">
+ <data output_type="i64" />
+ <input>
+ <port id="0" precision="I64" />
+ <port id="1" precision="I64" />
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="I64">
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="225" name="SliceAssign_128225/Reshape" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="I64">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="226" name="Constant_128214" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="227" name="Constant_128217" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="228" name="SliceAssign_128225/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="I64">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="I64">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="229" name="Constant_144475" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="266461336" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="230" name="SliceAssign_128225/Reshape_4" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="I64">
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="231" name="Constant_128097" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="232" name="Constant_128100" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="233" name="__module.model/aten::slice/Slice_5" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="217,227,232_1">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="234" name="__module.model/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="I64">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="I64" names="219">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="235" name="__module.model/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="I64">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="I64" names="220,221">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="236" name="__module.model/aten::add/Convert" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="237" name="__module.model/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="222,padding_mask.1">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="238" name="Constant_157420" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="0" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="239" name="__module.model/aten::eq/Equal" type="Equal" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="BOOL" names="223,padding_mask">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="240" name="134" type="Const" version="opset1">
+ <data element_type="f32" shape="" offset="266461312" size="4" />
+ <output>
+ <port id="0" precision="FP32" names="134" />
+ </output>
+ </layer>
+ <layer id="241" name="__module.model/aten::masked_fill/Select" type="Select" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="BOOL">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32" />
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="3" precision="FP32" names="228">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="242" name="__module.model/aten::copy_/ShapeOf" type="ShapeOf" version="opset3">
+ <data output_type="i64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="243" name="__module.model/aten::copy_/Broadcast" type="Broadcast" version="opset3">
+ <data mode="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="232">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="244" name="Constant_180538" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="245" name="SliceAssign_128225/Reshape_5" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="246" name="SliceAssign_128225/ScatterNDUpdate" type="ScatterNDUpdate" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="247" name="SliceAssign_128225/Reshape_6" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1143,1144,1145,1297,1298,1299,1451,1452,1453,1605,1606,1607,1759,1760,1761,1913,1914,1915,2067,2068,2069,212,2221,2222,2223,2375,2376,2377,2529,2530,2531,2683,2684,2685,373,374,375,527,528,529,681,682,683,835,836,837,989,990,991">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="248" name="Constant_129200" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="249" name="Constant_129203" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="250" name="__module.model.layers.0.self_attn/aten::slice/Slice_15" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1146,1300,1454,1608,1762,1916,2070,2224,2378,2532,2686,376,530,684,838,992">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="251" name="__module.model.layers.0.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="377,attn_output.1">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="252" name="__module.model.layers.0.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="253" name="__module.model.layers.0.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="378">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="254" name="Constant_154726" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="266461352" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="255" name="Constant_154727" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="256" name="Gather_154728" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>3</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="257" name="Constant_129277" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="258" name="__module.model.layers.0.self_attn/prim::ListConstruct_7" type="Concat" version="opset1">
+ <data axis="0" />
+ <input>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="I64">
+ <dim>3</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="259" name="__module.model.layers.0.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="381,a.1">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="260" name="self.model.layers.0.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="266461368" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="261" name="Convert_182124" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="262" name="self.model.layers.0.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="268558520" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="263" name="Convert_182126" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="264" name="self.model.layers.0.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="265" name="self.model.layers.0.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="268591288" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="266" name="self.model.layers.0.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="267" name="Constant_182131" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="268" name="Reshape_182132" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="269" name="__module.model.layers.0.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="270" name="__module.model.layers.0.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="383,385,hidden_states.17">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="271" name="__module.model.layers.0/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="388,390,401,hidden_states.19,hidden_states.21">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="272" name="Constant_157424" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="268722360" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="273" name="Constant_157423" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="274" name="Constant_157421" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="275" name="__module.model.layers.0.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="391">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="276" name="Constant_129305" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="392">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="277" name="__module.model.layers.0.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="393,variance.3">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="278" name="Constant_157422" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="279" name="__module.model.layers.0.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="394">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="280" name="__module.model.layers.0.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="281" name="__module.model.layers.0.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="395">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="282" name="__module.model.layers.0.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="396,397,hidden_states.23,hidden_states.25">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="283" name="__module.model.layers.0.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="398,400,a.3">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="284" name="self.model.layers.0.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="268730552" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="285" name="Convert_181644" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="286" name="self.model.layers.0.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="277119160" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="287" name="Convert_181646" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="288" name="self.model.layers.0.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="289" name="self.model.layers.0.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="277250232" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="290" name="self.model.layers.0.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="291" name="Constant_181651" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="292" name="Reshape_181652" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="293" name="__module.model.layers.0.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="294" name="__module.model.layers.0.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="406,input.1">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="295" name="__module.model.layers.0.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="407">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="296" name="self.model.layers.0.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="277774536" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="297" name="Convert_181634" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="298" name="self.model.layers.0.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="286163144" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="299" name="Convert_181636" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="300" name="self.model.layers.0.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="301" name="self.model.layers.0.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="286294216" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="302" name="self.model.layers.0.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="303" name="Constant_181641" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="304" name="Reshape_181642" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="305" name="__module.model.layers.0.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="306" name="__module.model.layers.0.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="409">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="307" name="__module.model.layers.0.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="410,a.5">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="308" name="self.model.layers.0.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="286818504" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="309" name="Convert_181654" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="310" name="self.model.layers.0.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="295207112" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="311" name="Convert_181656" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="312" name="self.model.layers.0.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="313" name="self.model.layers.0.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="295338184" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="314" name="self.model.layers.0.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="315" name="Constant_181661" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="316" name="Reshape_181662" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="317" name="__module.model.layers.0.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="318" name="__module.model.layers.0.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="412,hidden_states.27">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="319" name="__module.model.layers.0/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="413,415,423,434,hidden_states.29,hidden_states.31">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="320" name="Constant_157428" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="295862488" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="321" name="Constant_157427" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="322" name="Constant_157425" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="323" name="__module.model.layers.1.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="424">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="324" name="Constant_129379" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="425">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="325" name="__module.model.layers.1.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="426,variance.5">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="326" name="Constant_157426" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="327" name="__module.model.layers.1.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="427">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="328" name="__module.model.layers.1.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="329" name="__module.model.layers.1.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="428">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="330" name="__module.model.layers.1.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="429,430,hidden_states.33,hidden_states.35">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="331" name="__module.model.layers.1.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="431,433,hidden_states.37">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="332" name="self.model.layers.1.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="295870680" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="333" name="Convert_182134" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="334" name="self.model.layers.1.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="297967832" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="335" name="Convert_182136" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="336" name="self.model.layers.1.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="337" name="self.model.layers.1.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="298000600" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="338" name="self.model.layers.1.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="339" name="Constant_182141" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="340" name="Reshape_182142" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="341" name="__module.model.layers.1.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="342" name="__module.model.layers.1.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="442,query_states.3">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="343" name="Constant_162273" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="344" name="__module.model.layers.1.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="448">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="345" name="__module.model.layers.1.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="346" name="__module.model.layers.1.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="449,q.3">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="347" name="__module.model.layers.1.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="458">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="348" name="Constant_129776" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="349" name="Constant_129778" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="350" name="Constant_129780" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="351" name="__module.model.layers.1.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="468,x2.5">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="352" name="Constant_157429" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="353" name="__module.model.layers.1.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="469">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="354" name="Constant_129673" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="355" name="Constant_129717" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="356" name="Constant_129719" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="357" name="__module.model.layers.1.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="463">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="358" name="__module.model.layers.1.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="471">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="359" name="__module.model.layers.1.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="472">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="360" name="__module.model.layers.1.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="473">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="361" name="ReadValue_163967" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.1.keypresent.1.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.1.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="362" name="Constant_163524" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="363" name="Gather_163525" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="364" name="self.model.layers.1.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="298131672" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="365" name="Convert_182464" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="366" name="self.model.layers.1.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="298655960" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="367" name="Convert_182466" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="368" name="self.model.layers.1.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="369" name="self.model.layers.1.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="298664152" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="370" name="self.model.layers.1.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="371" name="Constant_182471" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="372" name="Reshape_182472" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="373" name="__module.model.layers.1.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="374" name="__module.model.layers.1.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="444,key_states.9">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="375" name="Constant_162274" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="376" name="__module.model.layers.1.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="451">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="377" name="__module.model.layers.1.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="378" name="__module.model.layers.1.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="452,k.3">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="379" name="__module.model.layers.1.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="474">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="380" name="Constant_129919" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="381" name="Constant_129921" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="382" name="Constant_129923" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="383" name="__module.model.layers.1.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="484,x2.7">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="384" name="Constant_157430" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="385" name="__module.model.layers.1.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="485">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="386" name="Constant_129816" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="387" name="Constant_129860" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="388" name="Constant_129862" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="389" name="__module.model.layers.1.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="479">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="390" name="__module.model.layers.1.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="487">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="391" name="__module.model.layers.1.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="488">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="392" name="__module.model.layers.1.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="489">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="393" name="__module.model.layers.1.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.1.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="394" name="__module.model.layers.1.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="501,502,503">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="395" name="__module.model.layers.1.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="505,hidden_states.41">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="396" name="Constant_162275" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="397" name="__module.model.layers.1.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="509,key_states.11">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="398" name="ReadValue_163969" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.1.valuepresent.1.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.1.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="399" name="Constant_163527" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="400" name="Gather_163528" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="401" name="self.model.layers.1.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="298696920" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="402" name="Convert_182454" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="403" name="self.model.layers.1.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="299221208" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="404" name="Convert_182456" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="405" name="self.model.layers.1.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="406" name="self.model.layers.1.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="299229400" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="407" name="self.model.layers.1.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="408" name="Constant_182461" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="409" name="Reshape_182462" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="410" name="__module.model.layers.1.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="411" name="__module.model.layers.1.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="446,value_states.3">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="412" name="Constant_162276" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="413" name="__module.model.layers.1.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="454">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="414" name="__module.model.layers.1.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="415" name="__module.model.layers.1.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="455">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="416" name="__module.model.layers.1.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.1.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="417" name="__module.model.layers.1.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="517,518,519">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="418" name="__module.model.layers.1.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="521,hidden_states.45">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="419" name="Constant_162277" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="420" name="__module.model.layers.1.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="525">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="421" name="__module.model.layers.1.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="531,attn_output.5">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="422" name="__module.model.layers.1.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="423" name="__module.model.layers.1.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="532">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="424" name="__module.model.layers.1.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="535,a.7">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="425" name="self.model.layers.1.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="299262168" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="426" name="Convert_182144" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="427" name="self.model.layers.1.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="301359320" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="428" name="Convert_182146" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="429" name="self.model.layers.1.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="430" name="self.model.layers.1.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="301392088" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="431" name="self.model.layers.1.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="432" name="Constant_182151" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="433" name="Reshape_182152" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="434" name="__module.model.layers.1.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="435" name="__module.model.layers.1.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="537,539,hidden_states.47">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="436" name="__module.model.layers.1/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="542,544,555,hidden_states.49,hidden_states.51">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="437" name="Constant_157434" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="301523160" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="438" name="Constant_157433" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="439" name="Constant_157431" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="440" name="__module.model.layers.1.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="545">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="441" name="Constant_130278" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="546">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="442" name="__module.model.layers.1.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="547,variance.7">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="443" name="Constant_157432" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="444" name="__module.model.layers.1.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="548">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="445" name="__module.model.layers.1.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="446" name="__module.model.layers.1.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="549">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="447" name="__module.model.layers.1.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="550,551,hidden_states.53,hidden_states.55">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="448" name="__module.model.layers.1.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="552,554,a.9">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="449" name="self.model.layers.1.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="301531352" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="450" name="Convert_181674" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="451" name="self.model.layers.1.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="309919960" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="452" name="Convert_181676" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="453" name="self.model.layers.1.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="454" name="self.model.layers.1.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="310051032" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="455" name="self.model.layers.1.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="456" name="Constant_181681" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="457" name="Reshape_181682" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="458" name="__module.model.layers.1.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="459" name="__module.model.layers.1.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="560,input.3">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="460" name="__module.model.layers.1.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="561">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="461" name="self.model.layers.1.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="310575320" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="462" name="Convert_181664" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="463" name="self.model.layers.1.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="318963928" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="464" name="Convert_181666" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="465" name="self.model.layers.1.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="466" name="self.model.layers.1.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="319095000" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="467" name="self.model.layers.1.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="468" name="Constant_181671" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="469" name="Reshape_181672" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="470" name="__module.model.layers.1.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="471" name="__module.model.layers.1.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="563">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="472" name="__module.model.layers.1.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="564,a.11">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="473" name="self.model.layers.1.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="319619288" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="474" name="Convert_181684" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="475" name="self.model.layers.1.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="328007896" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="476" name="Convert_181686" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="477" name="self.model.layers.1.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="478" name="self.model.layers.1.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="328138968" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="479" name="self.model.layers.1.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="480" name="Constant_181691" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="481" name="Reshape_181692" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="482" name="__module.model.layers.1.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="483" name="__module.model.layers.1.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="566,hidden_states.57">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="484" name="__module.model.layers.1/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="567,569,577,588,hidden_states.59,hidden_states.61">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="485" name="Constant_157438" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="328663256" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="486" name="Constant_157437" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="487" name="Constant_157435" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="488" name="__module.model.layers.2.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="578">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="489" name="Constant_130352" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="579">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="490" name="__module.model.layers.2.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="580,variance.9">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="491" name="Constant_157436" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="492" name="__module.model.layers.2.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="581">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="493" name="__module.model.layers.2.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="494" name="__module.model.layers.2.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="582">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="495" name="__module.model.layers.2.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="583,584,hidden_states.63,hidden_states.65">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="496" name="__module.model.layers.2.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="585,587,hidden_states.67">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="497" name="self.model.layers.2.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="328671448" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="498" name="Convert_182154" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="499" name="self.model.layers.2.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="330768600" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="500" name="Convert_182156" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="501" name="self.model.layers.2.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="502" name="self.model.layers.2.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="330801368" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="503" name="self.model.layers.2.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="504" name="Constant_182161" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="505" name="Reshape_182162" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="506" name="__module.model.layers.2.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="507" name="__module.model.layers.2.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="596,query_states.5">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="508" name="Constant_162278" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="509" name="__module.model.layers.2.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="602">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="510" name="__module.model.layers.2.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="511" name="__module.model.layers.2.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="603,q.5">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="512" name="__module.model.layers.2.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="612">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="513" name="Constant_130749" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="514" name="Constant_130751" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="515" name="Constant_130753" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="516" name="__module.model.layers.2.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="622,x2.9">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="517" name="Constant_157439" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="518" name="__module.model.layers.2.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="623">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="519" name="Constant_130646" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="520" name="Constant_130690" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="521" name="Constant_130692" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="522" name="__module.model.layers.2.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="617">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="523" name="__module.model.layers.2.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="625">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="524" name="__module.model.layers.2.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="626">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="525" name="__module.model.layers.2.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="627">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="526" name="ReadValue_163995" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.2.keypresent.2.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.2.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="527" name="Constant_163530" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="528" name="Gather_163531" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="529" name="self.model.layers.2.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="330932440" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="530" name="Convert_182484" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="531" name="self.model.layers.2.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="331456728" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="532" name="Convert_182486" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="533" name="self.model.layers.2.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="534" name="self.model.layers.2.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="331464920" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="535" name="self.model.layers.2.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="536" name="Constant_182491" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="537" name="Reshape_182492" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="538" name="__module.model.layers.2.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="539" name="__module.model.layers.2.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="598,key_states.13">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="540" name="Constant_162279" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="541" name="__module.model.layers.2.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="605">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="542" name="__module.model.layers.2.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="543" name="__module.model.layers.2.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="606,k.5">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="544" name="__module.model.layers.2.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="628">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="545" name="Constant_130892" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="546" name="Constant_130894" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="547" name="Constant_130896" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="548" name="__module.model.layers.2.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="638,x2.11">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="549" name="Constant_157440" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="550" name="__module.model.layers.2.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="639">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="551" name="Constant_130789" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="552" name="Constant_130833" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="553" name="Constant_130835" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="554" name="__module.model.layers.2.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="633">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="555" name="__module.model.layers.2.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="641">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="556" name="__module.model.layers.2.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="642">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="557" name="__module.model.layers.2.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="643">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="558" name="__module.model.layers.2.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.2.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="559" name="__module.model.layers.2.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="655,656,657">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="560" name="__module.model.layers.2.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="659,hidden_states.71">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="561" name="Constant_162280" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="562" name="__module.model.layers.2.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="663,key_states.15">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="563" name="ReadValue_163997" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.2.valuepresent.2.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.2.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="564" name="Constant_163533" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="565" name="Gather_163534" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="566" name="self.model.layers.2.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="331497688" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="567" name="Convert_182474" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="568" name="self.model.layers.2.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="332021976" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="569" name="Convert_182476" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="570" name="self.model.layers.2.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="571" name="self.model.layers.2.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="332030168" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="572" name="self.model.layers.2.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="573" name="Constant_182481" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="574" name="Reshape_182482" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="575" name="__module.model.layers.2.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="576" name="__module.model.layers.2.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="600,value_states.5">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="577" name="Constant_162281" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="578" name="__module.model.layers.2.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="608">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="579" name="__module.model.layers.2.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="580" name="__module.model.layers.2.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="609">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="581" name="__module.model.layers.2.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.2.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="582" name="__module.model.layers.2.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="671,672,673">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="583" name="__module.model.layers.2.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="675,hidden_states.75">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="584" name="Constant_162282" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="585" name="__module.model.layers.2.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="679">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="586" name="__module.model.layers.2.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="685,attn_output.9">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="587" name="__module.model.layers.2.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="588" name="__module.model.layers.2.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="686">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="589" name="__module.model.layers.2.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="689,a.13">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="590" name="self.model.layers.2.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="332062936" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="591" name="Convert_182164" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="592" name="self.model.layers.2.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="334160088" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="593" name="Convert_182166" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="594" name="self.model.layers.2.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="595" name="self.model.layers.2.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="334192856" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="596" name="self.model.layers.2.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="597" name="Constant_182171" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="598" name="Reshape_182172" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="599" name="__module.model.layers.2.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="600" name="__module.model.layers.2.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="691,693,hidden_states.77">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="601" name="__module.model.layers.2/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="696,698,709,hidden_states.79,hidden_states.81">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="602" name="Constant_157444" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="334323928" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="603" name="Constant_157443" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="604" name="Constant_157441" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="605" name="__module.model.layers.2.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="699">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="606" name="Constant_131251" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="700">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="607" name="__module.model.layers.2.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="701,variance.11">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="608" name="Constant_157442" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="609" name="__module.model.layers.2.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="702">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="610" name="__module.model.layers.2.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="611" name="__module.model.layers.2.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="703">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="612" name="__module.model.layers.2.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="704,705,hidden_states.83,hidden_states.85">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="613" name="__module.model.layers.2.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="706,708,a.15">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="614" name="self.model.layers.2.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="334332120" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="615" name="Convert_181704" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="616" name="self.model.layers.2.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="342720728" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="617" name="Convert_181706" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="618" name="self.model.layers.2.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="619" name="self.model.layers.2.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="342851800" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="620" name="self.model.layers.2.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="621" name="Constant_181711" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="622" name="Reshape_181712" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="623" name="__module.model.layers.2.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="624" name="__module.model.layers.2.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="714,input.5">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="625" name="__module.model.layers.2.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="715">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="626" name="self.model.layers.2.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="343376088" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="627" name="Convert_181694" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="628" name="self.model.layers.2.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="351764696" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="629" name="Convert_181696" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="630" name="self.model.layers.2.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="631" name="self.model.layers.2.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="351895768" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="632" name="self.model.layers.2.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="633" name="Constant_181701" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="634" name="Reshape_181702" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="635" name="__module.model.layers.2.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="636" name="__module.model.layers.2.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="717">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="637" name="__module.model.layers.2.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="718,a.17">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="638" name="self.model.layers.2.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="352420056" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="639" name="Convert_181714" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="640" name="self.model.layers.2.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="360808664" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="641" name="Convert_181716" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="642" name="self.model.layers.2.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="643" name="self.model.layers.2.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="360939736" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="644" name="self.model.layers.2.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="645" name="Constant_181721" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="646" name="Reshape_181722" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="647" name="__module.model.layers.2.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="648" name="__module.model.layers.2.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="720,hidden_states.87">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="649" name="__module.model.layers.2/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="721,723,731,742,hidden_states.89,hidden_states.91">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="650" name="Constant_157448" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="361464024" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="651" name="Constant_157447" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="652" name="Constant_157445" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="653" name="__module.model.layers.3.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="732">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="654" name="Constant_131325" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="733">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="655" name="__module.model.layers.3.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="734,variance.13">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="656" name="Constant_157446" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="657" name="__module.model.layers.3.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="735">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="658" name="__module.model.layers.3.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="659" name="__module.model.layers.3.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="736">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="660" name="__module.model.layers.3.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="737,738,hidden_states.93,hidden_states.95">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="661" name="__module.model.layers.3.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="739,741,hidden_states.97">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="662" name="self.model.layers.3.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="361472216" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="663" name="Convert_182174" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="664" name="self.model.layers.3.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="363569368" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="665" name="Convert_182176" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="666" name="self.model.layers.3.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="667" name="self.model.layers.3.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="363602136" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="668" name="self.model.layers.3.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="669" name="Constant_182181" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="670" name="Reshape_182182" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="671" name="__module.model.layers.3.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="672" name="__module.model.layers.3.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="750,query_states.7">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="673" name="Constant_162283" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="674" name="__module.model.layers.3.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="756">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="675" name="__module.model.layers.3.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="676" name="__module.model.layers.3.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="757,q.7">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="677" name="__module.model.layers.3.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="766">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="678" name="Constant_131722" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="679" name="Constant_131724" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="680" name="Constant_131726" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="681" name="__module.model.layers.3.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="776,x2.13">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="682" name="Constant_157449" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="683" name="__module.model.layers.3.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="777">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="684" name="Constant_131619" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="685" name="Constant_131663" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="686" name="Constant_131665" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="687" name="__module.model.layers.3.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="771">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="688" name="__module.model.layers.3.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="779">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="689" name="__module.model.layers.3.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="780">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="690" name="__module.model.layers.3.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="781">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="691" name="ReadValue_163999" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.3.keypresent.3.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.3.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="692" name="Constant_163536" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="693" name="Gather_163537" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="694" name="self.model.layers.3.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="363733208" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="695" name="Convert_182504" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="696" name="self.model.layers.3.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="364257496" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="697" name="Convert_182506" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="698" name="self.model.layers.3.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="699" name="self.model.layers.3.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="364265688" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="700" name="self.model.layers.3.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="701" name="Constant_182511" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="702" name="Reshape_182512" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="703" name="__module.model.layers.3.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="704" name="__module.model.layers.3.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="752,key_states.17">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="705" name="Constant_162284" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="706" name="__module.model.layers.3.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="759">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="707" name="__module.model.layers.3.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="708" name="__module.model.layers.3.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="760,k.7">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="709" name="__module.model.layers.3.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="782">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="710" name="Constant_131865" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="711" name="Constant_131867" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="712" name="Constant_131869" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="713" name="__module.model.layers.3.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="792,x2.15">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="714" name="Constant_157450" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="715" name="__module.model.layers.3.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="793">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="716" name="Constant_131762" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="717" name="Constant_131806" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="718" name="Constant_131808" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="719" name="__module.model.layers.3.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="787">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="720" name="__module.model.layers.3.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="795">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="721" name="__module.model.layers.3.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="796">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="722" name="__module.model.layers.3.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="797">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="723" name="__module.model.layers.3.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.3.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="724" name="__module.model.layers.3.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="809,810,811">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="725" name="__module.model.layers.3.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="813,hidden_states.101">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="726" name="Constant_162285" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="727" name="__module.model.layers.3.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="817,key_states.19">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="728" name="ReadValue_164001" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.3.valuepresent.3.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.3.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="729" name="Constant_163539" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="730" name="Gather_163540" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="731" name="self.model.layers.3.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="364298456" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="732" name="Convert_182494" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="733" name="self.model.layers.3.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="364822744" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="734" name="Convert_182496" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="735" name="self.model.layers.3.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="736" name="self.model.layers.3.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="364830936" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="737" name="self.model.layers.3.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="738" name="Constant_182501" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="739" name="Reshape_182502" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="740" name="__module.model.layers.3.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="741" name="__module.model.layers.3.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="754,value_states.7">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="742" name="Constant_162286" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="743" name="__module.model.layers.3.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="762">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="744" name="__module.model.layers.3.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="745" name="__module.model.layers.3.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="763">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="746" name="__module.model.layers.3.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.3.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="747" name="__module.model.layers.3.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="825,826,827">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="748" name="__module.model.layers.3.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="829,hidden_states.105">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="749" name="Constant_162287" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="750" name="__module.model.layers.3.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="833">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="751" name="__module.model.layers.3.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="839,attn_output.13">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="752" name="__module.model.layers.3.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="753" name="__module.model.layers.3.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="840">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="754" name="__module.model.layers.3.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="843,a.19">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="755" name="self.model.layers.3.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="364863704" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="756" name="Convert_182184" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="757" name="self.model.layers.3.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="366960856" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="758" name="Convert_182186" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="759" name="self.model.layers.3.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="760" name="self.model.layers.3.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="366993624" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="761" name="self.model.layers.3.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="762" name="Constant_182191" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="763" name="Reshape_182192" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="764" name="__module.model.layers.3.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="765" name="__module.model.layers.3.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="845,847,hidden_states.107">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="766" name="__module.model.layers.3/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="850,852,863,hidden_states.109,hidden_states.111">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="767" name="Constant_157454" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="367124696" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="768" name="Constant_157453" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="769" name="Constant_157451" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="770" name="__module.model.layers.3.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="853">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="771" name="Constant_132224" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="854">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="772" name="__module.model.layers.3.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="855,variance.15">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="773" name="Constant_157452" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="774" name="__module.model.layers.3.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="856">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="775" name="__module.model.layers.3.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="776" name="__module.model.layers.3.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="857">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="777" name="__module.model.layers.3.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="858,859,hidden_states.113,hidden_states.115">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="778" name="__module.model.layers.3.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="860,862,a.21">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="779" name="self.model.layers.3.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="367132888" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="780" name="Convert_181734" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="781" name="self.model.layers.3.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="375521496" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="782" name="Convert_181736" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="783" name="self.model.layers.3.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="784" name="self.model.layers.3.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="375652568" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="785" name="self.model.layers.3.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="786" name="Constant_181741" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="787" name="Reshape_181742" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="788" name="__module.model.layers.3.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="789" name="__module.model.layers.3.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="868,input.7">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="790" name="__module.model.layers.3.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="869">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="791" name="self.model.layers.3.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="376176856" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="792" name="Convert_181724" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="793" name="self.model.layers.3.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="384565464" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="794" name="Convert_181726" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="795" name="self.model.layers.3.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="796" name="self.model.layers.3.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="384696536" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="797" name="self.model.layers.3.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="798" name="Constant_181731" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="799" name="Reshape_181732" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="800" name="__module.model.layers.3.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="801" name="__module.model.layers.3.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="871">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="802" name="__module.model.layers.3.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="872,a.23">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="803" name="self.model.layers.3.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="385220824" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="804" name="Convert_181744" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="805" name="self.model.layers.3.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="393609432" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="806" name="Convert_181746" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="807" name="self.model.layers.3.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="808" name="self.model.layers.3.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="393740504" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="809" name="self.model.layers.3.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="810" name="Constant_181751" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="811" name="Reshape_181752" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="812" name="__module.model.layers.3.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="813" name="__module.model.layers.3.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="874,hidden_states.117">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="814" name="__module.model.layers.3/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="875,877,885,896,hidden_states.119,hidden_states.121">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="815" name="Constant_157458" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="394264792" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="816" name="Constant_157457" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="817" name="Constant_157455" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="818" name="__module.model.layers.4.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="886">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="819" name="Constant_132298" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="887">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="820" name="__module.model.layers.4.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="888,variance.17">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="821" name="Constant_157456" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="822" name="__module.model.layers.4.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="889">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="823" name="__module.model.layers.4.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="824" name="__module.model.layers.4.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="890">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="825" name="__module.model.layers.4.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="891,892,hidden_states.123,hidden_states.125">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="826" name="__module.model.layers.4.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="893,895,hidden_states.127">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="827" name="self.model.layers.4.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="394272984" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="828" name="Convert_182194" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="829" name="self.model.layers.4.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="396370136" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="830" name="Convert_182196" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="831" name="self.model.layers.4.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="832" name="self.model.layers.4.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="396402904" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="833" name="self.model.layers.4.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="834" name="Constant_182201" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="835" name="Reshape_182202" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="836" name="__module.model.layers.4.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="837" name="__module.model.layers.4.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="904,query_states.9">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="838" name="Constant_162288" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="839" name="__module.model.layers.4.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="910">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="840" name="__module.model.layers.4.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="841" name="__module.model.layers.4.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="911,q.9">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="842" name="__module.model.layers.4.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="920">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="843" name="Constant_132695" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="844" name="Constant_132697" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="845" name="Constant_132699" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="846" name="__module.model.layers.4.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="930,x2.17">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="847" name="Constant_157459" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="848" name="__module.model.layers.4.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="931">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="849" name="Constant_132592" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="850" name="Constant_132636" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="851" name="Constant_132638" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="852" name="__module.model.layers.4.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="925">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="853" name="__module.model.layers.4.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="933">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="854" name="__module.model.layers.4.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="934">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="855" name="__module.model.layers.4.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="935">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="856" name="ReadValue_164003" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.4.keypresent.4.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.4.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="857" name="Constant_163542" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="858" name="Gather_163543" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="859" name="self.model.layers.4.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="396533976" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="860" name="Convert_182524" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="861" name="self.model.layers.4.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="397058264" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="862" name="Convert_182526" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="863" name="self.model.layers.4.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="864" name="self.model.layers.4.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="397066456" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="865" name="self.model.layers.4.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="866" name="Constant_182531" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="867" name="Reshape_182532" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="868" name="__module.model.layers.4.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="869" name="__module.model.layers.4.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="906,key_states.21">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="870" name="Constant_162289" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="871" name="__module.model.layers.4.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="913">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="872" name="__module.model.layers.4.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="873" name="__module.model.layers.4.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="914,k.9">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="874" name="__module.model.layers.4.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="936">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="875" name="Constant_132838" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="876" name="Constant_132840" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="877" name="Constant_132842" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="878" name="__module.model.layers.4.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="946,x2.19">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="879" name="Constant_157460" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="880" name="__module.model.layers.4.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="947">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="881" name="Constant_132735" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="882" name="Constant_132779" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="883" name="Constant_132781" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="884" name="__module.model.layers.4.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="941">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="885" name="__module.model.layers.4.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="949">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="886" name="__module.model.layers.4.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="950">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="887" name="__module.model.layers.4.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="951">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="888" name="__module.model.layers.4.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.4.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="889" name="__module.model.layers.4.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="963,964,965">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="890" name="__module.model.layers.4.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="967,hidden_states.131">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="891" name="Constant_162290" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="892" name="__module.model.layers.4.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="971,key_states.23">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="893" name="ReadValue_164005" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.4.valuepresent.4.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.4.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="894" name="Constant_163545" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="895" name="Gather_163546" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="896" name="self.model.layers.4.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="397099224" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="897" name="Convert_182514" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="898" name="self.model.layers.4.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="397623512" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="899" name="Convert_182516" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="900" name="self.model.layers.4.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="901" name="self.model.layers.4.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="397631704" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="902" name="self.model.layers.4.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="903" name="Constant_182521" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="904" name="Reshape_182522" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="905" name="__module.model.layers.4.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="906" name="__module.model.layers.4.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="908,value_states.9">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="907" name="Constant_162291" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="908" name="__module.model.layers.4.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="916">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="909" name="__module.model.layers.4.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="910" name="__module.model.layers.4.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="917">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="911" name="__module.model.layers.4.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.4.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="912" name="__module.model.layers.4.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="979,980,981">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="913" name="__module.model.layers.4.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="983,hidden_states.135">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="914" name="Constant_162292" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="915" name="__module.model.layers.4.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="987">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="916" name="__module.model.layers.4.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="993,attn_output.17">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="917" name="__module.model.layers.4.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="918" name="__module.model.layers.4.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="994">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="919" name="__module.model.layers.4.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="997,a.25">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="920" name="self.model.layers.4.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="397664472" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="921" name="Convert_182204" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="922" name="self.model.layers.4.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="399761624" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="923" name="Convert_182206" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="924" name="self.model.layers.4.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="925" name="self.model.layers.4.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="399794392" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="926" name="self.model.layers.4.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="927" name="Constant_182211" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="928" name="Reshape_182212" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="929" name="__module.model.layers.4.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="930" name="__module.model.layers.4.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1001,999,hidden_states.137">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="931" name="__module.model.layers.4/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1004,1006,1017,hidden_states.139,hidden_states.141">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="932" name="Constant_157464" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="399925464" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="933" name="Constant_157463" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="934" name="Constant_157461" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="935" name="__module.model.layers.4.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1007">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="936" name="Constant_133197" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1008">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="937" name="__module.model.layers.4.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1009,variance.19">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="938" name="Constant_157462" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="939" name="__module.model.layers.4.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1010">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="940" name="__module.model.layers.4.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="941" name="__module.model.layers.4.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1011">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="942" name="__module.model.layers.4.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1012,1013,hidden_states.143,hidden_states.145">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="943" name="__module.model.layers.4.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1014,1016,a.27">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="944" name="self.model.layers.4.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="399933656" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="945" name="Convert_181764" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="946" name="self.model.layers.4.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="408322264" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="947" name="Convert_181766" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="948" name="self.model.layers.4.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="949" name="self.model.layers.4.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="408453336" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="950" name="self.model.layers.4.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="951" name="Constant_181771" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="952" name="Reshape_181772" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="953" name="__module.model.layers.4.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="954" name="__module.model.layers.4.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1022,input.9">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="955" name="__module.model.layers.4.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="1023">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="956" name="self.model.layers.4.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="408977624" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="957" name="Convert_181754" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="958" name="self.model.layers.4.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="417366232" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="959" name="Convert_181756" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="960" name="self.model.layers.4.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="961" name="self.model.layers.4.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="417497304" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="962" name="self.model.layers.4.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="963" name="Constant_181761" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="964" name="Reshape_181762" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="965" name="__module.model.layers.4.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="966" name="__module.model.layers.4.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1025">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="967" name="__module.model.layers.4.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1026,a.29">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="968" name="self.model.layers.4.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="418021592" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="969" name="Convert_181774" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="970" name="self.model.layers.4.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="426410200" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="971" name="Convert_181776" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="972" name="self.model.layers.4.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="973" name="self.model.layers.4.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="426541272" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="974" name="self.model.layers.4.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="975" name="Constant_181781" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="976" name="Reshape_181782" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="977" name="__module.model.layers.4.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="978" name="__module.model.layers.4.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1028,hidden_states.147">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="979" name="__module.model.layers.4/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1029,1031,1039,1050,hidden_states.149,hidden_states.151">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="980" name="Constant_157468" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="427065560" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="981" name="Constant_157467" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="982" name="Constant_157465" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="983" name="__module.model.layers.5.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1040">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="984" name="Constant_133271" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1041">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="985" name="__module.model.layers.5.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1042,variance.21">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="986" name="Constant_157466" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="987" name="__module.model.layers.5.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1043">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="988" name="__module.model.layers.5.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="989" name="__module.model.layers.5.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1044">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="990" name="__module.model.layers.5.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1045,1046,hidden_states.153,hidden_states.155">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="991" name="__module.model.layers.5.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1047,1049,hidden_states.157">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="992" name="self.model.layers.5.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="427073752" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="993" name="Convert_182214" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="994" name="self.model.layers.5.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="429170904" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="995" name="Convert_182216" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="996" name="self.model.layers.5.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="997" name="self.model.layers.5.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="429203672" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="998" name="self.model.layers.5.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="999" name="Constant_182221" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1000" name="Reshape_182222" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1001" name="__module.model.layers.5.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1002" name="__module.model.layers.5.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1058,query_states.11">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1003" name="Constant_162293" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1004" name="__module.model.layers.5.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1064">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1005" name="__module.model.layers.5.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1006" name="__module.model.layers.5.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1065,q.11">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1007" name="__module.model.layers.5.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1074">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1008" name="Constant_133668" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1009" name="Constant_133670" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1010" name="Constant_133672" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1011" name="__module.model.layers.5.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1084,x2.21">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1012" name="Constant_157469" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1013" name="__module.model.layers.5.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1085">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1014" name="Constant_133565" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1015" name="Constant_133609" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1016" name="Constant_133611" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1017" name="__module.model.layers.5.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1079">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1018" name="__module.model.layers.5.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1087">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1019" name="__module.model.layers.5.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1088">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1020" name="__module.model.layers.5.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1089">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1021" name="ReadValue_164007" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.5.keypresent.5.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.5.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1022" name="Constant_163548" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="1023" name="Gather_163549" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1024" name="self.model.layers.5.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="429334744" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1025" name="Convert_182544" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1026" name="self.model.layers.5.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="429859032" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1027" name="Convert_182546" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1028" name="self.model.layers.5.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1029" name="self.model.layers.5.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="429867224" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1030" name="self.model.layers.5.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1031" name="Constant_182551" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1032" name="Reshape_182552" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1033" name="__module.model.layers.5.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1034" name="__module.model.layers.5.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1060,key_states.25">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1035" name="Constant_162294" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1036" name="__module.model.layers.5.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1067">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1037" name="__module.model.layers.5.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1038" name="__module.model.layers.5.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1068,k.11">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1039" name="__module.model.layers.5.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1090">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1040" name="Constant_133811" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1041" name="Constant_133813" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1042" name="Constant_133815" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1043" name="__module.model.layers.5.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1100,x2.23">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1044" name="Constant_157470" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1045" name="__module.model.layers.5.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1101">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1046" name="Constant_133708" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1047" name="Constant_133752" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1048" name="Constant_133754" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1049" name="__module.model.layers.5.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1095">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1050" name="__module.model.layers.5.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1103">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1051" name="__module.model.layers.5.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1104">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1052" name="__module.model.layers.5.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1105">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1053" name="__module.model.layers.5.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.5.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1054" name="__module.model.layers.5.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1117,1118,1119">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1055" name="__module.model.layers.5.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1121,hidden_states.161">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1056" name="Constant_162295" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1057" name="__module.model.layers.5.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1125,key_states.27">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1058" name="ReadValue_164009" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.5.valuepresent.5.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.5.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1059" name="Constant_163551" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="1060" name="Gather_163552" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1061" name="self.model.layers.5.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="429899992" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1062" name="Convert_182534" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1063" name="self.model.layers.5.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="430424280" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1064" name="Convert_182536" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1065" name="self.model.layers.5.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1066" name="self.model.layers.5.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="430432472" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1067" name="self.model.layers.5.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1068" name="Constant_182541" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1069" name="Reshape_182542" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1070" name="__module.model.layers.5.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1071" name="__module.model.layers.5.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1062,value_states.11">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1072" name="Constant_162296" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1073" name="__module.model.layers.5.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1070">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1074" name="__module.model.layers.5.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1075" name="__module.model.layers.5.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1071">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1076" name="__module.model.layers.5.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.5.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1077" name="__module.model.layers.5.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1133,1134,1135">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1078" name="__module.model.layers.5.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1137,hidden_states.165">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1079" name="Constant_162297" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1080" name="__module.model.layers.5.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1141">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1081" name="__module.model.layers.5.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="1147,attn_output.21">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1082" name="__module.model.layers.5.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1083" name="__module.model.layers.5.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1148">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1084" name="__module.model.layers.5.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1151,a.31">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1085" name="self.model.layers.5.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="430465240" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1086" name="Convert_182224" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1087" name="self.model.layers.5.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="432562392" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1088" name="Convert_182226" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1089" name="self.model.layers.5.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1090" name="self.model.layers.5.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="432595160" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1091" name="self.model.layers.5.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1092" name="Constant_182231" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1093" name="Reshape_182232" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1094" name="__module.model.layers.5.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1095" name="__module.model.layers.5.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1153,1155,hidden_states.167">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1096" name="__module.model.layers.5/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1158,1160,1171,hidden_states.169,hidden_states.171">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1097" name="Constant_157474" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="432726232" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1098" name="Constant_157473" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1099" name="Constant_157471" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1100" name="__module.model.layers.5.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1161">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1101" name="Constant_134170" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1162">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1102" name="__module.model.layers.5.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1163,variance.23">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1103" name="Constant_157472" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1104" name="__module.model.layers.5.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1164">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1105" name="__module.model.layers.5.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1106" name="__module.model.layers.5.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1165">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1107" name="__module.model.layers.5.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1166,1167,hidden_states.173,hidden_states.175">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1108" name="__module.model.layers.5.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1168,1170,a.33">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1109" name="self.model.layers.5.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="432734424" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1110" name="Convert_181794" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1111" name="self.model.layers.5.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="441123032" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1112" name="Convert_181796" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1113" name="self.model.layers.5.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1114" name="self.model.layers.5.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="441254104" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1115" name="self.model.layers.5.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1116" name="Constant_181801" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1117" name="Reshape_181802" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1118" name="__module.model.layers.5.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1119" name="__module.model.layers.5.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1176,input.11">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1120" name="__module.model.layers.5.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="1177">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1121" name="self.model.layers.5.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="441778392" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1122" name="Convert_181784" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1123" name="self.model.layers.5.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="450167000" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1124" name="Convert_181786" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1125" name="self.model.layers.5.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1126" name="self.model.layers.5.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="450298072" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1127" name="self.model.layers.5.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1128" name="Constant_181791" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1129" name="Reshape_181792" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1130" name="__module.model.layers.5.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1131" name="__module.model.layers.5.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1179">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1132" name="__module.model.layers.5.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1180,a.35">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1133" name="self.model.layers.5.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="450822360" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1134" name="Convert_181804" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1135" name="self.model.layers.5.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="459210968" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1136" name="Convert_181806" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1137" name="self.model.layers.5.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1138" name="self.model.layers.5.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="459342040" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1139" name="self.model.layers.5.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1140" name="Constant_181811" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1141" name="Reshape_181812" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1142" name="__module.model.layers.5.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1143" name="__module.model.layers.5.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1182,hidden_states.177">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1144" name="__module.model.layers.5/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1183,1185,1193,1204,hidden_states.179,hidden_states.181">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1145" name="Constant_157478" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="459866328" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1146" name="Constant_157477" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1147" name="Constant_157475" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1148" name="__module.model.layers.6.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1194">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1149" name="Constant_134244" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1195">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1150" name="__module.model.layers.6.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1196,variance.25">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1151" name="Constant_157476" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1152" name="__module.model.layers.6.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1197">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1153" name="__module.model.layers.6.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1154" name="__module.model.layers.6.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1198">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1155" name="__module.model.layers.6.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1199,1200,hidden_states.183,hidden_states.185">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1156" name="__module.model.layers.6.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1201,1203,hidden_states.187">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1157" name="self.model.layers.6.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="459874520" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1158" name="Convert_182234" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1159" name="self.model.layers.6.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="461971672" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1160" name="Convert_182236" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1161" name="self.model.layers.6.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1162" name="self.model.layers.6.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="462004440" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1163" name="self.model.layers.6.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1164" name="Constant_182241" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1165" name="Reshape_182242" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1166" name="__module.model.layers.6.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1167" name="__module.model.layers.6.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1212,query_states.13">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1168" name="Constant_162298" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1169" name="__module.model.layers.6.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1218">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1170" name="__module.model.layers.6.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1171" name="__module.model.layers.6.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1219,q.13">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1172" name="__module.model.layers.6.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1228">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1173" name="Constant_134641" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1174" name="Constant_134643" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1175" name="Constant_134645" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1176" name="__module.model.layers.6.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1238,x2.25">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1177" name="Constant_157479" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1178" name="__module.model.layers.6.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1239">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1179" name="Constant_134538" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1180" name="Constant_134582" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1181" name="Constant_134584" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1182" name="__module.model.layers.6.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1233">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1183" name="__module.model.layers.6.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1241">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1184" name="__module.model.layers.6.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1242">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1185" name="__module.model.layers.6.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1243">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1186" name="ReadValue_164011" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.6.keypresent.6.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.6.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1187" name="Constant_163554" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="1188" name="Gather_163555" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1189" name="self.model.layers.6.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="462135512" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1190" name="Convert_182564" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1191" name="self.model.layers.6.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="462659800" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1192" name="Convert_182566" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1193" name="self.model.layers.6.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1194" name="self.model.layers.6.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="462667992" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1195" name="self.model.layers.6.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1196" name="Constant_182571" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1197" name="Reshape_182572" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1198" name="__module.model.layers.6.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1199" name="__module.model.layers.6.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1214,key_states.29">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1200" name="Constant_162299" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1201" name="__module.model.layers.6.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1221">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1202" name="__module.model.layers.6.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1203" name="__module.model.layers.6.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1222,k.13">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1204" name="__module.model.layers.6.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1244">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1205" name="Constant_134784" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1206" name="Constant_134786" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1207" name="Constant_134788" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1208" name="__module.model.layers.6.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1254,x2.27">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1209" name="Constant_157480" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1210" name="__module.model.layers.6.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1255">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1211" name="Constant_134681" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1212" name="Constant_134725" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1213" name="Constant_134727" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1214" name="__module.model.layers.6.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1249">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1215" name="__module.model.layers.6.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1257">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1216" name="__module.model.layers.6.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1258">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1217" name="__module.model.layers.6.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1259">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1218" name="__module.model.layers.6.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.6.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1219" name="__module.model.layers.6.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1271,1272,1273">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1220" name="__module.model.layers.6.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1275,hidden_states.191">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1221" name="Constant_162300" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1222" name="__module.model.layers.6.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1279,key_states.31">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1223" name="ReadValue_164013" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.6.valuepresent.6.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.6.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1224" name="Constant_163557" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="1225" name="Gather_163558" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1226" name="self.model.layers.6.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="462700760" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1227" name="Convert_182554" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1228" name="self.model.layers.6.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="463225048" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1229" name="Convert_182556" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1230" name="self.model.layers.6.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1231" name="self.model.layers.6.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="463233240" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1232" name="self.model.layers.6.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1233" name="Constant_182561" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1234" name="Reshape_182562" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1235" name="__module.model.layers.6.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1236" name="__module.model.layers.6.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1216,value_states.13">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1237" name="Constant_162301" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1238" name="__module.model.layers.6.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1224">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1239" name="__module.model.layers.6.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1240" name="__module.model.layers.6.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1225">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1241" name="__module.model.layers.6.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.6.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1242" name="__module.model.layers.6.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1287,1288,1289">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1243" name="__module.model.layers.6.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1291,hidden_states.195">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1244" name="Constant_162302" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1245" name="__module.model.layers.6.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1295">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1246" name="__module.model.layers.6.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="1301,attn_output.25">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1247" name="__module.model.layers.6.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1248" name="__module.model.layers.6.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1302">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1249" name="__module.model.layers.6.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1305,a.37">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1250" name="self.model.layers.6.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="463266008" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1251" name="Convert_182244" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1252" name="self.model.layers.6.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="465363160" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1253" name="Convert_182246" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1254" name="self.model.layers.6.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1255" name="self.model.layers.6.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="465395928" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1256" name="self.model.layers.6.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1257" name="Constant_182251" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1258" name="Reshape_182252" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1259" name="__module.model.layers.6.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1260" name="__module.model.layers.6.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1307,1309,hidden_states.197">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1261" name="__module.model.layers.6/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1312,1314,1325,hidden_states.199,hidden_states.201">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1262" name="Constant_157484" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="465527000" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1263" name="Constant_157483" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1264" name="Constant_157481" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1265" name="__module.model.layers.6.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1315">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1266" name="Constant_135143" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1316">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1267" name="__module.model.layers.6.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1317,variance.27">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1268" name="Constant_157482" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1269" name="__module.model.layers.6.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1318">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1270" name="__module.model.layers.6.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1271" name="__module.model.layers.6.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1319">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1272" name="__module.model.layers.6.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1320,1321,hidden_states.203,hidden_states.205">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1273" name="__module.model.layers.6.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1322,1324,a.39">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1274" name="self.model.layers.6.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="465535192" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1275" name="Convert_181824" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1276" name="self.model.layers.6.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="473923800" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1277" name="Convert_181826" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1278" name="self.model.layers.6.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1279" name="self.model.layers.6.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="474054872" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1280" name="self.model.layers.6.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1281" name="Constant_181831" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1282" name="Reshape_181832" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1283" name="__module.model.layers.6.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1284" name="__module.model.layers.6.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1330,input.13">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1285" name="__module.model.layers.6.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="1331">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1286" name="self.model.layers.6.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="474579160" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1287" name="Convert_181814" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1288" name="self.model.layers.6.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="482967768" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1289" name="Convert_181816" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1290" name="self.model.layers.6.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1291" name="self.model.layers.6.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="483098840" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1292" name="self.model.layers.6.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1293" name="Constant_181821" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1294" name="Reshape_181822" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1295" name="__module.model.layers.6.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1296" name="__module.model.layers.6.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1333">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1297" name="__module.model.layers.6.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1334,a.41">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1298" name="self.model.layers.6.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="483623128" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1299" name="Convert_181834" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1300" name="self.model.layers.6.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="492011736" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1301" name="Convert_181836" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1302" name="self.model.layers.6.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1303" name="self.model.layers.6.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="492142808" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1304" name="self.model.layers.6.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1305" name="Constant_181841" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1306" name="Reshape_181842" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1307" name="__module.model.layers.6.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1308" name="__module.model.layers.6.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1336,hidden_states.207">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1309" name="__module.model.layers.6/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1337,1339,1347,1358,hidden_states.209,hidden_states.211">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1310" name="Constant_157488" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="492667096" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1311" name="Constant_157487" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1312" name="Constant_157485" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1313" name="__module.model.layers.7.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1348">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1314" name="Constant_135217" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1349">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1315" name="__module.model.layers.7.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1350,variance.29">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1316" name="Constant_157486" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1317" name="__module.model.layers.7.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1351">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1318" name="__module.model.layers.7.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1319" name="__module.model.layers.7.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1352">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1320" name="__module.model.layers.7.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1353,1354,hidden_states.213,hidden_states.215">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1321" name="__module.model.layers.7.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1355,1357,hidden_states.217">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1322" name="self.model.layers.7.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="492675288" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1323" name="Convert_182254" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1324" name="self.model.layers.7.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="494772440" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1325" name="Convert_182256" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1326" name="self.model.layers.7.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1327" name="self.model.layers.7.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="494805208" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1328" name="self.model.layers.7.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1329" name="Constant_182261" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1330" name="Reshape_182262" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1331" name="__module.model.layers.7.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1332" name="__module.model.layers.7.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1366,query_states.15">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1333" name="Constant_162303" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1334" name="__module.model.layers.7.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1372">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1335" name="__module.model.layers.7.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1336" name="__module.model.layers.7.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1373,q.15">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1337" name="__module.model.layers.7.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1382">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1338" name="Constant_135614" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1339" name="Constant_135616" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1340" name="Constant_135618" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1341" name="__module.model.layers.7.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1392,x2.29">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1342" name="Constant_157489" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1343" name="__module.model.layers.7.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1393">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1344" name="Constant_135511" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1345" name="Constant_135555" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1346" name="Constant_135557" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1347" name="__module.model.layers.7.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1387">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1348" name="__module.model.layers.7.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1395">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1349" name="__module.model.layers.7.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1396">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1350" name="__module.model.layers.7.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1397">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1351" name="ReadValue_164015" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.7.keypresent.7.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.7.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1352" name="Constant_163560" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="1353" name="Gather_163561" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1354" name="self.model.layers.7.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="494936280" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1355" name="Convert_182584" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1356" name="self.model.layers.7.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="495460568" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1357" name="Convert_182586" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1358" name="self.model.layers.7.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1359" name="self.model.layers.7.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="495468760" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1360" name="self.model.layers.7.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1361" name="Constant_182591" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1362" name="Reshape_182592" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1363" name="__module.model.layers.7.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1364" name="__module.model.layers.7.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1368,key_states.33">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1365" name="Constant_162304" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1366" name="__module.model.layers.7.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1375">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1367" name="__module.model.layers.7.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1368" name="__module.model.layers.7.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1376,k.15">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1369" name="__module.model.layers.7.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1398">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1370" name="Constant_135757" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1371" name="Constant_135759" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1372" name="Constant_135761" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1373" name="__module.model.layers.7.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1408,x2.31">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1374" name="Constant_157490" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1375" name="__module.model.layers.7.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1409">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1376" name="Constant_135654" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1377" name="Constant_135698" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1378" name="Constant_135700" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1379" name="__module.model.layers.7.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1403">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1380" name="__module.model.layers.7.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1411">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1381" name="__module.model.layers.7.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1412">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1382" name="__module.model.layers.7.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1413">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1383" name="__module.model.layers.7.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.7.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1384" name="__module.model.layers.7.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1425,1426,1427">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1385" name="__module.model.layers.7.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1429,hidden_states.221">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1386" name="Constant_162305" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1387" name="__module.model.layers.7.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1433,key_states.35">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1388" name="ReadValue_164017" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.7.valuepresent.7.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.7.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1389" name="Constant_163563" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="1390" name="Gather_163564" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1391" name="self.model.layers.7.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="495501528" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1392" name="Convert_182574" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1393" name="self.model.layers.7.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="496025816" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1394" name="Convert_182576" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1395" name="self.model.layers.7.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1396" name="self.model.layers.7.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="496034008" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1397" name="self.model.layers.7.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1398" name="Constant_182581" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1399" name="Reshape_182582" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1400" name="__module.model.layers.7.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1401" name="__module.model.layers.7.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1370,value_states.15">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1402" name="Constant_162306" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1403" name="__module.model.layers.7.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1378">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1404" name="__module.model.layers.7.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1405" name="__module.model.layers.7.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1379">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1406" name="__module.model.layers.7.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.7.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1407" name="__module.model.layers.7.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1441,1442,1443">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1408" name="__module.model.layers.7.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1445,hidden_states.225">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1409" name="Constant_162307" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1410" name="__module.model.layers.7.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1449">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1411" name="__module.model.layers.7.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="1455,attn_output.29">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1412" name="__module.model.layers.7.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1413" name="__module.model.layers.7.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1456">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1414" name="__module.model.layers.7.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1459,a.43">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1415" name="self.model.layers.7.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="496066776" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1416" name="Convert_182264" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1417" name="self.model.layers.7.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="498163928" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1418" name="Convert_182266" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1419" name="self.model.layers.7.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1420" name="self.model.layers.7.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="498196696" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1421" name="self.model.layers.7.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1422" name="Constant_182271" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1423" name="Reshape_182272" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1424" name="__module.model.layers.7.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1425" name="__module.model.layers.7.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1461,1463,hidden_states.227">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1426" name="__module.model.layers.7/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1466,1468,1479,hidden_states.229,hidden_states.231">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1427" name="Constant_157494" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="498327768" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1428" name="Constant_157493" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1429" name="Constant_157491" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1430" name="__module.model.layers.7.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1469">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1431" name="Constant_136116" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1470">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1432" name="__module.model.layers.7.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1471,variance.31">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1433" name="Constant_157492" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1434" name="__module.model.layers.7.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1472">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1435" name="__module.model.layers.7.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1436" name="__module.model.layers.7.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1473">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1437" name="__module.model.layers.7.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1474,1475,hidden_states.233,hidden_states.235">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1438" name="__module.model.layers.7.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1476,1478,a.45">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1439" name="self.model.layers.7.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="498335960" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1440" name="Convert_181854" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1441" name="self.model.layers.7.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="506724568" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1442" name="Convert_181856" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1443" name="self.model.layers.7.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1444" name="self.model.layers.7.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="506855640" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1445" name="self.model.layers.7.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1446" name="Constant_181861" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1447" name="Reshape_181862" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1448" name="__module.model.layers.7.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1449" name="__module.model.layers.7.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1484,input.15">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1450" name="__module.model.layers.7.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="1485">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1451" name="self.model.layers.7.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="507379928" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1452" name="Convert_181844" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1453" name="self.model.layers.7.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="515768536" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1454" name="Convert_181846" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1455" name="self.model.layers.7.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1456" name="self.model.layers.7.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="515899608" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1457" name="self.model.layers.7.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1458" name="Constant_181851" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1459" name="Reshape_181852" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1460" name="__module.model.layers.7.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1461" name="__module.model.layers.7.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1487">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1462" name="__module.model.layers.7.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1488,a.47">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1463" name="self.model.layers.7.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="516423896" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1464" name="Convert_181864" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1465" name="self.model.layers.7.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="524812504" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1466" name="Convert_181866" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1467" name="self.model.layers.7.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1468" name="self.model.layers.7.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="524943576" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1469" name="self.model.layers.7.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1470" name="Constant_181871" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1471" name="Reshape_181872" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1472" name="__module.model.layers.7.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1473" name="__module.model.layers.7.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1490,hidden_states.237">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1474" name="__module.model.layers.7/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1491,1493,1501,1512,hidden_states.239,hidden_states.241">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1475" name="Constant_157498" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="525467864" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1476" name="Constant_157497" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1477" name="Constant_157495" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1478" name="__module.model.layers.8.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1502">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1479" name="Constant_136190" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1503">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1480" name="__module.model.layers.8.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1504,variance.33">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1481" name="Constant_157496" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1482" name="__module.model.layers.8.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1505">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1483" name="__module.model.layers.8.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1484" name="__module.model.layers.8.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1506">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1485" name="__module.model.layers.8.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1507,1508,hidden_states.243,hidden_states.245">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1486" name="__module.model.layers.8.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1509,1511,hidden_states.247">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1487" name="self.model.layers.8.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="525476056" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1488" name="Convert_182274" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1489" name="self.model.layers.8.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="527573208" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1490" name="Convert_182276" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1491" name="self.model.layers.8.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1492" name="self.model.layers.8.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="527605976" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1493" name="self.model.layers.8.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1494" name="Constant_182281" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1495" name="Reshape_182282" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1496" name="__module.model.layers.8.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1497" name="__module.model.layers.8.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1520,query_states.17">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1498" name="Constant_162308" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1499" name="__module.model.layers.8.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1526">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1500" name="__module.model.layers.8.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1501" name="__module.model.layers.8.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1527,q.17">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1502" name="__module.model.layers.8.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1536">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1503" name="Constant_136587" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1504" name="Constant_136589" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1505" name="Constant_136591" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1506" name="__module.model.layers.8.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1546,x2.33">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1507" name="Constant_157499" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1508" name="__module.model.layers.8.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1547">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1509" name="Constant_136484" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1510" name="Constant_136528" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1511" name="Constant_136530" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1512" name="__module.model.layers.8.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1541">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1513" name="__module.model.layers.8.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1549">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1514" name="__module.model.layers.8.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1550">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1515" name="__module.model.layers.8.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1551">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1516" name="ReadValue_164019" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.8.keypresent.8.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.8.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1517" name="Constant_163566" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="1518" name="Gather_163567" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1519" name="self.model.layers.8.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="527737048" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1520" name="Convert_182604" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1521" name="self.model.layers.8.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="528261336" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1522" name="Convert_182606" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1523" name="self.model.layers.8.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1524" name="self.model.layers.8.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="528269528" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1525" name="self.model.layers.8.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1526" name="Constant_182611" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1527" name="Reshape_182612" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1528" name="__module.model.layers.8.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1529" name="__module.model.layers.8.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1522,key_states.37">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1530" name="Constant_162309" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1531" name="__module.model.layers.8.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1529">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1532" name="__module.model.layers.8.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1533" name="__module.model.layers.8.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1530,k.17">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1534" name="__module.model.layers.8.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1552">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1535" name="Constant_136730" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1536" name="Constant_136732" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1537" name="Constant_136734" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1538" name="__module.model.layers.8.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1562,x2.35">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1539" name="Constant_157500" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1540" name="__module.model.layers.8.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1563">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1541" name="Constant_136627" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1542" name="Constant_136671" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1543" name="Constant_136673" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1544" name="__module.model.layers.8.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1557">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1545" name="__module.model.layers.8.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1565">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1546" name="__module.model.layers.8.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1566">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1547" name="__module.model.layers.8.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1567">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1548" name="__module.model.layers.8.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.8.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1549" name="__module.model.layers.8.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1579,1580,1581">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1550" name="__module.model.layers.8.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1583,hidden_states.251">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1551" name="Constant_162310" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1552" name="__module.model.layers.8.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1587,key_states.39">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1553" name="ReadValue_164021" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.8.valuepresent.8.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.8.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1554" name="Constant_163569" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="1555" name="Gather_163570" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1556" name="self.model.layers.8.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="528302296" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1557" name="Convert_182594" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1558" name="self.model.layers.8.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="528826584" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1559" name="Convert_182596" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1560" name="self.model.layers.8.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1561" name="self.model.layers.8.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="528834776" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1562" name="self.model.layers.8.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1563" name="Constant_182601" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1564" name="Reshape_182602" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1565" name="__module.model.layers.8.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1566" name="__module.model.layers.8.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1524,value_states.17">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1567" name="Constant_162311" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1568" name="__module.model.layers.8.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1532">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1569" name="__module.model.layers.8.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1570" name="__module.model.layers.8.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1533">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1571" name="__module.model.layers.8.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.8.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1572" name="__module.model.layers.8.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1595,1596,1597">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1573" name="__module.model.layers.8.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1599,hidden_states.255">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1574" name="Constant_162312" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1575" name="__module.model.layers.8.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1603">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1576" name="__module.model.layers.8.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="1609,attn_output.33">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1577" name="__module.model.layers.8.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1578" name="__module.model.layers.8.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1610">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1579" name="__module.model.layers.8.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1613,a.49">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1580" name="self.model.layers.8.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="528867544" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1581" name="Convert_182284" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1582" name="self.model.layers.8.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="530964696" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1583" name="Convert_182286" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1584" name="self.model.layers.8.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1585" name="self.model.layers.8.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="530997464" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1586" name="self.model.layers.8.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1587" name="Constant_182291" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1588" name="Reshape_182292" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1589" name="__module.model.layers.8.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1590" name="__module.model.layers.8.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1615,1617,hidden_states.257">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1591" name="__module.model.layers.8/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1620,1622,1633,hidden_states.259,hidden_states.261">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1592" name="Constant_157504" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="531128536" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1593" name="Constant_157503" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1594" name="Constant_157501" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1595" name="__module.model.layers.8.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1623">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1596" name="Constant_137089" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1624">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1597" name="__module.model.layers.8.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1625,variance.35">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1598" name="Constant_157502" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1599" name="__module.model.layers.8.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1626">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1600" name="__module.model.layers.8.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1601" name="__module.model.layers.8.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1627">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1602" name="__module.model.layers.8.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1628,1629,hidden_states.263,hidden_states.265">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1603" name="__module.model.layers.8.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1630,1632,a.51">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1604" name="self.model.layers.8.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="531136728" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1605" name="Convert_181884" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1606" name="self.model.layers.8.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="539525336" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1607" name="Convert_181886" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1608" name="self.model.layers.8.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1609" name="self.model.layers.8.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="539656408" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1610" name="self.model.layers.8.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1611" name="Constant_181891" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1612" name="Reshape_181892" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1613" name="__module.model.layers.8.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1614" name="__module.model.layers.8.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1638,input.17">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1615" name="__module.model.layers.8.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="1639">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1616" name="self.model.layers.8.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="540180696" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1617" name="Convert_181874" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1618" name="self.model.layers.8.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="548569304" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1619" name="Convert_181876" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1620" name="self.model.layers.8.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1621" name="self.model.layers.8.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="548700376" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1622" name="self.model.layers.8.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1623" name="Constant_181881" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1624" name="Reshape_181882" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1625" name="__module.model.layers.8.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1626" name="__module.model.layers.8.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1641">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1627" name="__module.model.layers.8.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1642,a.53">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1628" name="self.model.layers.8.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="549224664" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1629" name="Convert_181894" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1630" name="self.model.layers.8.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="557613272" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1631" name="Convert_181896" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1632" name="self.model.layers.8.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1633" name="self.model.layers.8.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="557744344" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1634" name="self.model.layers.8.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1635" name="Constant_181901" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1636" name="Reshape_181902" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1637" name="__module.model.layers.8.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1638" name="__module.model.layers.8.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1644,hidden_states.267">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1639" name="__module.model.layers.8/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1645,1647,1655,1666,hidden_states.269,hidden_states.271">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1640" name="Constant_157507" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1641" name="Constant_157505" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1642" name="__module.model.layers.9.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1656">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1643" name="Constant_137163" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1657">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1644" name="__module.model.layers.9.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1658,variance.37">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1645" name="Constant_157506" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1646" name="__module.model.layers.9.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1659">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1647" name="__module.model.layers.9.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1648" name="__module.model.layers.9.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1660">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1649" name="__module.model.layers.9.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1661,1662,hidden_states.273,hidden_states.275">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1650" name="__module.model.layers.9.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1663,1665,hidden_states.277">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1651" name="self.model.layers.9.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="558268632" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1652" name="Convert_182614" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1653" name="self.model.layers.9.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="558792920" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1654" name="Convert_182616" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1655" name="self.model.layers.9.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1656" name="self.model.layers.9.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="558801112" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1657" name="self.model.layers.9.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1658" name="Constant_182621" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1659" name="Reshape_182622" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1660" name="__module.model.layers.9.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1661" name="__module.model.layers.9.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1678,value_states.19">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1662" name="Constant_162316" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1663" name="__module.model.layers.9.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1686">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1664" name="__module.model.layers.9.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1665" name="__module.model.layers.9.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1687">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1666" name="__module.model.layers.9.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.9.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1668" name="ReadValue_164023" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.9.keypresent.9.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.9.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1669" name="Constant_163572" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="1670" name="Gather_163573" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1671" name="self.model.layers.9.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="558833880" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1672" name="Convert_182624" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1673" name="self.model.layers.9.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="559358168" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1674" name="Convert_182626" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1675" name="self.model.layers.9.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1676" name="self.model.layers.9.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="559366360" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1677" name="self.model.layers.9.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1678" name="Constant_182631" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1679" name="Reshape_182632" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1680" name="__module.model.layers.9.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1681" name="__module.model.layers.9.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1676,key_states.41">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1682" name="Constant_162314" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1683" name="__module.model.layers.9.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1683">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1684" name="__module.model.layers.9.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1685" name="__module.model.layers.9.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1684,k.19">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1686" name="__module.model.layers.9.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1706">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1687" name="Constant_137703" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1688" name="Constant_137705" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1689" name="Constant_137707" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1690" name="__module.model.layers.9.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1716,x2.39">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1691" name="Constant_157510" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1692" name="__module.model.layers.9.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1717">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1693" name="Constant_137600" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1694" name="Constant_137644" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1695" name="Constant_137646" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1696" name="__module.model.layers.9.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1711">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1697" name="__module.model.layers.9.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1719">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1698" name="__module.model.layers.9.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1720">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1699" name="__module.model.layers.9.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1721">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1700" name="__module.model.layers.9.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.9.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1716" name="ReadValue_163993" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.15.valuepresent.15.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.15.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1717" name="Constant_163611" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="1718" name="Gather_163612" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1719" name="Constant_157568" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="559399128" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1720" name="self.model.layers.9.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="559407320" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1721" name="Convert_182294" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1722" name="self.model.layers.9.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="561504472" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1723" name="Convert_182296" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1724" name="self.model.layers.9.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1725" name="self.model.layers.9.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="561537240" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1726" name="self.model.layers.9.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1727" name="Constant_182301" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1728" name="Reshape_182302" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1729" name="__module.model.layers.9.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1730" name="__module.model.layers.9.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1674,query_states.19">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1731" name="Constant_162313" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1732" name="__module.model.layers.9.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1680">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1733" name="__module.model.layers.9.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1734" name="__module.model.layers.9.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1681,q.19">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1735" name="__module.model.layers.9.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1690">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1736" name="Constant_137560" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1737" name="Constant_137562" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1738" name="Constant_137564" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1739" name="__module.model.layers.9.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1700,x2.37">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1740" name="Constant_157509" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1741" name="__module.model.layers.9.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1701">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1742" name="Constant_137457" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1743" name="Constant_137501" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1744" name="Constant_137503" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1745" name="__module.model.layers.9.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1695">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1746" name="__module.model.layers.9.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1703">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1747" name="__module.model.layers.9.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1704">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1748" name="__module.model.layers.9.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1705">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1749" name="__module.model.layers.9.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1733,1734,1735">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1750" name="__module.model.layers.9.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1737,hidden_states.281">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1751" name="Constant_162315" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1752" name="__module.model.layers.9.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1741,key_states.43">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1753" name="__module.model.layers.9.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1749,1750,1751">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1754" name="__module.model.layers.9.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1753,hidden_states.285">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1755" name="Constant_162317" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1756" name="__module.model.layers.9.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1757">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1757" name="__module.model.layers.9.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="1763,attn_output.37">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1758" name="__module.model.layers.9.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1759" name="__module.model.layers.9.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1764">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1760" name="__module.model.layers.9.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1767,a.55">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1761" name="self.model.layers.9.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="561668312" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1762" name="Convert_182304" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1763" name="self.model.layers.9.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="563765464" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1764" name="Convert_182306" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1765" name="self.model.layers.9.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1766" name="self.model.layers.9.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="563798232" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1767" name="self.model.layers.9.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1768" name="Constant_182311" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1769" name="Reshape_182312" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1770" name="__module.model.layers.9.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1771" name="__module.model.layers.9.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1769,1771,hidden_states.287">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1772" name="__module.model.layers.9/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1774,1776,1787,hidden_states.289,hidden_states.291">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1773" name="Constant_157514" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="563929304" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1774" name="Constant_157513" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1775" name="Constant_157511" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1776" name="__module.model.layers.9.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1777">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1777" name="Constant_138062" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1778">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1778" name="__module.model.layers.9.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1779,variance.39">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1779" name="Constant_157512" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1780" name="__module.model.layers.9.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1780">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1781" name="__module.model.layers.9.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1782" name="__module.model.layers.9.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1781">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1783" name="__module.model.layers.9.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1782,1783,hidden_states.293,hidden_states.295">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1784" name="__module.model.layers.9.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1784,1786,a.57">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1785" name="self.model.layers.9.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="563937496" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1786" name="Convert_181914" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1787" name="self.model.layers.9.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="572326104" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1788" name="Convert_181916" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1789" name="self.model.layers.9.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1790" name="self.model.layers.9.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="572457176" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1791" name="self.model.layers.9.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1792" name="Constant_181921" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1793" name="Reshape_181922" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1794" name="__module.model.layers.9.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1795" name="__module.model.layers.9.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1792,input.19">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1796" name="__module.model.layers.9.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="1793">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1797" name="self.model.layers.9.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="572981464" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1798" name="Convert_181904" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1799" name="self.model.layers.9.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="581370072" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1800" name="Convert_181906" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1801" name="self.model.layers.9.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1802" name="self.model.layers.9.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="581501144" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1803" name="self.model.layers.9.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1804" name="Constant_181911" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1805" name="Reshape_181912" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1806" name="__module.model.layers.9.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1807" name="__module.model.layers.9.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1795">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1808" name="__module.model.layers.9.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1796,a.59">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1809" name="self.model.layers.9.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="582025432" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1810" name="Convert_181924" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1811" name="self.model.layers.9.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="590414040" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1812" name="Convert_181926" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1813" name="self.model.layers.9.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1814" name="self.model.layers.9.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="590545112" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1815" name="self.model.layers.9.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1816" name="Constant_181931" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1817" name="Reshape_181932" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1818" name="__module.model.layers.9.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1819" name="__module.model.layers.9.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1798,hidden_states.297">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1820" name="__module.model.layers.9/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1799,1801,1809,1820,hidden_states.299,hidden_states.301">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1821" name="Constant_157518" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="591069400" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1822" name="Constant_157517" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1823" name="Constant_157515" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1824" name="__module.model.layers.10.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1810">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1825" name="Constant_138136" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1811">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1826" name="__module.model.layers.10.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1812,variance.41">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1827" name="Constant_157516" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1828" name="__module.model.layers.10.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1813">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1829" name="__module.model.layers.10.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1830" name="__module.model.layers.10.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1814">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1831" name="__module.model.layers.10.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1815,1816,hidden_states.303,hidden_states.305">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1832" name="__module.model.layers.10.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1817,1819,hidden_states.307">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1833" name="self.model.layers.10.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="591077592" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1834" name="Convert_182314" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1835" name="self.model.layers.10.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="593174744" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1836" name="Convert_182316" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1837" name="self.model.layers.10.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1838" name="self.model.layers.10.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="593207512" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1839" name="self.model.layers.10.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1840" name="Constant_182321" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1841" name="Reshape_182322" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1842" name="__module.model.layers.10.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1843" name="__module.model.layers.10.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1828,query_states.21">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1844" name="Constant_162318" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1845" name="__module.model.layers.10.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1834">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1846" name="__module.model.layers.10.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1847" name="__module.model.layers.10.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1835,q.21">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1848" name="__module.model.layers.10.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1844">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1849" name="Constant_138533" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1850" name="Constant_138535" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1851" name="Constant_138537" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1852" name="__module.model.layers.10.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1854,x2.41">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1853" name="Constant_157519" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1854" name="__module.model.layers.10.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1855">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1855" name="Constant_138430" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1856" name="Constant_138474" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1857" name="Constant_138476" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1858" name="__module.model.layers.10.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1849">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1859" name="__module.model.layers.10.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1857">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1860" name="__module.model.layers.10.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1858">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1861" name="__module.model.layers.10.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1859">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1862" name="ReadValue_163971" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.10.keypresent.10.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.10.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1863" name="Constant_163578" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="1864" name="Gather_163579" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1865" name="self.model.layers.10.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="593338584" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1866" name="Convert_182644" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1867" name="self.model.layers.10.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="593862872" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1868" name="Convert_182646" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1869" name="self.model.layers.10.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1870" name="self.model.layers.10.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="593871064" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1871" name="self.model.layers.10.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1872" name="Constant_182651" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1873" name="Reshape_182652" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1874" name="__module.model.layers.10.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1875" name="__module.model.layers.10.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1830,key_states.45">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1876" name="Constant_162319" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1877" name="__module.model.layers.10.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1837">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1878" name="__module.model.layers.10.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1879" name="__module.model.layers.10.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1838,k.21">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1880" name="__module.model.layers.10.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1860">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1881" name="Constant_138676" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1882" name="Constant_138678" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1883" name="Constant_138680" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1884" name="__module.model.layers.10.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1870,x2.43">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1885" name="Constant_157520" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1886" name="__module.model.layers.10.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1871">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1887" name="Constant_138573" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1888" name="Constant_138617" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1889" name="Constant_138619" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1890" name="__module.model.layers.10.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="1865">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1891" name="__module.model.layers.10.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1873">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1892" name="__module.model.layers.10.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1874">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1893" name="__module.model.layers.10.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1875">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1894" name="__module.model.layers.10.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.10.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1895" name="__module.model.layers.10.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1887,1888,1889">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1896" name="__module.model.layers.10.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1891,hidden_states.311">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1897" name="Constant_162320" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1898" name="__module.model.layers.10.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1895,key_states.47">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1899" name="ReadValue_163973" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.10.valuepresent.10.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.10.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1900" name="Constant_163581" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="1901" name="Gather_163582" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1902" name="self.model.layers.10.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="593903832" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1903" name="Convert_182634" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1904" name="self.model.layers.10.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="594428120" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1905" name="Convert_182636" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1906" name="self.model.layers.10.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1907" name="self.model.layers.10.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="594436312" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1908" name="self.model.layers.10.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1909" name="Constant_182641" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1910" name="Reshape_182642" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1911" name="__module.model.layers.10.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1912" name="__module.model.layers.10.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1832,value_states.21">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1913" name="Constant_162321" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1914" name="__module.model.layers.10.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1840">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1915" name="__module.model.layers.10.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1916" name="__module.model.layers.10.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1841">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1917" name="__module.model.layers.10.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.10.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1918" name="__module.model.layers.10.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1903,1904,1905">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1919" name="__module.model.layers.10.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1907,hidden_states.315">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1920" name="Constant_162322" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1921" name="__module.model.layers.10.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1911">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1922" name="__module.model.layers.10.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="1917,attn_output.41">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1923" name="__module.model.layers.10.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1924" name="__module.model.layers.10.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1918">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1925" name="__module.model.layers.10.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1921,a.61">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1926" name="self.model.layers.10.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="594469080" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1927" name="Convert_182324" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1928" name="self.model.layers.10.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="596566232" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1929" name="Convert_182326" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1930" name="self.model.layers.10.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1931" name="self.model.layers.10.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="596599000" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1932" name="self.model.layers.10.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1933" name="Constant_182331" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1934" name="Reshape_182332" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1935" name="__module.model.layers.10.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1936" name="__module.model.layers.10.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1923,1925,hidden_states.317">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1937" name="__module.model.layers.10/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1928,1930,1941,hidden_states.319,hidden_states.321">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1938" name="Constant_157524" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="596730072" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1939" name="Constant_157523" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1940" name="Constant_157521" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1941" name="__module.model.layers.10.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1931">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1942" name="Constant_139035" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1932">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1943" name="__module.model.layers.10.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1933,variance.43">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1944" name="Constant_157522" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1945" name="__module.model.layers.10.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1934">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1946" name="__module.model.layers.10.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1947" name="__module.model.layers.10.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1935">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1948" name="__module.model.layers.10.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1936,1937,hidden_states.323,hidden_states.325">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1949" name="__module.model.layers.10.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1938,1940,a.63">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1950" name="self.model.layers.10.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="596738264" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1951" name="Convert_181944" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1952" name="self.model.layers.10.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="605126872" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1953" name="Convert_181946" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1954" name="self.model.layers.10.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1955" name="self.model.layers.10.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="605257944" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1956" name="self.model.layers.10.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1957" name="Constant_181951" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1958" name="Reshape_181952" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1959" name="__module.model.layers.10.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1960" name="__module.model.layers.10.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1946,input.21">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1961" name="__module.model.layers.10.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="1947">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1962" name="self.model.layers.10.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="605782232" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1963" name="Convert_181934" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1964" name="self.model.layers.10.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="614170840" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1965" name="Convert_181936" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1966" name="self.model.layers.10.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1967" name="self.model.layers.10.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="614301912" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1968" name="self.model.layers.10.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1969" name="Constant_181941" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1970" name="Reshape_181942" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1971" name="__module.model.layers.10.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1972" name="__module.model.layers.10.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1949">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1973" name="__module.model.layers.10.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1950,a.65">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1974" name="self.model.layers.10.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="614826200" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1975" name="Convert_181954" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1976" name="self.model.layers.10.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="623214808" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1977" name="Convert_181956" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1978" name="self.model.layers.10.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1979" name="self.model.layers.10.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="623345880" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1980" name="self.model.layers.10.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1981" name="Constant_181961" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1982" name="Reshape_181962" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1983" name="__module.model.layers.10.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1984" name="__module.model.layers.10.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1952,hidden_states.327">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1985" name="__module.model.layers.10/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1953,1955,1963,1974,hidden_states.329,hidden_states.331">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1986" name="Constant_157528" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="623870168" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1987" name="Constant_157527" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1988" name="Constant_157525" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1989" name="__module.model.layers.11.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1964">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1990" name="Constant_139109" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="1965">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1991" name="__module.model.layers.11.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1966,variance.45">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1992" name="Constant_157526" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1993" name="__module.model.layers.11.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1967">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1994" name="__module.model.layers.11.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1995" name="__module.model.layers.11.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1968">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1996" name="__module.model.layers.11.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1969,1970,hidden_states.333,hidden_states.335">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1997" name="__module.model.layers.11.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1971,1973,hidden_states.337">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1998" name="self.model.layers.11.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="623878360" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1999" name="Convert_182334" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2000" name="self.model.layers.11.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="625975512" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2001" name="Convert_182336" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2002" name="self.model.layers.11.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2003" name="self.model.layers.11.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="626008280" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2004" name="self.model.layers.11.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2005" name="Constant_182341" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2006" name="Reshape_182342" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2007" name="__module.model.layers.11.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2008" name="__module.model.layers.11.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1982,query_states.23">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2009" name="Constant_162323" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2010" name="__module.model.layers.11.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1988">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2011" name="__module.model.layers.11.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2012" name="__module.model.layers.11.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1989,q.23">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2013" name="__module.model.layers.11.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1998">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2014" name="Constant_139506" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2015" name="Constant_139508" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2016" name="Constant_139510" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2017" name="__module.model.layers.11.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2008,x2.45">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2018" name="Constant_157529" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2019" name="__module.model.layers.11.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2009">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2020" name="Constant_139403" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2021" name="Constant_139447" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2022" name="Constant_139449" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2023" name="__module.model.layers.11.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2003">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2024" name="__module.model.layers.11.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2011">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2025" name="__module.model.layers.11.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2012">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2026" name="__module.model.layers.11.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2013">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2027" name="ReadValue_163975" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.11.keypresent.11.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.11.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2028" name="Constant_163584" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="2029" name="Gather_163585" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2030" name="self.model.layers.11.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="626139352" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2031" name="Convert_182664" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2032" name="self.model.layers.11.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="626663640" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2033" name="Convert_182666" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2034" name="self.model.layers.11.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2035" name="self.model.layers.11.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="626671832" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2036" name="self.model.layers.11.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2037" name="Constant_182671" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2038" name="Reshape_182672" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2039" name="__module.model.layers.11.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2040" name="__module.model.layers.11.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1984,key_states.49">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2041" name="Constant_162324" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2042" name="__module.model.layers.11.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1991">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2043" name="__module.model.layers.11.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2044" name="__module.model.layers.11.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1992,k.23">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2045" name="__module.model.layers.11.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2014">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2046" name="Constant_139649" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2047" name="Constant_139651" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2048" name="Constant_139653" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2049" name="__module.model.layers.11.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2024,x2.47">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2050" name="Constant_157530" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2051" name="__module.model.layers.11.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2025">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2052" name="Constant_139546" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2053" name="Constant_139590" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2054" name="Constant_139592" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2055" name="__module.model.layers.11.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2019">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2056" name="__module.model.layers.11.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2027">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2057" name="__module.model.layers.11.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2028">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2058" name="__module.model.layers.11.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2029">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2059" name="__module.model.layers.11.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.11.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2060" name="__module.model.layers.11.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2041,2042,2043">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2061" name="__module.model.layers.11.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2045,hidden_states.341">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2062" name="Constant_162325" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2063" name="__module.model.layers.11.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2049,key_states.51">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2064" name="ReadValue_163977" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.11.valuepresent.11.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.11.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2065" name="Constant_163587" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="2066" name="Gather_163588" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2067" name="self.model.layers.11.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="626704600" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2068" name="Convert_182654" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2069" name="self.model.layers.11.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="627228888" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2070" name="Convert_182656" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2071" name="self.model.layers.11.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2072" name="self.model.layers.11.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="627237080" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2073" name="self.model.layers.11.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2074" name="Constant_182661" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2075" name="Reshape_182662" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2076" name="__module.model.layers.11.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2077" name="__module.model.layers.11.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1986,value_states.23">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2078" name="Constant_162326" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2079" name="__module.model.layers.11.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1994">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2080" name="__module.model.layers.11.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2081" name="__module.model.layers.11.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="1995">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2082" name="__module.model.layers.11.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.11.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2083" name="__module.model.layers.11.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2057,2058,2059">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2084" name="__module.model.layers.11.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2061,hidden_states.345">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2085" name="Constant_162327" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2086" name="__module.model.layers.11.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2065">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2087" name="__module.model.layers.11.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="2071,attn_output.45">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2088" name="__module.model.layers.11.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2089" name="__module.model.layers.11.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2072">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2090" name="__module.model.layers.11.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2075,a.67">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2091" name="self.model.layers.11.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="627269848" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2092" name="Convert_182344" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2093" name="self.model.layers.11.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="629367000" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2094" name="Convert_182346" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2095" name="self.model.layers.11.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2096" name="self.model.layers.11.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="629399768" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2097" name="self.model.layers.11.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2098" name="Constant_182351" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2099" name="Reshape_182352" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2100" name="__module.model.layers.11.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2101" name="__module.model.layers.11.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2077,2079,hidden_states.347">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2102" name="__module.model.layers.11/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2082,2084,2095,hidden_states.349,hidden_states.351">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2103" name="Constant_157534" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="629530840" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2104" name="Constant_157533" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2105" name="Constant_157531" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2106" name="__module.model.layers.11.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2085">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2107" name="Constant_140008" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="2086">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2108" name="__module.model.layers.11.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2087,variance.47">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2109" name="Constant_157532" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2110" name="__module.model.layers.11.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2088">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2111" name="__module.model.layers.11.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2112" name="__module.model.layers.11.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2089">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2113" name="__module.model.layers.11.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2090,2091,hidden_states.353,hidden_states.355">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2114" name="__module.model.layers.11.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2092,2094,a.69">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2115" name="self.model.layers.11.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="629539032" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2116" name="Convert_181974" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2117" name="self.model.layers.11.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="637927640" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2118" name="Convert_181976" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2119" name="self.model.layers.11.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2120" name="self.model.layers.11.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="638058712" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2121" name="self.model.layers.11.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2122" name="Constant_181981" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2123" name="Reshape_181982" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2124" name="__module.model.layers.11.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2125" name="__module.model.layers.11.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2100,input.23">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2126" name="__module.model.layers.11.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="2101">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2127" name="self.model.layers.11.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="638583000" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2128" name="Convert_181964" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2129" name="self.model.layers.11.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="646971608" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2130" name="Convert_181966" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2131" name="self.model.layers.11.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2132" name="self.model.layers.11.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="647102680" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2133" name="self.model.layers.11.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2134" name="Constant_181971" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2135" name="Reshape_181972" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2136" name="__module.model.layers.11.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2137" name="__module.model.layers.11.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2103">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2138" name="__module.model.layers.11.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2104,a.71">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2139" name="self.model.layers.11.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="647626968" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2140" name="Convert_181984" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2141" name="self.model.layers.11.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="656015576" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2142" name="Convert_181986" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2143" name="self.model.layers.11.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2144" name="self.model.layers.11.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="656146648" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2145" name="self.model.layers.11.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2146" name="Constant_181991" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2147" name="Reshape_181992" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2148" name="__module.model.layers.11.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2149" name="__module.model.layers.11.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2106,hidden_states.357">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2150" name="__module.model.layers.11/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2107,2109,2117,2128,hidden_states.359,hidden_states.361">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2151" name="Constant_157538" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="656670936" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2152" name="Constant_157537" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2153" name="Constant_157535" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2154" name="__module.model.layers.12.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2118">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2155" name="Constant_140082" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="2119">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2156" name="__module.model.layers.12.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2120,variance.49">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2157" name="Constant_157536" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2158" name="__module.model.layers.12.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2121">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2159" name="__module.model.layers.12.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2160" name="__module.model.layers.12.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2122">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2161" name="__module.model.layers.12.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2123,2124,hidden_states.363,hidden_states.365">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2162" name="__module.model.layers.12.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2125,2127,hidden_states.367">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2163" name="self.model.layers.12.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="656679128" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2164" name="Convert_182354" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2165" name="self.model.layers.12.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="658776280" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2166" name="Convert_182356" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2167" name="self.model.layers.12.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2168" name="self.model.layers.12.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="658809048" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2169" name="self.model.layers.12.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2170" name="Constant_182361" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2171" name="Reshape_182362" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2172" name="__module.model.layers.12.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2173" name="__module.model.layers.12.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2136,query_states.25">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2174" name="Constant_162328" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2175" name="__module.model.layers.12.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2142">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2176" name="__module.model.layers.12.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2177" name="__module.model.layers.12.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2143,q.25">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2178" name="__module.model.layers.12.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2152">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2179" name="Constant_140479" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2180" name="Constant_140481" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2181" name="Constant_140483" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2182" name="__module.model.layers.12.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2162,x2.49">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2183" name="Constant_157539" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2184" name="__module.model.layers.12.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2163">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2185" name="Constant_140376" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2186" name="Constant_140420" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2187" name="Constant_140422" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2188" name="__module.model.layers.12.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2157">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2189" name="__module.model.layers.12.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2165">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2190" name="__module.model.layers.12.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2166">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2191" name="__module.model.layers.12.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2167">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2192" name="ReadValue_163979" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.12.keypresent.12.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.12.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2193" name="Constant_163590" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="2194" name="Gather_163591" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2195" name="self.model.layers.12.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="658940120" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2196" name="Convert_182684" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2197" name="self.model.layers.12.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="659464408" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2198" name="Convert_182686" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2199" name="self.model.layers.12.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2200" name="self.model.layers.12.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="659472600" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2201" name="self.model.layers.12.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2202" name="Constant_182691" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2203" name="Reshape_182692" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2204" name="__module.model.layers.12.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2205" name="__module.model.layers.12.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2138,key_states.53">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2206" name="Constant_162329" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2207" name="__module.model.layers.12.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2145">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2208" name="__module.model.layers.12.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2209" name="__module.model.layers.12.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2146,k.25">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2210" name="__module.model.layers.12.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2168">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2211" name="Constant_140622" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2212" name="Constant_140624" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2213" name="Constant_140626" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2214" name="__module.model.layers.12.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2178,x2.51">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2215" name="Constant_157540" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2216" name="__module.model.layers.12.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2179">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2217" name="Constant_140519" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2218" name="Constant_140563" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2219" name="Constant_140565" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2220" name="__module.model.layers.12.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2173">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2221" name="__module.model.layers.12.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2181">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2222" name="__module.model.layers.12.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2182">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2223" name="__module.model.layers.12.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2183">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2224" name="__module.model.layers.12.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.12.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2225" name="__module.model.layers.12.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2195,2196,2197">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2226" name="__module.model.layers.12.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2199,hidden_states.371">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2227" name="Constant_162330" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2228" name="__module.model.layers.12.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2203,key_states.55">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2229" name="ReadValue_163981" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.12.valuepresent.12.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.12.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2230" name="Constant_163593" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="2231" name="Gather_163594" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2232" name="self.model.layers.12.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="659505368" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2233" name="Convert_182674" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2234" name="self.model.layers.12.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="660029656" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2235" name="Convert_182676" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2236" name="self.model.layers.12.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2237" name="self.model.layers.12.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="660037848" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2238" name="self.model.layers.12.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2239" name="Constant_182681" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2240" name="Reshape_182682" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2241" name="__module.model.layers.12.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2242" name="__module.model.layers.12.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2140,value_states.25">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2243" name="Constant_162331" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2244" name="__module.model.layers.12.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2148">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2245" name="__module.model.layers.12.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2246" name="__module.model.layers.12.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2149">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2247" name="__module.model.layers.12.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.12.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2248" name="__module.model.layers.12.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2211,2212,2213">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2249" name="__module.model.layers.12.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2215,hidden_states.375">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2250" name="Constant_162332" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2251" name="__module.model.layers.12.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2219">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2252" name="__module.model.layers.12.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="2225,attn_output.49">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2253" name="__module.model.layers.12.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2254" name="__module.model.layers.12.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2226">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2255" name="__module.model.layers.12.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2229,a.73">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2256" name="self.model.layers.12.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="660070616" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2257" name="Convert_182364" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2258" name="self.model.layers.12.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="662167768" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2259" name="Convert_182366" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2260" name="self.model.layers.12.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2261" name="self.model.layers.12.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="662200536" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2262" name="self.model.layers.12.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2263" name="Constant_182371" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2264" name="Reshape_182372" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2265" name="__module.model.layers.12.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2266" name="__module.model.layers.12.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2231,2233,hidden_states.377">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2267" name="__module.model.layers.12/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2236,2238,2249,hidden_states.379,hidden_states.381">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2268" name="Constant_157544" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="662331608" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2269" name="Constant_157543" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2270" name="Constant_157541" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2271" name="__module.model.layers.12.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2239">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2272" name="Constant_140981" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="2240">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2273" name="__module.model.layers.12.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2241,variance.51">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2274" name="Constant_157542" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2275" name="__module.model.layers.12.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2242">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2276" name="__module.model.layers.12.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2277" name="__module.model.layers.12.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2243">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2278" name="__module.model.layers.12.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2244,2245,hidden_states.383,hidden_states.385">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2279" name="__module.model.layers.12.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2246,2248,a.75">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2280" name="self.model.layers.12.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="662339800" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2281" name="Convert_182004" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2282" name="self.model.layers.12.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="670728408" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2283" name="Convert_182006" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2284" name="self.model.layers.12.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2285" name="self.model.layers.12.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="670859480" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2286" name="self.model.layers.12.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2287" name="Constant_182011" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2288" name="Reshape_182012" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2289" name="__module.model.layers.12.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2290" name="__module.model.layers.12.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2254,input.25">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2291" name="__module.model.layers.12.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="2255">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2292" name="self.model.layers.12.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="671383768" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2293" name="Convert_181994" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2294" name="self.model.layers.12.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="679772376" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2295" name="Convert_181996" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2296" name="self.model.layers.12.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2297" name="self.model.layers.12.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="679903448" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2298" name="self.model.layers.12.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2299" name="Constant_182001" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2300" name="Reshape_182002" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2301" name="__module.model.layers.12.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2302" name="__module.model.layers.12.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2257">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2303" name="__module.model.layers.12.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2258,a.77">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2304" name="self.model.layers.12.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="680427736" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2305" name="Convert_182014" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2306" name="self.model.layers.12.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="688816344" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2307" name="Convert_182016" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2308" name="self.model.layers.12.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2309" name="self.model.layers.12.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="688947416" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2310" name="self.model.layers.12.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2311" name="Constant_182021" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2312" name="Reshape_182022" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2313" name="__module.model.layers.12.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2314" name="__module.model.layers.12.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2260,hidden_states.387">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2315" name="__module.model.layers.12/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2261,2263,2271,2282,hidden_states.389,hidden_states.391">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2316" name="Constant_157548" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="689471704" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2317" name="Constant_157547" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2318" name="Constant_157545" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2319" name="__module.model.layers.13.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2272">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2320" name="Constant_141055" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="2273">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2321" name="__module.model.layers.13.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2274,variance.53">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2322" name="Constant_157546" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2323" name="__module.model.layers.13.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2275">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2324" name="__module.model.layers.13.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2325" name="__module.model.layers.13.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2276">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2326" name="__module.model.layers.13.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2277,2278,hidden_states.393,hidden_states.395">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2327" name="__module.model.layers.13.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2279,2281,hidden_states.397">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2328" name="self.model.layers.13.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="689479896" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2329" name="Convert_182374" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2330" name="self.model.layers.13.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="691577048" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2331" name="Convert_182376" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2332" name="self.model.layers.13.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2333" name="self.model.layers.13.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="691609816" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2334" name="self.model.layers.13.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2335" name="Constant_182381" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2336" name="Reshape_182382" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2337" name="__module.model.layers.13.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2338" name="__module.model.layers.13.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2290,query_states.27">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2339" name="Constant_162333" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2340" name="__module.model.layers.13.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2296">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2341" name="__module.model.layers.13.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2342" name="__module.model.layers.13.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2297,q.27">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2343" name="__module.model.layers.13.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2306">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2344" name="Constant_141452" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2345" name="Constant_141454" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2346" name="Constant_141456" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2347" name="__module.model.layers.13.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2316,x2.53">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2348" name="Constant_157549" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2349" name="__module.model.layers.13.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2317">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2350" name="Constant_141349" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2351" name="Constant_141393" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2352" name="Constant_141395" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2353" name="__module.model.layers.13.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2311">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2354" name="__module.model.layers.13.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2319">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2355" name="__module.model.layers.13.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2320">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2356" name="__module.model.layers.13.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2321">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2357" name="ReadValue_163983" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.13.keypresent.13.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.13.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2358" name="Constant_163596" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="2359" name="Gather_163597" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2360" name="self.model.layers.13.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="691740888" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2361" name="Convert_182704" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2362" name="self.model.layers.13.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="692265176" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2363" name="Convert_182706" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2364" name="self.model.layers.13.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2365" name="self.model.layers.13.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="692273368" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2366" name="self.model.layers.13.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2367" name="Constant_182711" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2368" name="Reshape_182712" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2369" name="__module.model.layers.13.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2370" name="__module.model.layers.13.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2292,key_states.57">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2371" name="Constant_162334" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2372" name="__module.model.layers.13.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2299">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2373" name="__module.model.layers.13.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2374" name="__module.model.layers.13.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2300,k.27">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2375" name="__module.model.layers.13.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2322">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2376" name="Constant_141595" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2377" name="Constant_141597" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2378" name="Constant_141599" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2379" name="__module.model.layers.13.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2332,x2.55">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2380" name="Constant_157550" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2381" name="__module.model.layers.13.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2333">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2382" name="Constant_141492" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2383" name="Constant_141536" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2384" name="Constant_141538" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2385" name="__module.model.layers.13.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2327">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2386" name="__module.model.layers.13.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2335">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2387" name="__module.model.layers.13.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2336">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2388" name="__module.model.layers.13.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2337">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2389" name="__module.model.layers.13.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.13.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2390" name="__module.model.layers.13.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2349,2350,2351">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2391" name="__module.model.layers.13.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2353,hidden_states.401">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2392" name="Constant_162335" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2393" name="__module.model.layers.13.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2357,key_states.59">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2394" name="ReadValue_163985" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.13.valuepresent.13.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.13.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2395" name="Constant_163599" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="2396" name="Gather_163600" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2397" name="self.model.layers.13.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="692306136" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2398" name="Convert_182694" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2399" name="self.model.layers.13.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="692830424" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2400" name="Convert_182696" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2401" name="self.model.layers.13.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2402" name="self.model.layers.13.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="692838616" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2403" name="self.model.layers.13.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2404" name="Constant_182701" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2405" name="Reshape_182702" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2406" name="__module.model.layers.13.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2407" name="__module.model.layers.13.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2294,value_states.27">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2408" name="Constant_162336" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2409" name="__module.model.layers.13.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2302">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2410" name="__module.model.layers.13.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2411" name="__module.model.layers.13.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2303">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2412" name="__module.model.layers.13.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.13.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2413" name="__module.model.layers.13.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2365,2366,2367">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2414" name="__module.model.layers.13.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2369,hidden_states.405">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2415" name="Constant_162337" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2416" name="__module.model.layers.13.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2373">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2417" name="__module.model.layers.13.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="2379,attn_output.53">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2418" name="__module.model.layers.13.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2419" name="__module.model.layers.13.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2380">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2420" name="__module.model.layers.13.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2383,a.79">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2421" name="self.model.layers.13.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="692871384" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2422" name="Convert_182384" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2423" name="self.model.layers.13.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="694968536" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2424" name="Convert_182386" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2425" name="self.model.layers.13.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2426" name="self.model.layers.13.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="695001304" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2427" name="self.model.layers.13.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2428" name="Constant_182391" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2429" name="Reshape_182392" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2430" name="__module.model.layers.13.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2431" name="__module.model.layers.13.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2385,2387,hidden_states.407">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2432" name="__module.model.layers.13/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2390,2392,2403,hidden_states.409,hidden_states.411">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2433" name="Constant_157554" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="695132376" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2434" name="Constant_157553" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2435" name="Constant_157551" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2436" name="__module.model.layers.13.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2393">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2437" name="Constant_141954" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="2394">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2438" name="__module.model.layers.13.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2395,variance.55">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2439" name="Constant_157552" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2440" name="__module.model.layers.13.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2396">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2441" name="__module.model.layers.13.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2442" name="__module.model.layers.13.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2397">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2443" name="__module.model.layers.13.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2398,2399,hidden_states.413,hidden_states.415">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2444" name="__module.model.layers.13.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2400,2402,a.81">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2445" name="self.model.layers.13.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="695140568" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2446" name="Convert_182034" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2447" name="self.model.layers.13.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="703529176" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2448" name="Convert_182036" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2449" name="self.model.layers.13.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2450" name="self.model.layers.13.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="703660248" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2451" name="self.model.layers.13.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2452" name="Constant_182041" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2453" name="Reshape_182042" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2454" name="__module.model.layers.13.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2455" name="__module.model.layers.13.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2408,input.27">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2456" name="__module.model.layers.13.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="2409">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2457" name="self.model.layers.13.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="704184536" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2458" name="Convert_182024" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2459" name="self.model.layers.13.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="712573144" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2460" name="Convert_182026" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2461" name="self.model.layers.13.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2462" name="self.model.layers.13.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="712704216" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2463" name="self.model.layers.13.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2464" name="Constant_182031" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2465" name="Reshape_182032" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2466" name="__module.model.layers.13.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2467" name="__module.model.layers.13.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2411">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2468" name="__module.model.layers.13.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2412,a.83">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2469" name="self.model.layers.13.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="713228504" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2470" name="Convert_182044" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2471" name="self.model.layers.13.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="721617112" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2472" name="Convert_182046" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2473" name="self.model.layers.13.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2474" name="self.model.layers.13.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="721748184" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2475" name="self.model.layers.13.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2476" name="Constant_182051" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2477" name="Reshape_182052" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2478" name="__module.model.layers.13.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2479" name="__module.model.layers.13.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2414,hidden_states.417">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2480" name="__module.model.layers.13/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2415,2417,2425,2436,hidden_states.419,hidden_states.421">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2481" name="Constant_157558" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="722272472" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2482" name="Constant_157557" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2483" name="Constant_157555" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2484" name="__module.model.layers.14.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2426">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2485" name="Constant_142028" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="2427">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2486" name="__module.model.layers.14.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2428,variance.57">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2487" name="Constant_157556" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2488" name="__module.model.layers.14.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2429">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2489" name="__module.model.layers.14.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2490" name="__module.model.layers.14.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2430">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2491" name="__module.model.layers.14.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2431,2432,hidden_states.423,hidden_states.425">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2492" name="__module.model.layers.14.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2433,2435,hidden_states.427">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2493" name="self.model.layers.14.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="722280664" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2494" name="Convert_182394" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2495" name="self.model.layers.14.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="724377816" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2496" name="Convert_182396" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2497" name="self.model.layers.14.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2498" name="self.model.layers.14.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="724410584" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2499" name="self.model.layers.14.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2500" name="Constant_182401" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2501" name="Reshape_182402" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2502" name="__module.model.layers.14.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2503" name="__module.model.layers.14.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2444,query_states.29">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2504" name="Constant_162338" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2505" name="__module.model.layers.14.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2450">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2506" name="__module.model.layers.14.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2507" name="__module.model.layers.14.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2451,q.29">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2508" name="__module.model.layers.14.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2460">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2509" name="Constant_142425" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2510" name="Constant_142427" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2511" name="Constant_142429" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2512" name="__module.model.layers.14.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2470,x2.57">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2513" name="Constant_157559" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2514" name="__module.model.layers.14.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2471">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2515" name="Constant_142322" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2516" name="Constant_142366" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2517" name="Constant_142368" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2518" name="__module.model.layers.14.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2465">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2519" name="__module.model.layers.14.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2473">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2520" name="__module.model.layers.14.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2474">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2521" name="__module.model.layers.14.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2475">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2522" name="ReadValue_163987" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.14.keypresent.14.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.14.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2523" name="Constant_163602" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="2524" name="Gather_163603" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2525" name="self.model.layers.14.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="724541656" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2526" name="Convert_182724" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2527" name="self.model.layers.14.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="725065944" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2528" name="Convert_182726" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2529" name="self.model.layers.14.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2530" name="self.model.layers.14.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="725074136" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2531" name="self.model.layers.14.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2532" name="Constant_182731" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2533" name="Reshape_182732" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2534" name="__module.model.layers.14.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2535" name="__module.model.layers.14.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2446,key_states.61">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2536" name="Constant_162339" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2537" name="__module.model.layers.14.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2453">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2538" name="__module.model.layers.14.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2539" name="__module.model.layers.14.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2454,k.29">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2540" name="__module.model.layers.14.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2476">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2541" name="Constant_142568" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2542" name="Constant_142570" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2543" name="Constant_142572" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2544" name="__module.model.layers.14.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2486,x2.59">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2545" name="Constant_157560" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2546" name="__module.model.layers.14.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2487">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2547" name="Constant_142465" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2548" name="Constant_142509" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2549" name="Constant_142511" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2550" name="__module.model.layers.14.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2481">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2551" name="__module.model.layers.14.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2489">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2552" name="__module.model.layers.14.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2490">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2553" name="__module.model.layers.14.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2491">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2554" name="__module.model.layers.14.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.14.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2555" name="__module.model.layers.14.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2503,2504,2505">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2556" name="__module.model.layers.14.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2507,hidden_states.431">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2557" name="Constant_162340" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2558" name="__module.model.layers.14.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2511,key_states.63">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2559" name="ReadValue_163989" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.14.valuepresent.14.value" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.14.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2560" name="Constant_163605" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="2561" name="Gather_163606" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2562" name="self.model.layers.14.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="725106904" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2563" name="Convert_182714" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2564" name="self.model.layers.14.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="725631192" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2565" name="Convert_182716" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2566" name="self.model.layers.14.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2567" name="self.model.layers.14.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="725639384" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2568" name="self.model.layers.14.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2569" name="Constant_182721" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2570" name="Reshape_182722" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2571" name="__module.model.layers.14.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2572" name="__module.model.layers.14.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2448,value_states.29">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2573" name="Constant_162341" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2574" name="__module.model.layers.14.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2456">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2575" name="__module.model.layers.14.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2576" name="__module.model.layers.14.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2457">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2577" name="__module.model.layers.14.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.14.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2578" name="__module.model.layers.14.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2519,2520,2521">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2579" name="__module.model.layers.14.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2523,hidden_states.435">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2580" name="Constant_162342" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2581" name="__module.model.layers.14.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2527">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2582" name="__module.model.layers.14.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="2533,attn_output.57">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2583" name="__module.model.layers.14.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2584" name="__module.model.layers.14.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2534">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2585" name="__module.model.layers.14.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2537,a.85">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2586" name="self.model.layers.14.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="725672152" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2587" name="Convert_182404" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2588" name="self.model.layers.14.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="727769304" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2589" name="Convert_182406" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2590" name="self.model.layers.14.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2591" name="self.model.layers.14.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="727802072" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2592" name="self.model.layers.14.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2593" name="Constant_182411" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2594" name="Reshape_182412" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2595" name="__module.model.layers.14.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2596" name="__module.model.layers.14.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2539,2541,hidden_states.437">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2597" name="__module.model.layers.14/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2544,2546,2557,hidden_states.439,hidden_states.441">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2598" name="Constant_157564" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="727933144" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2599" name="Constant_157563" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2600" name="Constant_157561" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2601" name="__module.model.layers.14.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2547">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2602" name="Constant_142927" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="2548">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2603" name="__module.model.layers.14.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2549,variance.59">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2604" name="Constant_157562" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2605" name="__module.model.layers.14.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2550">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2606" name="__module.model.layers.14.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2607" name="__module.model.layers.14.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2551">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2608" name="__module.model.layers.14.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2552,2553,hidden_states.443,hidden_states.445">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2609" name="__module.model.layers.14.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2554,2556,a.87">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2610" name="self.model.layers.14.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="727941336" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2611" name="Convert_182064" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2612" name="self.model.layers.14.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="736329944" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2613" name="Convert_182066" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2614" name="self.model.layers.14.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2615" name="self.model.layers.14.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="736461016" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2616" name="self.model.layers.14.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2617" name="Constant_182071" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2618" name="Reshape_182072" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2619" name="__module.model.layers.14.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2620" name="__module.model.layers.14.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2562,input.29">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2621" name="__module.model.layers.14.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="2563">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2622" name="self.model.layers.14.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="736985304" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2623" name="Convert_182054" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2624" name="self.model.layers.14.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="745373912" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2625" name="Convert_182056" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2626" name="self.model.layers.14.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2627" name="self.model.layers.14.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="745504984" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2628" name="self.model.layers.14.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2629" name="Constant_182061" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2630" name="Reshape_182062" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2631" name="__module.model.layers.14.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2632" name="__module.model.layers.14.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2565">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2633" name="__module.model.layers.14.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2566,a.89">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2634" name="self.model.layers.14.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="746029272" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2635" name="Convert_182074" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2636" name="self.model.layers.14.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="754417880" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2637" name="Convert_182076" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2638" name="self.model.layers.14.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2639" name="self.model.layers.14.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="754548952" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2640" name="self.model.layers.14.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2641" name="Constant_182081" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2642" name="Reshape_182082" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2643" name="__module.model.layers.14.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2644" name="__module.model.layers.14.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2568,hidden_states.447">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2645" name="__module.model.layers.14/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2569,2571,2579,2590,hidden_states.449,hidden_states.451">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2646" name="Constant_157567" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2647" name="Constant_157565" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2648" name="__module.model.layers.15.input_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2580">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2649" name="Constant_143001" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="2581">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2650" name="__module.model.layers.15.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2582,variance.61">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2651" name="Constant_157566" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2652" name="__module.model.layers.15.input_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2583">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2653" name="__module.model.layers.15.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2654" name="__module.model.layers.15.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2584">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2655" name="__module.model.layers.15.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2585,2586,hidden_states.453,hidden_states.455">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2656" name="__module.model.layers.15.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2587,2589,hidden_states.457">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2657" name="self.model.layers.15.self_attn.v_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="755073240" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2658" name="Convert_182734" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2659" name="self.model.layers.15.self_attn.v_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="755597528" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2660" name="Convert_182736" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2661" name="self.model.layers.15.self_attn.v_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2662" name="self.model.layers.15.self_attn.v_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="755605720" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2663" name="self.model.layers.15.self_attn.v_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2664" name="Constant_182741" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2665" name="Reshape_182742" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2666" name="__module.model.layers.15.self_attn.v_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2667" name="__module.model.layers.15.self_attn.v_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2602,value_states">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2668" name="Constant_162343" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2669" name="__module.model.layers.15.self_attn/aten::view/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2610">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2670" name="__module.model.layers.15.self_attn/aten::transpose/Constant" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2671" name="__module.model.layers.15.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2611">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2672" name="__module.model.layers.15.self_attn/aten::cat/Concat" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.15.value">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2674" name="ReadValue_163991" type="ReadValue" version="opset6">
+ <data variable_id="past_key_values.15.keypresent.15.key" variable_type="f32" variable_shape="?,8,?,64" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>0</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="past_key_values.15.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2675" name="Constant_163608" type="Const" version="opset1">
+ <data element_type="i64" shape="" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64" />
+ </output>
+ </layer>
+ <layer id="2676" name="Gather_163609" type="Gather" version="opset8">
+ <data batch_dims="0" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>-1</dim>
+ </port>
+ <port id="2" precision="I64" />
+ </input>
+ <output>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2677" name="self.model.layers.15.self_attn.k_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 64" offset="755638488" size="524288" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2678" name="Convert_182744" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2679" name="self.model.layers.15.self_attn.k_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="512, 32, 1" offset="756162776" size="8192" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2680" name="Convert_182746" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2681" name="self.model.layers.15.self_attn.k_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2682" name="self.model.layers.15.self_attn.k_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="512, 32, 1" offset="756170968" size="32768" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2683" name="self.model.layers.15.self_attn.k_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2684" name="Constant_182751" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265895968" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2685" name="Reshape_182752" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2686" name="__module.model.layers.15.self_attn.k_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2687" name="__module.model.layers.15.self_attn.k_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>512</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2600,key_states.65">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2688" name="Constant_162344" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265895984" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2689" name="__module.model.layers.15.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>512</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2607">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2690" name="__module.model.layers.15.self_attn/aten::transpose/Constant_1" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2691" name="__module.model.layers.15.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2608,k">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2692" name="__module.model.layers.15.self_attn/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2630">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2693" name="Constant_143541" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2694" name="Constant_143543" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2695" name="Constant_143545" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2696" name="__module.model.layers.15.self_attn/aten::slice/Slice" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2640,x2">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2697" name="Constant_157569" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2698" name="__module.model.layers.15.self_attn/aten::neg/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2641">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2699" name="Constant_143438" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2700" name="Constant_143482" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2701" name="Constant_143484" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2702" name="__module.model.layers.15.self_attn/aten::slice/Slice_1" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2635">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2703" name="__module.model.layers.15.self_attn/aten::cat/Concat_1" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2643">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2704" name="__module.model.layers.15.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2644">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2705" name="__module.model.layers.15.self_attn/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2645">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2706" name="__module.model.layers.15.self_attn/aten::cat/Concat_2" type="Concat" version="opset1">
+ <data axis="-2" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="present.15.key">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2722" name="Constant_157578" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="756203736" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2723" name="self.model.layers.15.self_attn.q_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="756211928" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2724" name="Convert_182414" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2725" name="self.model.layers.15.self_attn.q_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="758309080" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2726" name="Convert_182416" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2727" name="self.model.layers.15.self_attn.q_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2728" name="self.model.layers.15.self_attn.q_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="758341848" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2729" name="self.model.layers.15.self_attn.q_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2730" name="Constant_182421" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2731" name="Reshape_182422" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2732" name="__module.model.layers.15.self_attn.q_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2733" name="__module.model.layers.15.self_attn.q_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2598,query_states">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2734" name="Constant_162345" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265330496" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2735" name="__module.model.layers.15.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2604">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2736" name="__module.model.layers.15.self_attn/aten::transpose/Constant_2" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2737" name="__module.model.layers.15.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2605,q">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2738" name="__module.model.layers.15.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2614">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2739" name="Constant_143398" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2740" name="Constant_143400" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330700" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2741" name="Constant_143402" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2742" name="__module.model.layers.15.self_attn/aten::slice/Slice_2" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2624,x2.61">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2743" name="Constant_157570" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1, 1" offset="265330716" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2744" name="__module.model.layers.15.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2625">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2745" name="Constant_143295" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="4" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2746" name="Constant_143339" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330692" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2747" name="Constant_143341" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="265330672" size="8" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2748" name="__module.model.layers.15.self_attn/aten::slice/Slice_3" type="Slice" version="opset8">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="2" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="3" precision="I64">
+ <dim>1</dim>
+ </port>
+ <port id="4" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="5" precision="FP32" names="2619">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2749" name="__module.model.layers.15.self_attn/aten::cat/Concat_3" type="Concat" version="opset1">
+ <data axis="-1" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2627">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2750" name="__module.model.layers.15.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2628">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2751" name="__module.model.layers.15.self_attn/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2629">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2752" name="__module.model.layers.15.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2657,2658,2659">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2753" name="__module.model.layers.15.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2661,hidden_states.461">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2754" name="Constant_162346" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2755" name="__module.model.layers.15.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2665,key_states">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2756" name="__module.model.layers.15.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64" />
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2673,2674,2675">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2757" name="__module.model.layers.15.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3">
+ <data mode="bidirectional" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>5</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2677,hidden_states.465">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2758" name="Constant_162347" type="Const" version="opset1">
+ <data element_type="i64" shape="4" offset="265896032" size="32" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2759" name="__module.model.layers.15.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1">
+ <data special_zero="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>4</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2681">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2760" name="__module.model.layers.15.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+ <data causal="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="2" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="3" precision="FP32">
+ <dim>-1</dim>
+ <dim>1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="4" precision="FP32" names="2687,attn_output.61">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2761" name="__module.model.layers.15.self_attn/aten::transpose/Constant_3" type="Const" version="opset1">
+ <data element_type="i32" shape="4" offset="265330528" size="16" />
+ <output>
+ <port id="0" precision="I32">
+ <dim>4</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2762" name="__module.model.layers.15.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I32">
+ <dim>4</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2688">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2763" name="__module.model.layers.15.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>3</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2691,a.91">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2764" name="self.model.layers.15.self_attn.o_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 64" offset="758472920" size="2097152" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2765" name="Convert_182424" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2766" name="self.model.layers.15.self_attn.o_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 32, 1" offset="760570072" size="32768" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2767" name="Convert_182426" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2768" name="self.model.layers.15.self_attn.o_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2769" name="self.model.layers.15.self_attn.o_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 32, 1" offset="760602840" size="131072" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2770" name="self.model.layers.15.self_attn.o_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2771" name="Constant_182431" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="265330480" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2772" name="Reshape_182432" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2773" name="__module.model.layers.15.self_attn.o_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2774" name="__module.model.layers.15.self_attn.o_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>-1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2693,2695,hidden_states.467">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2775" name="__module.model.layers.15/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2698,2700,2711,hidden_states.469,hidden_states.471">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2776" name="Constant_157574" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 2048" offset="760733912" size="8192" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2777" name="Constant_157573" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2778" name="Constant_157571" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2779" name="__module.model.layers.15.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2701">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2780" name="Constant_143900" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="2702">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2781" name="__module.model.layers.15.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2703,variance.63">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2782" name="Constant_157572" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2783" name="__module.model.layers.15.post_attention_layernorm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2704">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2784" name="__module.model.layers.15.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2785" name="__module.model.layers.15.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2705">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2786" name="__module.model.layers.15.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2706,2707,hidden_states.473,hidden_states.475">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2787" name="__module.model.layers.15.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2708,2710,a.93">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2788" name="self.model.layers.15.mlp.gate_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="760742104" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2789" name="Convert_182094" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2790" name="self.model.layers.15.mlp.gate_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="769130712" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2791" name="Convert_182096" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2792" name="self.model.layers.15.mlp.gate_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2793" name="self.model.layers.15.mlp.gate_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="769261784" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2794" name="self.model.layers.15.mlp.gate_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2795" name="Constant_182101" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2796" name="Reshape_182102" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2797" name="__module.model.layers.15.mlp.gate_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2798" name="__module.model.layers.15.mlp.gate_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2716,input">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2799" name="__module.model.layers.15.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32" names="2717">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2800" name="self.model.layers.15.mlp.up_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 64" offset="769786072" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2801" name="Convert_182084" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2802" name="self.model.layers.15.mlp.up_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="8192, 32, 1" offset="778174680" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2803" name="Convert_182086" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2804" name="self.model.layers.15.mlp.up_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2805" name="self.model.layers.15.mlp.up_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="8192, 32, 1" offset="778305752" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2806" name="self.model.layers.15.mlp.up_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2807" name="Constant_182091" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="277774520" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2808" name="Reshape_182092" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>32</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2809" name="__module.model.layers.15.mlp.up_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2810" name="__module.model.layers.15.mlp.up_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>8192</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2719">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2811" name="__module.model.layers.15.mlp/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2720,a.95">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2812" name="self.model.layers.15.mlp.down_proj.weight" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 64" offset="778830040" size="8388608" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2813" name="Convert_182104" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2814" name="self.model.layers.15.mlp.down_proj.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u4" shape="2048, 128, 1" offset="787218648" size="131072" />
+ <output>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2815" name="Convert_182106" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U4">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2816" name="self.model.layers.15.mlp.down_proj.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2817" name="self.model.layers.15.mlp.down_proj.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="2048, 128, 1" offset="787349720" size="524288" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2818" name="self.model.layers.15.mlp.down_proj.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2819" name="Constant_182111" type="Const" version="opset1">
+ <data element_type="i64" shape="2" offset="295862472" size="16" />
+ <output>
+ <port id="0" precision="I64">
+ <dim>2</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2820" name="Reshape_182112" type="Reshape" version="opset1">
+ <data special_zero="false" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>128</dim>
+ <dim>64</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>2</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2821" name="__module.model.layers.15.mlp.down_proj/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2822" name="__module.model.layers.15.mlp.down_proj/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>8192</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>2048</dim>
+ <dim>8192</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2722,hidden_states.477">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2823" name="__module.model.layers.15/aten::add/Add_1" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2723,2725,2729,hidden_states.479,hidden_states.481">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2824" name="Constant_157577" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069468" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2825" name="Constant_157575" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069472" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2826" name="__module.model.norm/aten::pow/Power" type="Power" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2730">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2827" name="Constant_143970" type="Const" version="opset1">
+ <data element_type="i64" shape="1" offset="263069476" size="8" />
+ <output>
+ <port id="0" precision="I64" names="2731">
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2828" name="__module.model.norm/aten::mean/ReduceMean" type="ReduceMean" version="opset1">
+ <data keep_dims="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="I64">
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2732,variance">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2829" name="Constant_157576" type="Const" version="opset1">
+ <data element_type="f32" shape="1, 1, 1" offset="263069484" size="4" />
+ <output>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2830" name="__module.model.norm/aten::add/Add" type="Add" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2733">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2831" name="__module.model.norm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2832" name="__module.model.norm/aten::rsqrt/Divide" type="Divide" version="opset1">
+ <data auto_broadcast="numpy" m_pythondiv="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>1</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2734">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2833" name="__module.model.norm/aten::mul/Multiply" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="2735,2736,hidden_states.483,hidden_states.485">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2834" name="__module.model.norm/aten::mul/Multiply_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>1</dim>
+ <dim>1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="103,2737,56,93,98,a,hidden_states">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2835" name="self.lm_head.weight" type="Const" version="opset1">
+ <data element_type="u8" shape="128256, 2048" offset="787874008" size="262668288" />
+ <output>
+ <port id="0" precision="U8">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2836" name="Convert_181625" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U8">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2837" name="self.lm_head.weight/zero_point" type="Const" version="opset1">
+ <data element_type="u8" shape="128256, 1" offset="1050542296" size="128256" />
+ <output>
+ <port id="0" precision="U8">
+ <dim>128256</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2838" name="Convert_181628" type="Convert" version="opset1">
+ <data destination_type="f16" />
+ <input>
+ <port id="0" precision="U8">
+ <dim>128256</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP16">
+ <dim>128256</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2839" name="self.lm_head.weight/zero_point/subtract" type="Subtract" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>128256</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2840" name="self.lm_head.weight/scale" type="Const" version="opset1">
+ <data element_type="f16" shape="128256, 1" offset="1050670552" size="256512" />
+ <output>
+ <port id="0" precision="FP16">
+ <dim>128256</dim>
+ <dim>1</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2841" name="self.lm_head.weight/fq_weights_1" type="Multiply" version="opset1">
+ <data auto_broadcast="numpy" />
+ <input>
+ <port id="0" precision="FP16">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP16">
+ <dim>128256</dim>
+ <dim>1</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP16">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2842" name="__module.lm_head/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+ <data destination_type="f32" />
+ <rt_info>
+ <attribute name="decompression" version="0" />
+ </rt_info>
+ <input>
+ <port id="0" precision="FP16">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2843" name="__module.lm_head/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+ <data transpose_a="false" transpose_b="true" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>2048</dim>
+ </port>
+ <port id="1" precision="FP32">
+ <dim>128256</dim>
+ <dim>2048</dim>
+ </port>
+ </input>
+ <output>
+ <port id="2" precision="FP32" names="logits">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>128256</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2721" name="Assign_163964" type="Assign" version="opset6">
+ <data variable_id="past_key_values.0.keypresent.0.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2720" name="Assign_163966" type="Assign" version="opset6">
+ <data variable_id="past_key_values.0.valuepresent.0.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2719" name="Assign_163968" type="Assign" version="opset6">
+ <data variable_id="past_key_values.1.keypresent.1.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2718" name="Assign_163970" type="Assign" version="opset6">
+ <data variable_id="past_key_values.1.valuepresent.1.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2717" name="Assign_163972" type="Assign" version="opset6">
+ <data variable_id="past_key_values.10.keypresent.10.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2716" name="Assign_163974" type="Assign" version="opset6">
+ <data variable_id="past_key_values.10.valuepresent.10.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2715" name="Assign_163976" type="Assign" version="opset6">
+ <data variable_id="past_key_values.11.keypresent.11.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2714" name="Assign_163978" type="Assign" version="opset6">
+ <data variable_id="past_key_values.11.valuepresent.11.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2713" name="Assign_163980" type="Assign" version="opset6">
+ <data variable_id="past_key_values.12.keypresent.12.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2712" name="Assign_163982" type="Assign" version="opset6">
+ <data variable_id="past_key_values.12.valuepresent.12.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2711" name="Assign_163984" type="Assign" version="opset6">
+ <data variable_id="past_key_values.13.keypresent.13.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2710" name="Assign_163986" type="Assign" version="opset6">
+ <data variable_id="past_key_values.13.valuepresent.13.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2709" name="Assign_163988" type="Assign" version="opset6">
+ <data variable_id="past_key_values.14.keypresent.14.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2708" name="Assign_163990" type="Assign" version="opset6">
+ <data variable_id="past_key_values.14.valuepresent.14.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2707" name="Assign_163992" type="Assign" version="opset6">
+ <data variable_id="past_key_values.15.keypresent.15.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2673" name="Assign_163994" type="Assign" version="opset6">
+ <data variable_id="past_key_values.15.valuepresent.15.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1715" name="Assign_163996" type="Assign" version="opset6">
+ <data variable_id="past_key_values.2.keypresent.2.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1714" name="Assign_163998" type="Assign" version="opset6">
+ <data variable_id="past_key_values.2.valuepresent.2.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1713" name="Assign_164000" type="Assign" version="opset6">
+ <data variable_id="past_key_values.3.keypresent.3.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1712" name="Assign_164002" type="Assign" version="opset6">
+ <data variable_id="past_key_values.3.valuepresent.3.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1711" name="Assign_164004" type="Assign" version="opset6">
+ <data variable_id="past_key_values.4.keypresent.4.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1710" name="Assign_164006" type="Assign" version="opset6">
+ <data variable_id="past_key_values.4.valuepresent.4.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1709" name="Assign_164008" type="Assign" version="opset6">
+ <data variable_id="past_key_values.5.keypresent.5.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1708" name="Assign_164010" type="Assign" version="opset6">
+ <data variable_id="past_key_values.5.valuepresent.5.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1707" name="Assign_164012" type="Assign" version="opset6">
+ <data variable_id="past_key_values.6.keypresent.6.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1706" name="Assign_164014" type="Assign" version="opset6">
+ <data variable_id="past_key_values.6.valuepresent.6.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1705" name="Assign_164016" type="Assign" version="opset6">
+ <data variable_id="past_key_values.7.keypresent.7.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1704" name="Assign_164018" type="Assign" version="opset6">
+ <data variable_id="past_key_values.7.valuepresent.7.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1703" name="Assign_164020" type="Assign" version="opset6">
+ <data variable_id="past_key_values.8.keypresent.8.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1702" name="Assign_164022" type="Assign" version="opset6">
+ <data variable_id="past_key_values.8.valuepresent.8.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1701" name="Assign_164024" type="Assign" version="opset6">
+ <data variable_id="past_key_values.9.keypresent.9.key" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="1667" name="Assign_164026" type="Assign" version="opset6">
+ <data variable_id="past_key_values.9.valuepresent.9.value" />
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </input>
+ <output>
+ <port id="1" precision="FP32">
+ <dim>-1</dim>
+ <dim>8</dim>
+ <dim>-1</dim>
+ <dim>64</dim>
+ </port>
+ </output>
+ </layer>
+ <layer id="2844" name="Result_144375" type="Result" version="opset1">
+ <input>
+ <port id="0" precision="FP32">
+ <dim>-1</dim>
+ <dim>-1</dim>
+ <dim>128256</dim>
+ </port>
+ </input>
+ </layer>
+ </layers>
+ <edges>
+ <edge from-layer="0" from-port="0" to-layer="1225" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="895" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="1901" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="1023" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="2029" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="1060" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="2066" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="1188" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="2194" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="1864" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="2231" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="1353" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="2359" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="1390" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="2396" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="2524" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="2676" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="2561" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="1555" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="93" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="147" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="363" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="400" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="528" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="565" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="858" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="16" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="1670" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="693" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="1718" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="1518" to-port="1" />
+ <edge from-layer="0" from-port="0" to-layer="730" to-port="1" />
+ <edge from-layer="1" from-port="0" to-layer="66" to-port="0" />
+ <edge from-layer="2" from-port="0" to-layer="234" to-port="0" />
+ <edge from-layer="2" from-port="0" to-layer="171" to-port="0" />
+ <edge from-layer="3" from-port="0" to-layer="26" to-port="0" />
+ <edge from-layer="3" from-port="0" to-layer="5" to-port="0" />
+ <edge from-layer="4" from-port="0" to-layer="13" to-port="0" />
+ <edge from-layer="5" from-port="1" to-layer="8" to-port="0" />
+ <edge from-layer="6" from-port="0" to-layer="8" to-port="1" />
+ <edge from-layer="7" from-port="0" to-layer="8" to-port="2" />
+ <edge from-layer="8" from-port="3" to-layer="12" to-port="0" />
+ <edge from-layer="9" from-port="0" to-layer="12" to-port="1" />
+ <edge from-layer="10" from-port="0" to-layer="12" to-port="2" />
+ <edge from-layer="11" from-port="0" to-layer="12" to-port="3" />
+ <edge from-layer="12" from-port="4" to-layer="13" to-port="1" />
+ <edge from-layer="13" from-port="2" to-layer="398" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="728" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="1716" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="691" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="1668" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="563" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="526" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="1516" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="361" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="145" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="1553" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="91" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="1186" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="2559" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="2522" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="2394" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="1388" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="2357" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="1351" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="2229" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="1223" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="2192" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="2674" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="2064" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="1058" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="2027" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="1021" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="1899" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="893" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="1862" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="856" to-port="0" />
+ <edge from-layer="13" from-port="2" to-layer="14" to-port="0" />
+ <edge from-layer="14" from-port="1" to-layer="16" to-port="0" />
+ <edge from-layer="15" from-port="0" to-layer="16" to-port="2" />
+ <edge from-layer="16" from-port="3" to-layer="1666" to-port="0" />
+ <edge from-layer="17" from-port="0" to-layer="1650" to-port="0" />
+ <edge from-layer="18" from-port="0" to-layer="19" to-port="0" />
+ <edge from-layer="19" from-port="1" to-layer="22" to-port="0" />
+ <edge from-layer="20" from-port="0" to-layer="21" to-port="0" />
+ <edge from-layer="21" from-port="1" to-layer="22" to-port="1" />
+ <edge from-layer="22" from-port="2" to-layer="24" to-port="0" />
+ <edge from-layer="23" from-port="0" to-layer="24" to-port="1" />
+ <edge from-layer="24" from-port="2" to-layer="25" to-port="0" />
+ <edge from-layer="25" from-port="1" to-layer="28" to-port="0" />
+ <edge from-layer="26" from-port="1" to-layer="28" to-port="1" />
+ <edge from-layer="27" from-port="0" to-layer="28" to-port="2" />
+ <edge from-layer="28" from-port="3" to-layer="271" to-port="0" />
+ <edge from-layer="28" from-port="3" to-layer="39" to-port="0" />
+ <edge from-layer="28" from-port="3" to-layer="32" to-port="0" />
+ <edge from-layer="29" from-port="0" to-layer="40" to-port="0" />
+ <edge from-layer="30" from-port="0" to-layer="38" to-port="0" />
+ <edge from-layer="31" from-port="0" to-layer="32" to-port="1" />
+ <edge from-layer="32" from-port="2" to-layer="34" to-port="0" />
+ <edge from-layer="33" from-port="0" to-layer="34" to-port="1" />
+ <edge from-layer="34" from-port="2" to-layer="36" to-port="0" />
+ <edge from-layer="35" from-port="0" to-layer="36" to-port="1" />
+ <edge from-layer="36" from-port="2" to-layer="37" to-port="0" />
+ <edge from-layer="37" from-port="1" to-layer="38" to-port="1" />
+ <edge from-layer="38" from-port="2" to-layer="39" to-port="1" />
+ <edge from-layer="39" from-port="2" to-layer="40" to-port="1" />
+ <edge from-layer="40" from-port="2" to-layer="51" to-port="0" />
+ <edge from-layer="40" from-port="2" to-layer="104" to-port="0" />
+ <edge from-layer="40" from-port="2" to-layer="158" to-port="0" />
+ <edge from-layer="40" from-port="2" to-layer="57" to-port="0" />
+ <edge from-layer="41" from-port="0" to-layer="42" to-port="0" />
+ <edge from-layer="42" from-port="1" to-layer="45" to-port="0" />
+ <edge from-layer="43" from-port="0" to-layer="44" to-port="0" />
+ <edge from-layer="44" from-port="1" to-layer="45" to-port="1" />
+ <edge from-layer="45" from-port="2" to-layer="47" to-port="0" />
+ <edge from-layer="46" from-port="0" to-layer="47" to-port="1" />
+ <edge from-layer="47" from-port="2" to-layer="49" to-port="0" />
+ <edge from-layer="48" from-port="0" to-layer="49" to-port="1" />
+ <edge from-layer="49" from-port="2" to-layer="50" to-port="0" />
+ <edge from-layer="50" from-port="1" to-layer="51" to-port="1" />
+ <edge from-layer="51" from-port="2" to-layer="53" to-port="0" />
+ <edge from-layer="52" from-port="0" to-layer="53" to-port="1" />
+ <edge from-layer="53" from-port="2" to-layer="55" to-port="0" />
+ <edge from-layer="54" from-port="0" to-layer="55" to-port="1" />
+ <edge from-layer="55" from-port="2" to-layer="74" to-port="0" />
+ <edge from-layer="55" from-port="2" to-layer="85" to-port="0" />
+ <edge from-layer="55" from-port="2" to-layer="79" to-port="0" />
+ <edge from-layer="56" from-port="0" to-layer="64" to-port="0" />
+ <edge from-layer="57" from-port="1" to-layer="60" to-port="0" />
+ <edge from-layer="57" from-port="1" to-layer="136" to-port="0" />
+ <edge from-layer="57" from-port="1" to-layer="256" to-port="0" />
+ <edge from-layer="58" from-port="0" to-layer="60" to-port="1" />
+ <edge from-layer="59" from-port="0" to-layer="60" to-port="2" />
+ <edge from-layer="60" from-port="3" to-layer="63" to-port="0" />
+ <edge from-layer="60" from-port="3" to-layer="141" to-port="0" />
+ <edge from-layer="60" from-port="3" to-layer="216" to-port="0" />
+ <edge from-layer="61" from-port="0" to-layer="63" to-port="1" />
+ <edge from-layer="62" from-port="0" to-layer="63" to-port="2" />
+ <edge from-layer="63" from-port="3" to-layer="64" to-port="1" />
+ <edge from-layer="64" from-port="2" to-layer="68" to-port="0" />
+ <edge from-layer="65" from-port="0" to-layer="73" to-port="1" />
+ <edge from-layer="65" from-port="0" to-layer="66" to-port="1" />
+ <edge from-layer="65" from-port="0" to-layer="212" to-port="1" />
+ <edge from-layer="65" from-port="0" to-layer="88" to-port="1" />
+ <edge from-layer="65" from-port="0" to-layer="234" to-port="1" />
+ <edge from-layer="66" from-port="2" to-layer="67" to-port="0" />
+ <edge from-layer="67" from-port="1" to-layer="68" to-port="1" />
+ <edge from-layer="68" from-port="2" to-layer="70" to-port="0" />
+ <edge from-layer="69" from-port="0" to-layer="70" to-port="1" />
+ <edge from-layer="70" from-port="2" to-layer="71" to-port="1" />
+ <edge from-layer="70" from-port="2" to-layer="71" to-port="0" />
+ <edge from-layer="71" from-port="2" to-layer="87" to-port="0" />
+ <edge from-layer="71" from-port="2" to-layer="72" to-port="0" />
+ <edge from-layer="72" from-port="1" to-layer="73" to-port="0" />
+ <edge from-layer="73" from-port="2" to-layer="74" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="544" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="2375" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="2508" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="512" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="2540" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="2692" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="2738" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="379" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="1534" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="1369" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="347" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="1337" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="1204" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="1172" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="1039" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="1007" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="2343" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="1686" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="1735" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="842" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="1848" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="1880" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="1502" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="2013" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="874" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="2045" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="709" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="2178" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="2210" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="677" to-port="1" />
+ <edge from-layer="73" from-port="2" to-layer="109" to-port="1" />
+ <edge from-layer="74" from-port="2" to-layer="90" to-port="0" />
+ <edge from-layer="75" from-port="0" to-layer="79" to-port="1" />
+ <edge from-layer="76" from-port="0" to-layer="79" to-port="2" />
+ <edge from-layer="77" from-port="0" to-layer="79" to-port="3" />
+ <edge from-layer="78" from-port="0" to-layer="713" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2544" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2518" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2512" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="681" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2385" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2379" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="687" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2347" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="113" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2220" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2214" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="250" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2188" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2182" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2550" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="884" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="119" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="719" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="878" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="233" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2055" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2049" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="228" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2023" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2017" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1506" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1890" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1176" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="389" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="383" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="516" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="522" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="548" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="554" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1011" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1017" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="79" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="357" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1043" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1049" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="351" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2353" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1182" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="85" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1208" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1214" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1341" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1347" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1373" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1379" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2748" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2742" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2702" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="2696" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1858" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1544" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1696" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1852" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1690" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="852" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="846" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1884" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1745" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1739" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1538" to-port="4" />
+ <edge from-layer="78" from-port="0" to-layer="1512" to-port="4" />
+ <edge from-layer="79" from-port="5" to-layer="81" to-port="0" />
+ <edge from-layer="80" from-port="0" to-layer="81" to-port="1" />
+ <edge from-layer="81" from-port="2" to-layer="86" to-port="0" />
+ <edge from-layer="82" from-port="0" to-layer="85" to-port="1" />
+ <edge from-layer="83" from-port="0" to-layer="85" to-port="2" />
+ <edge from-layer="84" from-port="0" to-layer="85" to-port="3" />
+ <edge from-layer="85" from-port="5" to-layer="86" to-port="1" />
+ <edge from-layer="86" from-port="2" to-layer="89" to-port="0" />
+ <edge from-layer="87" from-port="1" to-layer="88" to-port="0" />
+ <edge from-layer="88" from-port="2" to-layer="2222" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="2355" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="689" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="359" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="1019" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="1546" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="1216" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="556" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="391" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="2190" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="886" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="2387" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="89" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="1051" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="2520" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="1747" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="2552" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="1184" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="1698" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="2704" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="2750" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="1381" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="1349" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="121" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="721" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="2025" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="1892" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="1860" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="2057" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="1514" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="854" to-port="1" />
+ <edge from-layer="88" from-port="2" to-layer="524" to-port="1" />
+ <edge from-layer="89" from-port="2" to-layer="90" to-port="1" />
+ <edge from-layer="90" from-port="2" to-layer="251" to-port="0" />
+ <edge from-layer="91" from-port="1" to-layer="93" to-port="0" />
+ <edge from-layer="92" from-port="0" to-layer="93" to-port="2" />
+ <edge from-layer="93" from-port="3" to-layer="123" to-port="0" />
+ <edge from-layer="93" from-port="3" to-layer="128" to-port="0" />
+ <edge from-layer="94" from-port="0" to-layer="95" to-port="0" />
+ <edge from-layer="95" from-port="1" to-layer="98" to-port="0" />
+ <edge from-layer="96" from-port="0" to-layer="97" to-port="0" />
+ <edge from-layer="97" from-port="1" to-layer="98" to-port="1" />
+ <edge from-layer="98" from-port="2" to-layer="100" to-port="0" />
+ <edge from-layer="99" from-port="0" to-layer="100" to-port="1" />
+ <edge from-layer="100" from-port="2" to-layer="102" to-port="0" />
+ <edge from-layer="101" from-port="0" to-layer="102" to-port="1" />
+ <edge from-layer="102" from-port="2" to-layer="103" to-port="0" />
+ <edge from-layer="103" from-port="1" to-layer="104" to-port="1" />
+ <edge from-layer="104" from-port="2" to-layer="106" to-port="0" />
+ <edge from-layer="105" from-port="0" to-layer="106" to-port="1" />
+ <edge from-layer="106" from-port="2" to-layer="108" to-port="0" />
+ <edge from-layer="107" from-port="0" to-layer="108" to-port="1" />
+ <edge from-layer="108" from-port="2" to-layer="119" to-port="0" />
+ <edge from-layer="108" from-port="2" to-layer="113" to-port="0" />
+ <edge from-layer="108" from-port="2" to-layer="109" to-port="0" />
+ <edge from-layer="109" from-port="2" to-layer="122" to-port="0" />
+ <edge from-layer="110" from-port="0" to-layer="113" to-port="1" />
+ <edge from-layer="111" from-port="0" to-layer="113" to-port="2" />
+ <edge from-layer="112" from-port="0" to-layer="113" to-port="3" />
+ <edge from-layer="113" from-port="5" to-layer="115" to-port="0" />
+ <edge from-layer="114" from-port="0" to-layer="115" to-port="1" />
+ <edge from-layer="115" from-port="2" to-layer="120" to-port="0" />
+ <edge from-layer="116" from-port="0" to-layer="119" to-port="1" />
+ <edge from-layer="117" from-port="0" to-layer="119" to-port="2" />
+ <edge from-layer="118" from-port="0" to-layer="119" to-port="3" />
+ <edge from-layer="119" from-port="5" to-layer="120" to-port="1" />
+ <edge from-layer="120" from-port="2" to-layer="121" to-port="0" />
+ <edge from-layer="121" from-port="2" to-layer="122" to-port="1" />
+ <edge from-layer="122" from-port="2" to-layer="123" to-port="1" />
+ <edge from-layer="123" from-port="2" to-layer="2721" to-port="0" />
+ <edge from-layer="123" from-port="2" to-layer="125" to-port="0" />
+ <edge from-layer="124" from-port="0" to-layer="724" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="2578" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="417" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="747" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="2555" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="1384" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="2752" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="1077" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="2756" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="1407" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="1749" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="125" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="889" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="1219" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="1895" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="2083" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="559" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="394" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="1549" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="2225" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="582" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="1242" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="2248" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="2060" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="1918" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="235" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="1753" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="164" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="2390" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="912" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="2413" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="1054" to-port="1" />
+ <edge from-layer="124" from-port="0" to-layer="1572" to-port="1" />
+ <edge from-layer="125" from-port="2" to-layer="142" to-port="0" />
+ <edge from-layer="126" from-port="0" to-layer="141" to-port="1" />
+ <edge from-layer="127" from-port="0" to-layer="141" to-port="2" />
+ <edge from-layer="128" from-port="1" to-layer="131" to-port="0" />
+ <edge from-layer="129" from-port="0" to-layer="131" to-port="1" />
+ <edge from-layer="130" from-port="0" to-layer="131" to-port="2" />
+ <edge from-layer="131" from-port="3" to-layer="204" to-port="0" />
+ <edge from-layer="131" from-port="3" to-layer="133" to-port="0" />
+ <edge from-layer="131" from-port="3" to-layer="202" to-port="0" />
+ <edge from-layer="132" from-port="0" to-layer="133" to-port="1" />
+ <edge from-layer="133" from-port="2" to-layer="139" to-port="0" />
+ <edge from-layer="134" from-port="0" to-layer="136" to-port="1" />
+ <edge from-layer="135" from-port="0" to-layer="136" to-port="2" />
+ <edge from-layer="136" from-port="3" to-layer="138" to-port="0" />
+ <edge from-layer="136" from-port="3" to-layer="202" to-port="1" />
+ <edge from-layer="137" from-port="0" to-layer="138" to-port="1" />
+ <edge from-layer="138" from-port="2" to-layer="139" to-port="1" />
+ <edge from-layer="138" from-port="2" to-layer="177" to-port="0" />
+ <edge from-layer="139" from-port="2" to-layer="141" to-port="3" />
+ <edge from-layer="139" from-port="2" to-layer="250" to-port="2" />
+ <edge from-layer="140" from-port="0" to-layer="141" to-port="4" />
+ <edge from-layer="141" from-port="5" to-layer="1078" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="1573" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="165" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="560" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="583" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="142" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="395" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="1550" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="1055" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="418" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="2556" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="2249" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="2226" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="890" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="2084" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="2061" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="1754" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="2391" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="748" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="913" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="2414" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="1750" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="2579" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="725" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="2753" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="2757" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="1919" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="1408" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="1385" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="1896" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="1220" to-port="1" />
+ <edge from-layer="141" from-port="5" to-layer="1243" to-port="1" />
+ <edge from-layer="142" from-port="2" to-layer="144" to-port="0" />
+ <edge from-layer="143" from-port="0" to-layer="144" to-port="1" />
+ <edge from-layer="144" from-port="2" to-layer="251" to-port="1" />
+ <edge from-layer="145" from-port="1" to-layer="147" to-port="0" />
+ <edge from-layer="146" from-port="0" to-layer="147" to-port="2" />
+ <edge from-layer="147" from-port="3" to-layer="163" to-port="0" />
+ <edge from-layer="148" from-port="0" to-layer="149" to-port="0" />
+ <edge from-layer="149" from-port="1" to-layer="152" to-port="0" />
+ <edge from-layer="150" from-port="0" to-layer="151" to-port="0" />
+ <edge from-layer="151" from-port="1" to-layer="152" to-port="1" />
+ <edge from-layer="152" from-port="2" to-layer="154" to-port="0" />
+ <edge from-layer="153" from-port="0" to-layer="154" to-port="1" />
+ <edge from-layer="154" from-port="2" to-layer="156" to-port="0" />
+ <edge from-layer="155" from-port="0" to-layer="156" to-port="1" />
+ <edge from-layer="156" from-port="2" to-layer="157" to-port="0" />
+ <edge from-layer="157" from-port="1" to-layer="158" to-port="1" />
+ <edge from-layer="158" from-port="2" to-layer="160" to-port="0" />
+ <edge from-layer="159" from-port="0" to-layer="160" to-port="1" />
+ <edge from-layer="160" from-port="2" to-layer="162" to-port="0" />
+ <edge from-layer="161" from-port="0" to-layer="162" to-port="1" />
+ <edge from-layer="162" from-port="2" to-layer="163" to-port="1" />
+ <edge from-layer="163" from-port="2" to-layer="164" to-port="0" />
+ <edge from-layer="163" from-port="2" to-layer="2720" to-port="0" />
+ <edge from-layer="164" from-port="2" to-layer="165" to-port="0" />
+ <edge from-layer="165" from-port="2" to-layer="167" to-port="0" />
+ <edge from-layer="166" from-port="0" to-layer="167" to-port="1" />
+ <edge from-layer="167" from-port="2" to-layer="251" to-port="2" />
+ <edge from-layer="168" from-port="0" to-layer="186" to-port="1" />
+ <edge from-layer="168" from-port="0" to-layer="185" to-port="0" />
+ <edge from-layer="169" from-port="0" to-layer="179" to-port="0" />
+ <edge from-layer="170" from-port="0" to-layer="178" to-port="0" />
+ <edge from-layer="171" from-port="1" to-layer="174" to-port="0" />
+ <edge from-layer="172" from-port="0" to-layer="174" to-port="1" />
+ <edge from-layer="173" from-port="0" to-layer="174" to-port="2" />
+ <edge from-layer="174" from-port="3" to-layer="176" to-port="0" />
+ <edge from-layer="174" from-port="3" to-layer="199" to-port="1" />
+ <edge from-layer="175" from-port="0" to-layer="176" to-port="1" />
+ <edge from-layer="176" from-port="2" to-layer="228" to-port="2" />
+ <edge from-layer="176" from-port="2" to-layer="177" to-port="1" />
+ <edge from-layer="176" from-port="2" to-layer="233" to-port="2" />
+ <edge from-layer="177" from-port="2" to-layer="178" to-port="1" />
+ <edge from-layer="178" from-port="2" to-layer="179" to-port="1" />
+ <edge from-layer="179" from-port="2" to-layer="180" to-port="0" />
+ <edge from-layer="179" from-port="2" to-layer="196" to-port="1" />
+ <edge from-layer="180" from-port="1" to-layer="183" to-port="0" />
+ <edge from-layer="180" from-port="1" to-layer="190" to-port="0" />
+ <edge from-layer="181" from-port="0" to-layer="183" to-port="1" />
+ <edge from-layer="182" from-port="0" to-layer="183" to-port="2" />
+ <edge from-layer="183" from-port="3" to-layer="185" to-port="1" />
+ <edge from-layer="184" from-port="0" to-layer="193" to-port="1" />
+ <edge from-layer="184" from-port="0" to-layer="185" to-port="2" />
+ <edge from-layer="184" from-port="0" to-layer="192" to-port="2" />
+ <edge from-layer="185" from-port="3" to-layer="186" to-port="0" />
+ <edge from-layer="186" from-port="2" to-layer="194" to-port="0" />
+ <edge from-layer="187" from-port="0" to-layer="192" to-port="0" />
+ <edge from-layer="187" from-port="0" to-layer="191" to-port="1" />
+ <edge from-layer="188" from-port="0" to-layer="190" to-port="1" />
+ <edge from-layer="189" from-port="0" to-layer="190" to-port="2" />
+ <edge from-layer="190" from-port="3" to-layer="191" to-port="0" />
+ <edge from-layer="191" from-port="2" to-layer="192" to-port="1" />
+ <edge from-layer="192" from-port="3" to-layer="193" to-port="0" />
+ <edge from-layer="193" from-port="2" to-layer="194" to-port="1" />
+ <edge from-layer="194" from-port="2" to-layer="196" to-port="0" />
+ <edge from-layer="195" from-port="0" to-layer="196" to-port="2" />
+ <edge from-layer="196" from-port="3" to-layer="209" to-port="0" />
+ <edge from-layer="197" from-port="0" to-layer="199" to-port="0" />
+ <edge from-layer="198" from-port="0" to-layer="199" to-port="2" />
+ <edge from-layer="199" from-port="3" to-layer="200" to-port="0" />
+ <edge from-layer="200" from-port="1" to-layer="201" to-port="0" />
+ <edge from-layer="201" from-port="1" to-layer="207" to-port="0" />
+ <edge from-layer="202" from-port="2" to-layer="204" to-port="1" />
+ <edge from-layer="203" from-port="0" to-layer="204" to-port="2" />
+ <edge from-layer="204" from-port="3" to-layer="206" to-port="0" />
+ <edge from-layer="205" from-port="0" to-layer="206" to-port="1" />
+ <edge from-layer="206" from-port="2" to-layer="207" to-port="1" />
+ <edge from-layer="207" from-port="2" to-layer="208" to-port="0" />
+ <edge from-layer="208" from-port="1" to-layer="209" to-port="1" />
+ <edge from-layer="209" from-port="2" to-layer="211" to-port="0" />
+ <edge from-layer="210" from-port="0" to-layer="211" to-port="1" />
+ <edge from-layer="211" from-port="2" to-layer="212" to-port="0" />
+ <edge from-layer="212" from-port="2" to-layer="217" to-port="0" />
+ <edge from-layer="213" from-port="0" to-layer="216" to-port="1" />
+ <edge from-layer="214" from-port="0" to-layer="216" to-port="2" />
+ <edge from-layer="215" from-port="0" to-layer="216" to-port="3" />
+ <edge from-layer="216" from-port="4" to-layer="217" to-port="1" />
+ <edge from-layer="217" from-port="2" to-layer="233" to-port="0" />
+ <edge from-layer="217" from-port="2" to-layer="221" to-port="0" />
+ <edge from-layer="217" from-port="2" to-layer="219" to-port="0" />
+ <edge from-layer="218" from-port="0" to-layer="219" to-port="1" />
+ <edge from-layer="219" from-port="2" to-layer="246" to-port="0" />
+ <edge from-layer="220" from-port="0" to-layer="222" to-port="1" />
+ <edge from-layer="220" from-port="0" to-layer="224" to-port="0" />
+ <edge from-layer="221" from-port="1" to-layer="222" to-port="0" />
+ <edge from-layer="221" from-port="1" to-layer="247" to-port="1" />
+ <edge from-layer="221" from-port="1" to-layer="225" to-port="1" />
+ <edge from-layer="222" from-port="2" to-layer="224" to-port="1" />
+ <edge from-layer="223" from-port="0" to-layer="224" to-port="2" />
+ <edge from-layer="224" from-port="3" to-layer="225" to-port="0" />
+ <edge from-layer="225" from-port="2" to-layer="228" to-port="0" />
+ <edge from-layer="226" from-port="0" to-layer="228" to-port="1" />
+ <edge from-layer="227" from-port="0" to-layer="228" to-port="3" />
+ <edge from-layer="228" from-port="5" to-layer="230" to-port="0" />
+ <edge from-layer="229" from-port="0" to-layer="230" to-port="1" />
+ <edge from-layer="230" from-port="2" to-layer="246" to-port="1" />
+ <edge from-layer="231" from-port="0" to-layer="233" to-port="1" />
+ <edge from-layer="232" from-port="0" to-layer="233" to-port="3" />
+ <edge from-layer="233" from-port="5" to-layer="242" to-port="0" />
+ <edge from-layer="233" from-port="5" to-layer="237" to-port="0" />
+ <edge from-layer="233" from-port="5" to-layer="241" to-port="2" />
+ <edge from-layer="234" from-port="2" to-layer="235" to-port="0" />
+ <edge from-layer="235" from-port="2" to-layer="236" to-port="0" />
+ <edge from-layer="236" from-port="1" to-layer="237" to-port="1" />
+ <edge from-layer="237" from-port="2" to-layer="239" to-port="0" />
+ <edge from-layer="238" from-port="0" to-layer="239" to-port="1" />
+ <edge from-layer="239" from-port="2" to-layer="241" to-port="0" />
+ <edge from-layer="240" from-port="0" to-layer="241" to-port="1" />
+ <edge from-layer="241" from-port="3" to-layer="243" to-port="0" />
+ <edge from-layer="242" from-port="1" to-layer="243" to-port="1" />
+ <edge from-layer="243" from-port="2" to-layer="245" to-port="0" />
+ <edge from-layer="244" from-port="0" to-layer="245" to-port="1" />
+ <edge from-layer="245" from-port="2" to-layer="246" to-port="2" />
+ <edge from-layer="246" from-port="3" to-layer="247" to-port="0" />
+ <edge from-layer="247" from-port="2" to-layer="250" to-port="0" />
+ <edge from-layer="248" from-port="0" to-layer="250" to-port="1" />
+ <edge from-layer="249" from-port="0" to-layer="250" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="751" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="2417" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="2252" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="1576" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="586" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="1081" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="421" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="1246" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="2582" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="1411" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="916" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="1757" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="1922" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="2760" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="2087" to-port="3" />
+ <edge from-layer="250" from-port="5" to-layer="251" to-port="3" />
+ <edge from-layer="251" from-port="4" to-layer="253" to-port="0" />
+ <edge from-layer="252" from-port="0" to-layer="253" to-port="1" />
+ <edge from-layer="253" from-port="2" to-layer="259" to-port="0" />
+ <edge from-layer="254" from-port="0" to-layer="256" to-port="1" />
+ <edge from-layer="255" from-port="0" to-layer="256" to-port="2" />
+ <edge from-layer="256" from-port="3" to-layer="258" to-port="0" />
+ <edge from-layer="257" from-port="0" to-layer="258" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="1760" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="2420" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="919" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="424" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="1249" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="259" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="1925" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="2585" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="1084" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="1414" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="589" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="2763" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="1579" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="2090" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="754" to-port="1" />
+ <edge from-layer="258" from-port="2" to-layer="2255" to-port="1" />
+ <edge from-layer="259" from-port="2" to-layer="270" to-port="0" />
+ <edge from-layer="260" from-port="0" to-layer="261" to-port="0" />
+ <edge from-layer="261" from-port="1" to-layer="264" to-port="0" />
+ <edge from-layer="262" from-port="0" to-layer="263" to-port="0" />
+ <edge from-layer="263" from-port="1" to-layer="264" to-port="1" />
+ <edge from-layer="264" from-port="2" to-layer="266" to-port="0" />
+ <edge from-layer="265" from-port="0" to-layer="266" to-port="1" />
+ <edge from-layer="266" from-port="2" to-layer="268" to-port="0" />
+ <edge from-layer="267" from-port="0" to-layer="268" to-port="1" />
+ <edge from-layer="268" from-port="2" to-layer="269" to-port="0" />
+ <edge from-layer="269" from-port="1" to-layer="270" to-port="1" />
+ <edge from-layer="270" from-port="2" to-layer="271" to-port="1" />
+ <edge from-layer="271" from-port="2" to-layer="282" to-port="0" />
+ <edge from-layer="271" from-port="2" to-layer="319" to-port="0" />
+ <edge from-layer="271" from-port="2" to-layer="275" to-port="0" />
+ <edge from-layer="272" from-port="0" to-layer="283" to-port="0" />
+ <edge from-layer="273" from-port="0" to-layer="281" to-port="0" />
+ <edge from-layer="274" from-port="0" to-layer="275" to-port="1" />
+ <edge from-layer="275" from-port="2" to-layer="277" to-port="0" />
+ <edge from-layer="276" from-port="0" to-layer="277" to-port="1" />
+ <edge from-layer="277" from-port="2" to-layer="279" to-port="0" />
+ <edge from-layer="278" from-port="0" to-layer="279" to-port="1" />
+ <edge from-layer="279" from-port="2" to-layer="280" to-port="0" />
+ <edge from-layer="280" from-port="1" to-layer="281" to-port="1" />
+ <edge from-layer="281" from-port="2" to-layer="282" to-port="1" />
+ <edge from-layer="282" from-port="2" to-layer="283" to-port="1" />
+ <edge from-layer="283" from-port="2" to-layer="294" to-port="0" />
+ <edge from-layer="283" from-port="2" to-layer="306" to-port="0" />
+ <edge from-layer="284" from-port="0" to-layer="285" to-port="0" />
+ <edge from-layer="285" from-port="1" to-layer="288" to-port="0" />
+ <edge from-layer="286" from-port="0" to-layer="287" to-port="0" />
+ <edge from-layer="287" from-port="1" to-layer="288" to-port="1" />
+ <edge from-layer="288" from-port="2" to-layer="290" to-port="0" />
+ <edge from-layer="289" from-port="0" to-layer="290" to-port="1" />
+ <edge from-layer="290" from-port="2" to-layer="292" to-port="0" />
+ <edge from-layer="291" from-port="0" to-layer="292" to-port="1" />
+ <edge from-layer="292" from-port="2" to-layer="293" to-port="0" />
+ <edge from-layer="293" from-port="1" to-layer="294" to-port="1" />
+ <edge from-layer="294" from-port="2" to-layer="295" to-port="0" />
+ <edge from-layer="295" from-port="1" to-layer="307" to-port="0" />
+ <edge from-layer="296" from-port="0" to-layer="297" to-port="0" />
+ <edge from-layer="297" from-port="1" to-layer="300" to-port="0" />
+ <edge from-layer="298" from-port="0" to-layer="299" to-port="0" />
+ <edge from-layer="299" from-port="1" to-layer="300" to-port="1" />
+ <edge from-layer="300" from-port="2" to-layer="302" to-port="0" />
+ <edge from-layer="301" from-port="0" to-layer="302" to-port="1" />
+ <edge from-layer="302" from-port="2" to-layer="304" to-port="0" />
+ <edge from-layer="303" from-port="0" to-layer="304" to-port="1" />
+ <edge from-layer="304" from-port="2" to-layer="305" to-port="0" />
+ <edge from-layer="305" from-port="1" to-layer="306" to-port="1" />
+ <edge from-layer="306" from-port="2" to-layer="307" to-port="1" />
+ <edge from-layer="307" from-port="2" to-layer="318" to-port="0" />
+ <edge from-layer="308" from-port="0" to-layer="309" to-port="0" />
+ <edge from-layer="309" from-port="1" to-layer="312" to-port="0" />
+ <edge from-layer="310" from-port="0" to-layer="311" to-port="0" />
+ <edge from-layer="311" from-port="1" to-layer="312" to-port="1" />
+ <edge from-layer="312" from-port="2" to-layer="314" to-port="0" />
+ <edge from-layer="313" from-port="0" to-layer="314" to-port="1" />
+ <edge from-layer="314" from-port="2" to-layer="316" to-port="0" />
+ <edge from-layer="315" from-port="0" to-layer="316" to-port="1" />
+ <edge from-layer="316" from-port="2" to-layer="317" to-port="0" />
+ <edge from-layer="317" from-port="1" to-layer="318" to-port="1" />
+ <edge from-layer="318" from-port="2" to-layer="319" to-port="1" />
+ <edge from-layer="319" from-port="2" to-layer="436" to-port="0" />
+ <edge from-layer="319" from-port="2" to-layer="323" to-port="0" />
+ <edge from-layer="319" from-port="2" to-layer="330" to-port="0" />
+ <edge from-layer="320" from-port="0" to-layer="331" to-port="0" />
+ <edge from-layer="321" from-port="0" to-layer="329" to-port="0" />
+ <edge from-layer="322" from-port="0" to-layer="323" to-port="1" />
+ <edge from-layer="323" from-port="2" to-layer="325" to-port="0" />
+ <edge from-layer="324" from-port="0" to-layer="325" to-port="1" />
+ <edge from-layer="325" from-port="2" to-layer="327" to-port="0" />
+ <edge from-layer="326" from-port="0" to-layer="327" to-port="1" />
+ <edge from-layer="327" from-port="2" to-layer="328" to-port="0" />
+ <edge from-layer="328" from-port="1" to-layer="329" to-port="1" />
+ <edge from-layer="329" from-port="2" to-layer="330" to-port="1" />
+ <edge from-layer="330" from-port="2" to-layer="331" to-port="1" />
+ <edge from-layer="331" from-port="2" to-layer="411" to-port="0" />
+ <edge from-layer="331" from-port="2" to-layer="342" to-port="0" />
+ <edge from-layer="331" from-port="2" to-layer="374" to-port="0" />
+ <edge from-layer="332" from-port="0" to-layer="333" to-port="0" />
+ <edge from-layer="333" from-port="1" to-layer="336" to-port="0" />
+ <edge from-layer="334" from-port="0" to-layer="335" to-port="0" />
+ <edge from-layer="335" from-port="1" to-layer="336" to-port="1" />
+ <edge from-layer="336" from-port="2" to-layer="338" to-port="0" />
+ <edge from-layer="337" from-port="0" to-layer="338" to-port="1" />
+ <edge from-layer="338" from-port="2" to-layer="340" to-port="0" />
+ <edge from-layer="339" from-port="0" to-layer="340" to-port="1" />
+ <edge from-layer="340" from-port="2" to-layer="341" to-port="0" />
+ <edge from-layer="341" from-port="1" to-layer="342" to-port="1" />
+ <edge from-layer="342" from-port="2" to-layer="344" to-port="0" />
+ <edge from-layer="343" from-port="0" to-layer="344" to-port="1" />
+ <edge from-layer="344" from-port="2" to-layer="346" to-port="0" />
+ <edge from-layer="345" from-port="0" to-layer="346" to-port="1" />
+ <edge from-layer="346" from-port="2" to-layer="347" to-port="0" />
+ <edge from-layer="346" from-port="2" to-layer="351" to-port="0" />
+ <edge from-layer="346" from-port="2" to-layer="357" to-port="0" />
+ <edge from-layer="347" from-port="2" to-layer="360" to-port="0" />
+ <edge from-layer="348" from-port="0" to-layer="351" to-port="1" />
+ <edge from-layer="349" from-port="0" to-layer="351" to-port="2" />
+ <edge from-layer="350" from-port="0" to-layer="351" to-port="3" />
+ <edge from-layer="351" from-port="5" to-layer="353" to-port="0" />
+ <edge from-layer="352" from-port="0" to-layer="353" to-port="1" />
+ <edge from-layer="353" from-port="2" to-layer="358" to-port="0" />
+ <edge from-layer="354" from-port="0" to-layer="357" to-port="1" />
+ <edge from-layer="355" from-port="0" to-layer="357" to-port="2" />
+ <edge from-layer="356" from-port="0" to-layer="357" to-port="3" />
+ <edge from-layer="357" from-port="5" to-layer="358" to-port="1" />
+ <edge from-layer="358" from-port="2" to-layer="359" to-port="0" />
+ <edge from-layer="359" from-port="2" to-layer="360" to-port="1" />
+ <edge from-layer="360" from-port="2" to-layer="421" to-port="0" />
+ <edge from-layer="361" from-port="1" to-layer="363" to-port="0" />
+ <edge from-layer="362" from-port="0" to-layer="363" to-port="2" />
+ <edge from-layer="363" from-port="3" to-layer="393" to-port="0" />
+ <edge from-layer="364" from-port="0" to-layer="365" to-port="0" />
+ <edge from-layer="365" from-port="1" to-layer="368" to-port="0" />
+ <edge from-layer="366" from-port="0" to-layer="367" to-port="0" />
+ <edge from-layer="367" from-port="1" to-layer="368" to-port="1" />
+ <edge from-layer="368" from-port="2" to-layer="370" to-port="0" />
+ <edge from-layer="369" from-port="0" to-layer="370" to-port="1" />
+ <edge from-layer="370" from-port="2" to-layer="372" to-port="0" />
+ <edge from-layer="371" from-port="0" to-layer="372" to-port="1" />
+ <edge from-layer="372" from-port="2" to-layer="373" to-port="0" />
+ <edge from-layer="373" from-port="1" to-layer="374" to-port="1" />
+ <edge from-layer="374" from-port="2" to-layer="376" to-port="0" />
+ <edge from-layer="375" from-port="0" to-layer="376" to-port="1" />
+ <edge from-layer="376" from-port="2" to-layer="378" to-port="0" />
+ <edge from-layer="377" from-port="0" to-layer="378" to-port="1" />
+ <edge from-layer="378" from-port="2" to-layer="389" to-port="0" />
+ <edge from-layer="378" from-port="2" to-layer="379" to-port="0" />
+ <edge from-layer="378" from-port="2" to-layer="383" to-port="0" />
+ <edge from-layer="379" from-port="2" to-layer="392" to-port="0" />
+ <edge from-layer="380" from-port="0" to-layer="383" to-port="1" />
+ <edge from-layer="381" from-port="0" to-layer="383" to-port="2" />
+ <edge from-layer="382" from-port="0" to-layer="383" to-port="3" />
+ <edge from-layer="383" from-port="5" to-layer="385" to-port="0" />
+ <edge from-layer="384" from-port="0" to-layer="385" to-port="1" />
+ <edge from-layer="385" from-port="2" to-layer="390" to-port="0" />
+ <edge from-layer="386" from-port="0" to-layer="389" to-port="1" />
+ <edge from-layer="387" from-port="0" to-layer="389" to-port="2" />
+ <edge from-layer="388" from-port="0" to-layer="389" to-port="3" />
+ <edge from-layer="389" from-port="5" to-layer="390" to-port="1" />
+ <edge from-layer="390" from-port="2" to-layer="391" to-port="0" />
+ <edge from-layer="391" from-port="2" to-layer="392" to-port="1" />
+ <edge from-layer="392" from-port="2" to-layer="393" to-port="1" />
+ <edge from-layer="393" from-port="2" to-layer="394" to-port="0" />
+ <edge from-layer="393" from-port="2" to-layer="2719" to-port="0" />
+ <edge from-layer="394" from-port="2" to-layer="395" to-port="0" />
+ <edge from-layer="395" from-port="2" to-layer="397" to-port="0" />
+ <edge from-layer="396" from-port="0" to-layer="397" to-port="1" />
+ <edge from-layer="397" from-port="2" to-layer="421" to-port="1" />
+ <edge from-layer="398" from-port="1" to-layer="400" to-port="0" />
+ <edge from-layer="399" from-port="0" to-layer="400" to-port="2" />
+ <edge from-layer="400" from-port="3" to-layer="416" to-port="0" />
+ <edge from-layer="401" from-port="0" to-layer="402" to-port="0" />
+ <edge from-layer="402" from-port="1" to-layer="405" to-port="0" />
+ <edge from-layer="403" from-port="0" to-layer="404" to-port="0" />
+ <edge from-layer="404" from-port="1" to-layer="405" to-port="1" />
+ <edge from-layer="405" from-port="2" to-layer="407" to-port="0" />
+ <edge from-layer="406" from-port="0" to-layer="407" to-port="1" />
+ <edge from-layer="407" from-port="2" to-layer="409" to-port="0" />
+ <edge from-layer="408" from-port="0" to-layer="409" to-port="1" />
+ <edge from-layer="409" from-port="2" to-layer="410" to-port="0" />
+ <edge from-layer="410" from-port="1" to-layer="411" to-port="1" />
+ <edge from-layer="411" from-port="2" to-layer="413" to-port="0" />
+ <edge from-layer="412" from-port="0" to-layer="413" to-port="1" />
+ <edge from-layer="413" from-port="2" to-layer="415" to-port="0" />
+ <edge from-layer="414" from-port="0" to-layer="415" to-port="1" />
+ <edge from-layer="415" from-port="2" to-layer="416" to-port="1" />
+ <edge from-layer="416" from-port="2" to-layer="417" to-port="0" />
+ <edge from-layer="416" from-port="2" to-layer="2718" to-port="0" />
+ <edge from-layer="417" from-port="2" to-layer="418" to-port="0" />
+ <edge from-layer="418" from-port="2" to-layer="420" to-port="0" />
+ <edge from-layer="419" from-port="0" to-layer="420" to-port="1" />
+ <edge from-layer="420" from-port="2" to-layer="421" to-port="2" />
+ <edge from-layer="421" from-port="4" to-layer="423" to-port="0" />
+ <edge from-layer="422" from-port="0" to-layer="423" to-port="1" />
+ <edge from-layer="423" from-port="2" to-layer="424" to-port="0" />
+ <edge from-layer="424" from-port="2" to-layer="435" to-port="0" />
+ <edge from-layer="425" from-port="0" to-layer="426" to-port="0" />
+ <edge from-layer="426" from-port="1" to-layer="429" to-port="0" />
+ <edge from-layer="427" from-port="0" to-layer="428" to-port="0" />
+ <edge from-layer="428" from-port="1" to-layer="429" to-port="1" />
+ <edge from-layer="429" from-port="2" to-layer="431" to-port="0" />
+ <edge from-layer="430" from-port="0" to-layer="431" to-port="1" />
+ <edge from-layer="431" from-port="2" to-layer="433" to-port="0" />
+ <edge from-layer="432" from-port="0" to-layer="433" to-port="1" />
+ <edge from-layer="433" from-port="2" to-layer="434" to-port="0" />
+ <edge from-layer="434" from-port="1" to-layer="435" to-port="1" />
+ <edge from-layer="435" from-port="2" to-layer="436" to-port="1" />
+ <edge from-layer="436" from-port="2" to-layer="484" to-port="0" />
+ <edge from-layer="436" from-port="2" to-layer="447" to-port="0" />
+ <edge from-layer="436" from-port="2" to-layer="440" to-port="0" />
+ <edge from-layer="437" from-port="0" to-layer="448" to-port="0" />
+ <edge from-layer="438" from-port="0" to-layer="446" to-port="0" />
+ <edge from-layer="439" from-port="0" to-layer="440" to-port="1" />
+ <edge from-layer="440" from-port="2" to-layer="442" to-port="0" />
+ <edge from-layer="441" from-port="0" to-layer="442" to-port="1" />
+ <edge from-layer="442" from-port="2" to-layer="444" to-port="0" />
+ <edge from-layer="443" from-port="0" to-layer="444" to-port="1" />
+ <edge from-layer="444" from-port="2" to-layer="445" to-port="0" />
+ <edge from-layer="445" from-port="1" to-layer="446" to-port="1" />
+ <edge from-layer="446" from-port="2" to-layer="447" to-port="1" />
+ <edge from-layer="447" from-port="2" to-layer="448" to-port="1" />
+ <edge from-layer="448" from-port="2" to-layer="459" to-port="0" />
+ <edge from-layer="448" from-port="2" to-layer="471" to-port="0" />
+ <edge from-layer="449" from-port="0" to-layer="450" to-port="0" />
+ <edge from-layer="450" from-port="1" to-layer="453" to-port="0" />
+ <edge from-layer="451" from-port="0" to-layer="452" to-port="0" />
+ <edge from-layer="452" from-port="1" to-layer="453" to-port="1" />
+ <edge from-layer="453" from-port="2" to-layer="455" to-port="0" />
+ <edge from-layer="454" from-port="0" to-layer="455" to-port="1" />
+ <edge from-layer="455" from-port="2" to-layer="457" to-port="0" />
+ <edge from-layer="456" from-port="0" to-layer="457" to-port="1" />
+ <edge from-layer="457" from-port="2" to-layer="458" to-port="0" />
+ <edge from-layer="458" from-port="1" to-layer="459" to-port="1" />
+ <edge from-layer="459" from-port="2" to-layer="460" to-port="0" />
+ <edge from-layer="460" from-port="1" to-layer="472" to-port="0" />
+ <edge from-layer="461" from-port="0" to-layer="462" to-port="0" />
+ <edge from-layer="462" from-port="1" to-layer="465" to-port="0" />
+ <edge from-layer="463" from-port="0" to-layer="464" to-port="0" />
+ <edge from-layer="464" from-port="1" to-layer="465" to-port="1" />
+ <edge from-layer="465" from-port="2" to-layer="467" to-port="0" />
+ <edge from-layer="466" from-port="0" to-layer="467" to-port="1" />
+ <edge from-layer="467" from-port="2" to-layer="469" to-port="0" />
+ <edge from-layer="468" from-port="0" to-layer="469" to-port="1" />
+ <edge from-layer="469" from-port="2" to-layer="470" to-port="0" />
+ <edge from-layer="470" from-port="1" to-layer="471" to-port="1" />
+ <edge from-layer="471" from-port="2" to-layer="472" to-port="1" />
+ <edge from-layer="472" from-port="2" to-layer="483" to-port="0" />
+ <edge from-layer="473" from-port="0" to-layer="474" to-port="0" />
+ <edge from-layer="474" from-port="1" to-layer="477" to-port="0" />
+ <edge from-layer="475" from-port="0" to-layer="476" to-port="0" />
+ <edge from-layer="476" from-port="1" to-layer="477" to-port="1" />
+ <edge from-layer="477" from-port="2" to-layer="479" to-port="0" />
+ <edge from-layer="478" from-port="0" to-layer="479" to-port="1" />
+ <edge from-layer="479" from-port="2" to-layer="481" to-port="0" />
+ <edge from-layer="480" from-port="0" to-layer="481" to-port="1" />
+ <edge from-layer="481" from-port="2" to-layer="482" to-port="0" />
+ <edge from-layer="482" from-port="1" to-layer="483" to-port="1" />
+ <edge from-layer="483" from-port="2" to-layer="484" to-port="1" />
+ <edge from-layer="484" from-port="2" to-layer="601" to-port="0" />
+ <edge from-layer="484" from-port="2" to-layer="488" to-port="0" />
+ <edge from-layer="484" from-port="2" to-layer="495" to-port="0" />
+ <edge from-layer="485" from-port="0" to-layer="496" to-port="0" />
+ <edge from-layer="486" from-port="0" to-layer="494" to-port="0" />
+ <edge from-layer="487" from-port="0" to-layer="488" to-port="1" />
+ <edge from-layer="488" from-port="2" to-layer="490" to-port="0" />
+ <edge from-layer="489" from-port="0" to-layer="490" to-port="1" />
+ <edge from-layer="490" from-port="2" to-layer="492" to-port="0" />
+ <edge from-layer="491" from-port="0" to-layer="492" to-port="1" />
+ <edge from-layer="492" from-port="2" to-layer="493" to-port="0" />
+ <edge from-layer="493" from-port="1" to-layer="494" to-port="1" />
+ <edge from-layer="494" from-port="2" to-layer="495" to-port="1" />
+ <edge from-layer="495" from-port="2" to-layer="496" to-port="1" />
+ <edge from-layer="496" from-port="2" to-layer="576" to-port="0" />
+ <edge from-layer="496" from-port="2" to-layer="539" to-port="0" />
+ <edge from-layer="496" from-port="2" to-layer="507" to-port="0" />
+ <edge from-layer="497" from-port="0" to-layer="498" to-port="0" />
+ <edge from-layer="498" from-port="1" to-layer="501" to-port="0" />
+ <edge from-layer="499" from-port="0" to-layer="500" to-port="0" />
+ <edge from-layer="500" from-port="1" to-layer="501" to-port="1" />
+ <edge from-layer="501" from-port="2" to-layer="503" to-port="0" />
+ <edge from-layer="502" from-port="0" to-layer="503" to-port="1" />
+ <edge from-layer="503" from-port="2" to-layer="505" to-port="0" />
+ <edge from-layer="504" from-port="0" to-layer="505" to-port="1" />
+ <edge from-layer="505" from-port="2" to-layer="506" to-port="0" />
+ <edge from-layer="506" from-port="1" to-layer="507" to-port="1" />
+ <edge from-layer="507" from-port="2" to-layer="509" to-port="0" />
+ <edge from-layer="508" from-port="0" to-layer="509" to-port="1" />
+ <edge from-layer="509" from-port="2" to-layer="511" to-port="0" />
+ <edge from-layer="510" from-port="0" to-layer="511" to-port="1" />
+ <edge from-layer="511" from-port="2" to-layer="516" to-port="0" />
+ <edge from-layer="511" from-port="2" to-layer="522" to-port="0" />
+ <edge from-layer="511" from-port="2" to-layer="512" to-port="0" />
+ <edge from-layer="512" from-port="2" to-layer="525" to-port="0" />
+ <edge from-layer="513" from-port="0" to-layer="516" to-port="1" />
+ <edge from-layer="514" from-port="0" to-layer="516" to-port="2" />
+ <edge from-layer="515" from-port="0" to-layer="516" to-port="3" />
+ <edge from-layer="516" from-port="5" to-layer="518" to-port="0" />
+ <edge from-layer="517" from-port="0" to-layer="518" to-port="1" />
+ <edge from-layer="518" from-port="2" to-layer="523" to-port="0" />
+ <edge from-layer="519" from-port="0" to-layer="522" to-port="1" />
+ <edge from-layer="520" from-port="0" to-layer="522" to-port="2" />
+ <edge from-layer="521" from-port="0" to-layer="522" to-port="3" />
+ <edge from-layer="522" from-port="5" to-layer="523" to-port="1" />
+ <edge from-layer="523" from-port="2" to-layer="524" to-port="0" />
+ <edge from-layer="524" from-port="2" to-layer="525" to-port="1" />
+ <edge from-layer="525" from-port="2" to-layer="586" to-port="0" />
+ <edge from-layer="526" from-port="1" to-layer="528" to-port="0" />
+ <edge from-layer="527" from-port="0" to-layer="528" to-port="2" />
+ <edge from-layer="528" from-port="3" to-layer="558" to-port="0" />
+ <edge from-layer="529" from-port="0" to-layer="530" to-port="0" />
+ <edge from-layer="530" from-port="1" to-layer="533" to-port="0" />
+ <edge from-layer="531" from-port="0" to-layer="532" to-port="0" />
+ <edge from-layer="532" from-port="1" to-layer="533" to-port="1" />
+ <edge from-layer="533" from-port="2" to-layer="535" to-port="0" />
+ <edge from-layer="534" from-port="0" to-layer="535" to-port="1" />
+ <edge from-layer="535" from-port="2" to-layer="537" to-port="0" />
+ <edge from-layer="536" from-port="0" to-layer="537" to-port="1" />
+ <edge from-layer="537" from-port="2" to-layer="538" to-port="0" />
+ <edge from-layer="538" from-port="1" to-layer="539" to-port="1" />
+ <edge from-layer="539" from-port="2" to-layer="541" to-port="0" />
+ <edge from-layer="540" from-port="0" to-layer="541" to-port="1" />
+ <edge from-layer="541" from-port="2" to-layer="543" to-port="0" />
+ <edge from-layer="542" from-port="0" to-layer="543" to-port="1" />
+ <edge from-layer="543" from-port="2" to-layer="544" to-port="0" />
+ <edge from-layer="543" from-port="2" to-layer="554" to-port="0" />
+ <edge from-layer="543" from-port="2" to-layer="548" to-port="0" />
+ <edge from-layer="544" from-port="2" to-layer="557" to-port="0" />
+ <edge from-layer="545" from-port="0" to-layer="548" to-port="1" />
+ <edge from-layer="546" from-port="0" to-layer="548" to-port="2" />
+ <edge from-layer="547" from-port="0" to-layer="548" to-port="3" />
+ <edge from-layer="548" from-port="5" to-layer="550" to-port="0" />
+ <edge from-layer="549" from-port="0" to-layer="550" to-port="1" />
+ <edge from-layer="550" from-port="2" to-layer="555" to-port="0" />
+ <edge from-layer="551" from-port="0" to-layer="554" to-port="1" />
+ <edge from-layer="552" from-port="0" to-layer="554" to-port="2" />
+ <edge from-layer="553" from-port="0" to-layer="554" to-port="3" />
+ <edge from-layer="554" from-port="5" to-layer="555" to-port="1" />
+ <edge from-layer="555" from-port="2" to-layer="556" to-port="0" />
+ <edge from-layer="556" from-port="2" to-layer="557" to-port="1" />
+ <edge from-layer="557" from-port="2" to-layer="558" to-port="1" />
+ <edge from-layer="558" from-port="2" to-layer="559" to-port="0" />
+ <edge from-layer="558" from-port="2" to-layer="1715" to-port="0" />
+ <edge from-layer="559" from-port="2" to-layer="560" to-port="0" />
+ <edge from-layer="560" from-port="2" to-layer="562" to-port="0" />
+ <edge from-layer="561" from-port="0" to-layer="562" to-port="1" />
+ <edge from-layer="562" from-port="2" to-layer="586" to-port="1" />
+ <edge from-layer="563" from-port="1" to-layer="565" to-port="0" />
+ <edge from-layer="564" from-port="0" to-layer="565" to-port="2" />
+ <edge from-layer="565" from-port="3" to-layer="581" to-port="0" />
+ <edge from-layer="566" from-port="0" to-layer="567" to-port="0" />
+ <edge from-layer="567" from-port="1" to-layer="570" to-port="0" />
+ <edge from-layer="568" from-port="0" to-layer="569" to-port="0" />
+ <edge from-layer="569" from-port="1" to-layer="570" to-port="1" />
+ <edge from-layer="570" from-port="2" to-layer="572" to-port="0" />
+ <edge from-layer="571" from-port="0" to-layer="572" to-port="1" />
+ <edge from-layer="572" from-port="2" to-layer="574" to-port="0" />
+ <edge from-layer="573" from-port="0" to-layer="574" to-port="1" />
+ <edge from-layer="574" from-port="2" to-layer="575" to-port="0" />
+ <edge from-layer="575" from-port="1" to-layer="576" to-port="1" />
+ <edge from-layer="576" from-port="2" to-layer="578" to-port="0" />
+ <edge from-layer="577" from-port="0" to-layer="578" to-port="1" />
+ <edge from-layer="578" from-port="2" to-layer="580" to-port="0" />
+ <edge from-layer="579" from-port="0" to-layer="580" to-port="1" />
+ <edge from-layer="580" from-port="2" to-layer="581" to-port="1" />
+ <edge from-layer="581" from-port="2" to-layer="1714" to-port="0" />
+ <edge from-layer="581" from-port="2" to-layer="582" to-port="0" />
+ <edge from-layer="582" from-port="2" to-layer="583" to-port="0" />
+ <edge from-layer="583" from-port="2" to-layer="585" to-port="0" />
+ <edge from-layer="584" from-port="0" to-layer="585" to-port="1" />
+ <edge from-layer="585" from-port="2" to-layer="586" to-port="2" />
+ <edge from-layer="586" from-port="4" to-layer="588" to-port="0" />
+ <edge from-layer="587" from-port="0" to-layer="588" to-port="1" />
+ <edge from-layer="588" from-port="2" to-layer="589" to-port="0" />
+ <edge from-layer="589" from-port="2" to-layer="600" to-port="0" />
+ <edge from-layer="590" from-port="0" to-layer="591" to-port="0" />
+ <edge from-layer="591" from-port="1" to-layer="594" to-port="0" />
+ <edge from-layer="592" from-port="0" to-layer="593" to-port="0" />
+ <edge from-layer="593" from-port="1" to-layer="594" to-port="1" />
+ <edge from-layer="594" from-port="2" to-layer="596" to-port="0" />
+ <edge from-layer="595" from-port="0" to-layer="596" to-port="1" />
+ <edge from-layer="596" from-port="2" to-layer="598" to-port="0" />
+ <edge from-layer="597" from-port="0" to-layer="598" to-port="1" />
+ <edge from-layer="598" from-port="2" to-layer="599" to-port="0" />
+ <edge from-layer="599" from-port="1" to-layer="600" to-port="1" />
+ <edge from-layer="600" from-port="2" to-layer="601" to-port="1" />
+ <edge from-layer="601" from-port="2" to-layer="605" to-port="0" />
+ <edge from-layer="601" from-port="2" to-layer="612" to-port="0" />
+ <edge from-layer="601" from-port="2" to-layer="649" to-port="0" />
+ <edge from-layer="602" from-port="0" to-layer="613" to-port="0" />
+ <edge from-layer="603" from-port="0" to-layer="611" to-port="0" />
+ <edge from-layer="604" from-port="0" to-layer="605" to-port="1" />
+ <edge from-layer="605" from-port="2" to-layer="607" to-port="0" />
+ <edge from-layer="606" from-port="0" to-layer="607" to-port="1" />
+ <edge from-layer="607" from-port="2" to-layer="609" to-port="0" />
+ <edge from-layer="608" from-port="0" to-layer="609" to-port="1" />
+ <edge from-layer="609" from-port="2" to-layer="610" to-port="0" />
+ <edge from-layer="610" from-port="1" to-layer="611" to-port="1" />
+ <edge from-layer="611" from-port="2" to-layer="612" to-port="1" />
+ <edge from-layer="612" from-port="2" to-layer="613" to-port="1" />
+ <edge from-layer="613" from-port="2" to-layer="636" to-port="0" />
+ <edge from-layer="613" from-port="2" to-layer="624" to-port="0" />
+ <edge from-layer="614" from-port="0" to-layer="615" to-port="0" />
+ <edge from-layer="615" from-port="1" to-layer="618" to-port="0" />
+ <edge from-layer="616" from-port="0" to-layer="617" to-port="0" />
+ <edge from-layer="617" from-port="1" to-layer="618" to-port="1" />
+ <edge from-layer="618" from-port="2" to-layer="620" to-port="0" />
+ <edge from-layer="619" from-port="0" to-layer="620" to-port="1" />
+ <edge from-layer="620" from-port="2" to-layer="622" to-port="0" />
+ <edge from-layer="621" from-port="0" to-layer="622" to-port="1" />
+ <edge from-layer="622" from-port="2" to-layer="623" to-port="0" />
+ <edge from-layer="623" from-port="1" to-layer="624" to-port="1" />
+ <edge from-layer="624" from-port="2" to-layer="625" to-port="0" />
+ <edge from-layer="625" from-port="1" to-layer="637" to-port="0" />
+ <edge from-layer="626" from-port="0" to-layer="627" to-port="0" />
+ <edge from-layer="627" from-port="1" to-layer="630" to-port="0" />
+ <edge from-layer="628" from-port="0" to-layer="629" to-port="0" />
+ <edge from-layer="629" from-port="1" to-layer="630" to-port="1" />
+ <edge from-layer="630" from-port="2" to-layer="632" to-port="0" />
+ <edge from-layer="631" from-port="0" to-layer="632" to-port="1" />
+ <edge from-layer="632" from-port="2" to-layer="634" to-port="0" />
+ <edge from-layer="633" from-port="0" to-layer="634" to-port="1" />
+ <edge from-layer="634" from-port="2" to-layer="635" to-port="0" />
+ <edge from-layer="635" from-port="1" to-layer="636" to-port="1" />
+ <edge from-layer="636" from-port="2" to-layer="637" to-port="1" />
+ <edge from-layer="637" from-port="2" to-layer="648" to-port="0" />
+ <edge from-layer="638" from-port="0" to-layer="639" to-port="0" />
+ <edge from-layer="639" from-port="1" to-layer="642" to-port="0" />
+ <edge from-layer="640" from-port="0" to-layer="641" to-port="0" />
+ <edge from-layer="641" from-port="1" to-layer="642" to-port="1" />
+ <edge from-layer="642" from-port="2" to-layer="644" to-port="0" />
+ <edge from-layer="643" from-port="0" to-layer="644" to-port="1" />
+ <edge from-layer="644" from-port="2" to-layer="646" to-port="0" />
+ <edge from-layer="645" from-port="0" to-layer="646" to-port="1" />
+ <edge from-layer="646" from-port="2" to-layer="647" to-port="0" />
+ <edge from-layer="647" from-port="1" to-layer="648" to-port="1" />
+ <edge from-layer="648" from-port="2" to-layer="649" to-port="1" />
+ <edge from-layer="649" from-port="2" to-layer="766" to-port="0" />
+ <edge from-layer="649" from-port="2" to-layer="653" to-port="0" />
+ <edge from-layer="649" from-port="2" to-layer="660" to-port="0" />
+ <edge from-layer="650" from-port="0" to-layer="661" to-port="0" />
+ <edge from-layer="651" from-port="0" to-layer="659" to-port="0" />
+ <edge from-layer="652" from-port="0" to-layer="653" to-port="1" />
+ <edge from-layer="653" from-port="2" to-layer="655" to-port="0" />
+ <edge from-layer="654" from-port="0" to-layer="655" to-port="1" />
+ <edge from-layer="655" from-port="2" to-layer="657" to-port="0" />
+ <edge from-layer="656" from-port="0" to-layer="657" to-port="1" />
+ <edge from-layer="657" from-port="2" to-layer="658" to-port="0" />
+ <edge from-layer="658" from-port="1" to-layer="659" to-port="1" />
+ <edge from-layer="659" from-port="2" to-layer="660" to-port="1" />
+ <edge from-layer="660" from-port="2" to-layer="661" to-port="1" />
+ <edge from-layer="661" from-port="2" to-layer="741" to-port="0" />
+ <edge from-layer="661" from-port="2" to-layer="704" to-port="0" />
+ <edge from-layer="661" from-port="2" to-layer="672" to-port="0" />
+ <edge from-layer="662" from-port="0" to-layer="663" to-port="0" />
+ <edge from-layer="663" from-port="1" to-layer="666" to-port="0" />
+ <edge from-layer="664" from-port="0" to-layer="665" to-port="0" />
+ <edge from-layer="665" from-port="1" to-layer="666" to-port="1" />
+ <edge from-layer="666" from-port="2" to-layer="668" to-port="0" />
+ <edge from-layer="667" from-port="0" to-layer="668" to-port="1" />
+ <edge from-layer="668" from-port="2" to-layer="670" to-port="0" />
+ <edge from-layer="669" from-port="0" to-layer="670" to-port="1" />
+ <edge from-layer="670" from-port="2" to-layer="671" to-port="0" />
+ <edge from-layer="671" from-port="1" to-layer="672" to-port="1" />
+ <edge from-layer="672" from-port="2" to-layer="674" to-port="0" />
+ <edge from-layer="673" from-port="0" to-layer="674" to-port="1" />
+ <edge from-layer="674" from-port="2" to-layer="676" to-port="0" />
+ <edge from-layer="675" from-port="0" to-layer="676" to-port="1" />
+ <edge from-layer="676" from-port="2" to-layer="677" to-port="0" />
+ <edge from-layer="676" from-port="2" to-layer="681" to-port="0" />
+ <edge from-layer="676" from-port="2" to-layer="687" to-port="0" />
+ <edge from-layer="677" from-port="2" to-layer="690" to-port="0" />
+ <edge from-layer="678" from-port="0" to-layer="681" to-port="1" />
+ <edge from-layer="679" from-port="0" to-layer="681" to-port="2" />
+ <edge from-layer="680" from-port="0" to-layer="681" to-port="3" />
+ <edge from-layer="681" from-port="5" to-layer="683" to-port="0" />
+ <edge from-layer="682" from-port="0" to-layer="683" to-port="1" />
+ <edge from-layer="683" from-port="2" to-layer="688" to-port="0" />
+ <edge from-layer="684" from-port="0" to-layer="687" to-port="1" />
+ <edge from-layer="685" from-port="0" to-layer="687" to-port="2" />
+ <edge from-layer="686" from-port="0" to-layer="687" to-port="3" />
+ <edge from-layer="687" from-port="5" to-layer="688" to-port="1" />
+ <edge from-layer="688" from-port="2" to-layer="689" to-port="0" />
+ <edge from-layer="689" from-port="2" to-layer="690" to-port="1" />
+ <edge from-layer="690" from-port="2" to-layer="751" to-port="0" />
+ <edge from-layer="691" from-port="1" to-layer="693" to-port="0" />
+ <edge from-layer="692" from-port="0" to-layer="693" to-port="2" />
+ <edge from-layer="693" from-port="3" to-layer="723" to-port="0" />
+ <edge from-layer="694" from-port="0" to-layer="695" to-port="0" />
+ <edge from-layer="695" from-port="1" to-layer="698" to-port="0" />
+ <edge from-layer="696" from-port="0" to-layer="697" to-port="0" />
+ <edge from-layer="697" from-port="1" to-layer="698" to-port="1" />
+ <edge from-layer="698" from-port="2" to-layer="700" to-port="0" />
+ <edge from-layer="699" from-port="0" to-layer="700" to-port="1" />
+ <edge from-layer="700" from-port="2" to-layer="702" to-port="0" />
+ <edge from-layer="701" from-port="0" to-layer="702" to-port="1" />
+ <edge from-layer="702" from-port="2" to-layer="703" to-port="0" />
+ <edge from-layer="703" from-port="1" to-layer="704" to-port="1" />
+ <edge from-layer="704" from-port="2" to-layer="706" to-port="0" />
+ <edge from-layer="705" from-port="0" to-layer="706" to-port="1" />
+ <edge from-layer="706" from-port="2" to-layer="708" to-port="0" />
+ <edge from-layer="707" from-port="0" to-layer="708" to-port="1" />
+ <edge from-layer="708" from-port="2" to-layer="709" to-port="0" />
+ <edge from-layer="708" from-port="2" to-layer="713" to-port="0" />
+ <edge from-layer="708" from-port="2" to-layer="719" to-port="0" />
+ <edge from-layer="709" from-port="2" to-layer="722" to-port="0" />
+ <edge from-layer="710" from-port="0" to-layer="713" to-port="1" />
+ <edge from-layer="711" from-port="0" to-layer="713" to-port="2" />
+ <edge from-layer="712" from-port="0" to-layer="713" to-port="3" />
+ <edge from-layer="713" from-port="5" to-layer="715" to-port="0" />
+ <edge from-layer="714" from-port="0" to-layer="715" to-port="1" />
+ <edge from-layer="715" from-port="2" to-layer="720" to-port="0" />
+ <edge from-layer="716" from-port="0" to-layer="719" to-port="1" />
+ <edge from-layer="717" from-port="0" to-layer="719" to-port="2" />
+ <edge from-layer="718" from-port="0" to-layer="719" to-port="3" />
+ <edge from-layer="719" from-port="5" to-layer="720" to-port="1" />
+ <edge from-layer="720" from-port="2" to-layer="721" to-port="0" />
+ <edge from-layer="721" from-port="2" to-layer="722" to-port="1" />
+ <edge from-layer="722" from-port="2" to-layer="723" to-port="1" />
+ <edge from-layer="723" from-port="2" to-layer="1713" to-port="0" />
+ <edge from-layer="723" from-port="2" to-layer="724" to-port="0" />
+ <edge from-layer="724" from-port="2" to-layer="725" to-port="0" />
+ <edge from-layer="725" from-port="2" to-layer="727" to-port="0" />
+ <edge from-layer="726" from-port="0" to-layer="727" to-port="1" />
+ <edge from-layer="727" from-port="2" to-layer="751" to-port="1" />
+ <edge from-layer="728" from-port="1" to-layer="730" to-port="0" />
+ <edge from-layer="729" from-port="0" to-layer="730" to-port="2" />
+ <edge from-layer="730" from-port="3" to-layer="746" to-port="0" />
+ <edge from-layer="731" from-port="0" to-layer="732" to-port="0" />
+ <edge from-layer="732" from-port="1" to-layer="735" to-port="0" />
+ <edge from-layer="733" from-port="0" to-layer="734" to-port="0" />
+ <edge from-layer="734" from-port="1" to-layer="735" to-port="1" />
+ <edge from-layer="735" from-port="2" to-layer="737" to-port="0" />
+ <edge from-layer="736" from-port="0" to-layer="737" to-port="1" />
+ <edge from-layer="737" from-port="2" to-layer="739" to-port="0" />
+ <edge from-layer="738" from-port="0" to-layer="739" to-port="1" />
+ <edge from-layer="739" from-port="2" to-layer="740" to-port="0" />
+ <edge from-layer="740" from-port="1" to-layer="741" to-port="1" />
+ <edge from-layer="741" from-port="2" to-layer="743" to-port="0" />
+ <edge from-layer="742" from-port="0" to-layer="743" to-port="1" />
+ <edge from-layer="743" from-port="2" to-layer="745" to-port="0" />
+ <edge from-layer="744" from-port="0" to-layer="745" to-port="1" />
+ <edge from-layer="745" from-port="2" to-layer="746" to-port="1" />
+ <edge from-layer="746" from-port="2" to-layer="747" to-port="0" />
+ <edge from-layer="746" from-port="2" to-layer="1712" to-port="0" />
+ <edge from-layer="747" from-port="2" to-layer="748" to-port="0" />
+ <edge from-layer="748" from-port="2" to-layer="750" to-port="0" />
+ <edge from-layer="749" from-port="0" to-layer="750" to-port="1" />
+ <edge from-layer="750" from-port="2" to-layer="751" to-port="2" />
+ <edge from-layer="751" from-port="4" to-layer="753" to-port="0" />
+ <edge from-layer="752" from-port="0" to-layer="753" to-port="1" />
+ <edge from-layer="753" from-port="2" to-layer="754" to-port="0" />
+ <edge from-layer="754" from-port="2" to-layer="765" to-port="0" />
+ <edge from-layer="755" from-port="0" to-layer="756" to-port="0" />
+ <edge from-layer="756" from-port="1" to-layer="759" to-port="0" />
+ <edge from-layer="757" from-port="0" to-layer="758" to-port="0" />
+ <edge from-layer="758" from-port="1" to-layer="759" to-port="1" />
+ <edge from-layer="759" from-port="2" to-layer="761" to-port="0" />
+ <edge from-layer="760" from-port="0" to-layer="761" to-port="1" />
+ <edge from-layer="761" from-port="2" to-layer="763" to-port="0" />
+ <edge from-layer="762" from-port="0" to-layer="763" to-port="1" />
+ <edge from-layer="763" from-port="2" to-layer="764" to-port="0" />
+ <edge from-layer="764" from-port="1" to-layer="765" to-port="1" />
+ <edge from-layer="765" from-port="2" to-layer="766" to-port="1" />
+ <edge from-layer="766" from-port="2" to-layer="770" to-port="0" />
+ <edge from-layer="766" from-port="2" to-layer="814" to-port="0" />
+ <edge from-layer="766" from-port="2" to-layer="777" to-port="0" />
+ <edge from-layer="767" from-port="0" to-layer="778" to-port="0" />
+ <edge from-layer="768" from-port="0" to-layer="776" to-port="0" />
+ <edge from-layer="769" from-port="0" to-layer="770" to-port="1" />
+ <edge from-layer="770" from-port="2" to-layer="772" to-port="0" />
+ <edge from-layer="771" from-port="0" to-layer="772" to-port="1" />
+ <edge from-layer="772" from-port="2" to-layer="774" to-port="0" />
+ <edge from-layer="773" from-port="0" to-layer="774" to-port="1" />
+ <edge from-layer="774" from-port="2" to-layer="775" to-port="0" />
+ <edge from-layer="775" from-port="1" to-layer="776" to-port="1" />
+ <edge from-layer="776" from-port="2" to-layer="777" to-port="1" />
+ <edge from-layer="777" from-port="2" to-layer="778" to-port="1" />
+ <edge from-layer="778" from-port="2" to-layer="801" to-port="0" />
+ <edge from-layer="778" from-port="2" to-layer="789" to-port="0" />
+ <edge from-layer="779" from-port="0" to-layer="780" to-port="0" />
+ <edge from-layer="780" from-port="1" to-layer="783" to-port="0" />
+ <edge from-layer="781" from-port="0" to-layer="782" to-port="0" />
+ <edge from-layer="782" from-port="1" to-layer="783" to-port="1" />
+ <edge from-layer="783" from-port="2" to-layer="785" to-port="0" />
+ <edge from-layer="784" from-port="0" to-layer="785" to-port="1" />
+ <edge from-layer="785" from-port="2" to-layer="787" to-port="0" />
+ <edge from-layer="786" from-port="0" to-layer="787" to-port="1" />
+ <edge from-layer="787" from-port="2" to-layer="788" to-port="0" />
+ <edge from-layer="788" from-port="1" to-layer="789" to-port="1" />
+ <edge from-layer="789" from-port="2" to-layer="790" to-port="0" />
+ <edge from-layer="790" from-port="1" to-layer="802" to-port="0" />
+ <edge from-layer="791" from-port="0" to-layer="792" to-port="0" />
+ <edge from-layer="792" from-port="1" to-layer="795" to-port="0" />
+ <edge from-layer="793" from-port="0" to-layer="794" to-port="0" />
+ <edge from-layer="794" from-port="1" to-layer="795" to-port="1" />
+ <edge from-layer="795" from-port="2" to-layer="797" to-port="0" />
+ <edge from-layer="796" from-port="0" to-layer="797" to-port="1" />
+ <edge from-layer="797" from-port="2" to-layer="799" to-port="0" />
+ <edge from-layer="798" from-port="0" to-layer="799" to-port="1" />
+ <edge from-layer="799" from-port="2" to-layer="800" to-port="0" />
+ <edge from-layer="800" from-port="1" to-layer="801" to-port="1" />
+ <edge from-layer="801" from-port="2" to-layer="802" to-port="1" />
+ <edge from-layer="802" from-port="2" to-layer="813" to-port="0" />
+ <edge from-layer="803" from-port="0" to-layer="804" to-port="0" />
+ <edge from-layer="804" from-port="1" to-layer="807" to-port="0" />
+ <edge from-layer="805" from-port="0" to-layer="806" to-port="0" />
+ <edge from-layer="806" from-port="1" to-layer="807" to-port="1" />
+ <edge from-layer="807" from-port="2" to-layer="809" to-port="0" />
+ <edge from-layer="808" from-port="0" to-layer="809" to-port="1" />
+ <edge from-layer="809" from-port="2" to-layer="811" to-port="0" />
+ <edge from-layer="810" from-port="0" to-layer="811" to-port="1" />
+ <edge from-layer="811" from-port="2" to-layer="812" to-port="0" />
+ <edge from-layer="812" from-port="1" to-layer="813" to-port="1" />
+ <edge from-layer="813" from-port="2" to-layer="814" to-port="1" />
+ <edge from-layer="814" from-port="2" to-layer="931" to-port="0" />
+ <edge from-layer="814" from-port="2" to-layer="825" to-port="0" />
+ <edge from-layer="814" from-port="2" to-layer="818" to-port="0" />
+ <edge from-layer="815" from-port="0" to-layer="826" to-port="0" />
+ <edge from-layer="816" from-port="0" to-layer="824" to-port="0" />
+ <edge from-layer="817" from-port="0" to-layer="818" to-port="1" />
+ <edge from-layer="818" from-port="2" to-layer="820" to-port="0" />
+ <edge from-layer="819" from-port="0" to-layer="820" to-port="1" />
+ <edge from-layer="820" from-port="2" to-layer="822" to-port="0" />
+ <edge from-layer="821" from-port="0" to-layer="822" to-port="1" />
+ <edge from-layer="822" from-port="2" to-layer="823" to-port="0" />
+ <edge from-layer="823" from-port="1" to-layer="824" to-port="1" />
+ <edge from-layer="824" from-port="2" to-layer="825" to-port="1" />
+ <edge from-layer="825" from-port="2" to-layer="826" to-port="1" />
+ <edge from-layer="826" from-port="2" to-layer="837" to-port="0" />
+ <edge from-layer="826" from-port="2" to-layer="906" to-port="0" />
+ <edge from-layer="826" from-port="2" to-layer="869" to-port="0" />
+ <edge from-layer="827" from-port="0" to-layer="828" to-port="0" />
+ <edge from-layer="828" from-port="1" to-layer="831" to-port="0" />
+ <edge from-layer="829" from-port="0" to-layer="830" to-port="0" />
+ <edge from-layer="830" from-port="1" to-layer="831" to-port="1" />
+ <edge from-layer="831" from-port="2" to-layer="833" to-port="0" />
+ <edge from-layer="832" from-port="0" to-layer="833" to-port="1" />
+ <edge from-layer="833" from-port="2" to-layer="835" to-port="0" />
+ <edge from-layer="834" from-port="0" to-layer="835" to-port="1" />
+ <edge from-layer="835" from-port="2" to-layer="836" to-port="0" />
+ <edge from-layer="836" from-port="1" to-layer="837" to-port="1" />
+ <edge from-layer="837" from-port="2" to-layer="839" to-port="0" />
+ <edge from-layer="838" from-port="0" to-layer="839" to-port="1" />
+ <edge from-layer="839" from-port="2" to-layer="841" to-port="0" />
+ <edge from-layer="840" from-port="0" to-layer="841" to-port="1" />
+ <edge from-layer="841" from-port="2" to-layer="852" to-port="0" />
+ <edge from-layer="841" from-port="2" to-layer="842" to-port="0" />
+ <edge from-layer="841" from-port="2" to-layer="846" to-port="0" />
+ <edge from-layer="842" from-port="2" to-layer="855" to-port="0" />
+ <edge from-layer="843" from-port="0" to-layer="846" to-port="1" />
+ <edge from-layer="844" from-port="0" to-layer="846" to-port="2" />
+ <edge from-layer="845" from-port="0" to-layer="846" to-port="3" />
+ <edge from-layer="846" from-port="5" to-layer="848" to-port="0" />
+ <edge from-layer="847" from-port="0" to-layer="848" to-port="1" />
+ <edge from-layer="848" from-port="2" to-layer="853" to-port="0" />
+ <edge from-layer="849" from-port="0" to-layer="852" to-port="1" />
+ <edge from-layer="850" from-port="0" to-layer="852" to-port="2" />
+ <edge from-layer="851" from-port="0" to-layer="852" to-port="3" />
+ <edge from-layer="852" from-port="5" to-layer="853" to-port="1" />
+ <edge from-layer="853" from-port="2" to-layer="854" to-port="0" />
+ <edge from-layer="854" from-port="2" to-layer="855" to-port="1" />
+ <edge from-layer="855" from-port="2" to-layer="916" to-port="0" />
+ <edge from-layer="856" from-port="1" to-layer="858" to-port="0" />
+ <edge from-layer="857" from-port="0" to-layer="858" to-port="2" />
+ <edge from-layer="858" from-port="3" to-layer="888" to-port="0" />
+ <edge from-layer="859" from-port="0" to-layer="860" to-port="0" />
+ <edge from-layer="860" from-port="1" to-layer="863" to-port="0" />
+ <edge from-layer="861" from-port="0" to-layer="862" to-port="0" />
+ <edge from-layer="862" from-port="1" to-layer="863" to-port="1" />
+ <edge from-layer="863" from-port="2" to-layer="865" to-port="0" />
+ <edge from-layer="864" from-port="0" to-layer="865" to-port="1" />
+ <edge from-layer="865" from-port="2" to-layer="867" to-port="0" />
+ <edge from-layer="866" from-port="0" to-layer="867" to-port="1" />
+ <edge from-layer="867" from-port="2" to-layer="868" to-port="0" />
+ <edge from-layer="868" from-port="1" to-layer="869" to-port="1" />
+ <edge from-layer="869" from-port="2" to-layer="871" to-port="0" />
+ <edge from-layer="870" from-port="0" to-layer="871" to-port="1" />
+ <edge from-layer="871" from-port="2" to-layer="873" to-port="0" />
+ <edge from-layer="872" from-port="0" to-layer="873" to-port="1" />
+ <edge from-layer="873" from-port="2" to-layer="884" to-port="0" />
+ <edge from-layer="873" from-port="2" to-layer="874" to-port="0" />
+ <edge from-layer="873" from-port="2" to-layer="878" to-port="0" />
+ <edge from-layer="874" from-port="2" to-layer="887" to-port="0" />
+ <edge from-layer="875" from-port="0" to-layer="878" to-port="1" />
+ <edge from-layer="876" from-port="0" to-layer="878" to-port="2" />
+ <edge from-layer="877" from-port="0" to-layer="878" to-port="3" />
+ <edge from-layer="878" from-port="5" to-layer="880" to-port="0" />
+ <edge from-layer="879" from-port="0" to-layer="880" to-port="1" />
+ <edge from-layer="880" from-port="2" to-layer="885" to-port="0" />
+ <edge from-layer="881" from-port="0" to-layer="884" to-port="1" />
+ <edge from-layer="882" from-port="0" to-layer="884" to-port="2" />
+ <edge from-layer="883" from-port="0" to-layer="884" to-port="3" />
+ <edge from-layer="884" from-port="5" to-layer="885" to-port="1" />
+ <edge from-layer="885" from-port="2" to-layer="886" to-port="0" />
+ <edge from-layer="886" from-port="2" to-layer="887" to-port="1" />
+ <edge from-layer="887" from-port="2" to-layer="888" to-port="1" />
+ <edge from-layer="888" from-port="2" to-layer="1711" to-port="0" />
+ <edge from-layer="888" from-port="2" to-layer="889" to-port="0" />
+ <edge from-layer="889" from-port="2" to-layer="890" to-port="0" />
+ <edge from-layer="890" from-port="2" to-layer="892" to-port="0" />
+ <edge from-layer="891" from-port="0" to-layer="892" to-port="1" />
+ <edge from-layer="892" from-port="2" to-layer="916" to-port="1" />
+ <edge from-layer="893" from-port="1" to-layer="895" to-port="0" />
+ <edge from-layer="894" from-port="0" to-layer="895" to-port="2" />
+ <edge from-layer="895" from-port="3" to-layer="911" to-port="0" />
+ <edge from-layer="896" from-port="0" to-layer="897" to-port="0" />
+ <edge from-layer="897" from-port="1" to-layer="900" to-port="0" />
+ <edge from-layer="898" from-port="0" to-layer="899" to-port="0" />
+ <edge from-layer="899" from-port="1" to-layer="900" to-port="1" />
+ <edge from-layer="900" from-port="2" to-layer="902" to-port="0" />
+ <edge from-layer="901" from-port="0" to-layer="902" to-port="1" />
+ <edge from-layer="902" from-port="2" to-layer="904" to-port="0" />
+ <edge from-layer="903" from-port="0" to-layer="904" to-port="1" />
+ <edge from-layer="904" from-port="2" to-layer="905" to-port="0" />
+ <edge from-layer="905" from-port="1" to-layer="906" to-port="1" />
+ <edge from-layer="906" from-port="2" to-layer="908" to-port="0" />
+ <edge from-layer="907" from-port="0" to-layer="908" to-port="1" />
+ <edge from-layer="908" from-port="2" to-layer="910" to-port="0" />
+ <edge from-layer="909" from-port="0" to-layer="910" to-port="1" />
+ <edge from-layer="910" from-port="2" to-layer="911" to-port="1" />
+ <edge from-layer="911" from-port="2" to-layer="912" to-port="0" />
+ <edge from-layer="911" from-port="2" to-layer="1710" to-port="0" />
+ <edge from-layer="912" from-port="2" to-layer="913" to-port="0" />
+ <edge from-layer="913" from-port="2" to-layer="915" to-port="0" />
+ <edge from-layer="914" from-port="0" to-layer="915" to-port="1" />
+ <edge from-layer="915" from-port="2" to-layer="916" to-port="2" />
+ <edge from-layer="916" from-port="4" to-layer="918" to-port="0" />
+ <edge from-layer="917" from-port="0" to-layer="918" to-port="1" />
+ <edge from-layer="918" from-port="2" to-layer="919" to-port="0" />
+ <edge from-layer="919" from-port="2" to-layer="930" to-port="0" />
+ <edge from-layer="920" from-port="0" to-layer="921" to-port="0" />
+ <edge from-layer="921" from-port="1" to-layer="924" to-port="0" />
+ <edge from-layer="922" from-port="0" to-layer="923" to-port="0" />
+ <edge from-layer="923" from-port="1" to-layer="924" to-port="1" />
+ <edge from-layer="924" from-port="2" to-layer="926" to-port="0" />
+ <edge from-layer="925" from-port="0" to-layer="926" to-port="1" />
+ <edge from-layer="926" from-port="2" to-layer="928" to-port="0" />
+ <edge from-layer="927" from-port="0" to-layer="928" to-port="1" />
+ <edge from-layer="928" from-port="2" to-layer="929" to-port="0" />
+ <edge from-layer="929" from-port="1" to-layer="930" to-port="1" />
+ <edge from-layer="930" from-port="2" to-layer="931" to-port="1" />
+ <edge from-layer="931" from-port="2" to-layer="935" to-port="0" />
+ <edge from-layer="931" from-port="2" to-layer="942" to-port="0" />
+ <edge from-layer="931" from-port="2" to-layer="979" to-port="0" />
+ <edge from-layer="932" from-port="0" to-layer="943" to-port="0" />
+ <edge from-layer="933" from-port="0" to-layer="941" to-port="0" />
+ <edge from-layer="934" from-port="0" to-layer="935" to-port="1" />
+ <edge from-layer="935" from-port="2" to-layer="937" to-port="0" />
+ <edge from-layer="936" from-port="0" to-layer="937" to-port="1" />
+ <edge from-layer="937" from-port="2" to-layer="939" to-port="0" />
+ <edge from-layer="938" from-port="0" to-layer="939" to-port="1" />
+ <edge from-layer="939" from-port="2" to-layer="940" to-port="0" />
+ <edge from-layer="940" from-port="1" to-layer="941" to-port="1" />
+ <edge from-layer="941" from-port="2" to-layer="942" to-port="1" />
+ <edge from-layer="942" from-port="2" to-layer="943" to-port="1" />
+ <edge from-layer="943" from-port="2" to-layer="954" to-port="0" />
+ <edge from-layer="943" from-port="2" to-layer="966" to-port="0" />
+ <edge from-layer="944" from-port="0" to-layer="945" to-port="0" />
+ <edge from-layer="945" from-port="1" to-layer="948" to-port="0" />
+ <edge from-layer="946" from-port="0" to-layer="947" to-port="0" />
+ <edge from-layer="947" from-port="1" to-layer="948" to-port="1" />
+ <edge from-layer="948" from-port="2" to-layer="950" to-port="0" />
+ <edge from-layer="949" from-port="0" to-layer="950" to-port="1" />
+ <edge from-layer="950" from-port="2" to-layer="952" to-port="0" />
+ <edge from-layer="951" from-port="0" to-layer="952" to-port="1" />
+ <edge from-layer="952" from-port="2" to-layer="953" to-port="0" />
+ <edge from-layer="953" from-port="1" to-layer="954" to-port="1" />
+ <edge from-layer="954" from-port="2" to-layer="955" to-port="0" />
+ <edge from-layer="955" from-port="1" to-layer="967" to-port="0" />
+ <edge from-layer="956" from-port="0" to-layer="957" to-port="0" />
+ <edge from-layer="957" from-port="1" to-layer="960" to-port="0" />
+ <edge from-layer="958" from-port="0" to-layer="959" to-port="0" />
+ <edge from-layer="959" from-port="1" to-layer="960" to-port="1" />
+ <edge from-layer="960" from-port="2" to-layer="962" to-port="0" />
+ <edge from-layer="961" from-port="0" to-layer="962" to-port="1" />
+ <edge from-layer="962" from-port="2" to-layer="964" to-port="0" />
+ <edge from-layer="963" from-port="0" to-layer="964" to-port="1" />
+ <edge from-layer="964" from-port="2" to-layer="965" to-port="0" />
+ <edge from-layer="965" from-port="1" to-layer="966" to-port="1" />
+ <edge from-layer="966" from-port="2" to-layer="967" to-port="1" />
+ <edge from-layer="967" from-port="2" to-layer="978" to-port="0" />
+ <edge from-layer="968" from-port="0" to-layer="969" to-port="0" />
+ <edge from-layer="969" from-port="1" to-layer="972" to-port="0" />
+ <edge from-layer="970" from-port="0" to-layer="971" to-port="0" />
+ <edge from-layer="971" from-port="1" to-layer="972" to-port="1" />
+ <edge from-layer="972" from-port="2" to-layer="974" to-port="0" />
+ <edge from-layer="973" from-port="0" to-layer="974" to-port="1" />
+ <edge from-layer="974" from-port="2" to-layer="976" to-port="0" />
+ <edge from-layer="975" from-port="0" to-layer="976" to-port="1" />
+ <edge from-layer="976" from-port="2" to-layer="977" to-port="0" />
+ <edge from-layer="977" from-port="1" to-layer="978" to-port="1" />
+ <edge from-layer="978" from-port="2" to-layer="979" to-port="1" />
+ <edge from-layer="979" from-port="2" to-layer="1096" to-port="0" />
+ <edge from-layer="979" from-port="2" to-layer="990" to-port="0" />
+ <edge from-layer="979" from-port="2" to-layer="983" to-port="0" />
+ <edge from-layer="980" from-port="0" to-layer="991" to-port="0" />
+ <edge from-layer="981" from-port="0" to-layer="989" to-port="0" />
+ <edge from-layer="982" from-port="0" to-layer="983" to-port="1" />
+ <edge from-layer="983" from-port="2" to-layer="985" to-port="0" />
+ <edge from-layer="984" from-port="0" to-layer="985" to-port="1" />
+ <edge from-layer="985" from-port="2" to-layer="987" to-port="0" />
+ <edge from-layer="986" from-port="0" to-layer="987" to-port="1" />
+ <edge from-layer="987" from-port="2" to-layer="988" to-port="0" />
+ <edge from-layer="988" from-port="1" to-layer="989" to-port="1" />
+ <edge from-layer="989" from-port="2" to-layer="990" to-port="1" />
+ <edge from-layer="990" from-port="2" to-layer="991" to-port="1" />
+ <edge from-layer="991" from-port="2" to-layer="1034" to-port="0" />
+ <edge from-layer="991" from-port="2" to-layer="1002" to-port="0" />
+ <edge from-layer="991" from-port="2" to-layer="1071" to-port="0" />
+ <edge from-layer="992" from-port="0" to-layer="993" to-port="0" />
+ <edge from-layer="993" from-port="1" to-layer="996" to-port="0" />
+ <edge from-layer="994" from-port="0" to-layer="995" to-port="0" />
+ <edge from-layer="995" from-port="1" to-layer="996" to-port="1" />
+ <edge from-layer="996" from-port="2" to-layer="998" to-port="0" />
+ <edge from-layer="997" from-port="0" to-layer="998" to-port="1" />
+ <edge from-layer="998" from-port="2" to-layer="1000" to-port="0" />
+ <edge from-layer="999" from-port="0" to-layer="1000" to-port="1" />
+ <edge from-layer="1000" from-port="2" to-layer="1001" to-port="0" />
+ <edge from-layer="1001" from-port="1" to-layer="1002" to-port="1" />
+ <edge from-layer="1002" from-port="2" to-layer="1004" to-port="0" />
+ <edge from-layer="1003" from-port="0" to-layer="1004" to-port="1" />
+ <edge from-layer="1004" from-port="2" to-layer="1006" to-port="0" />
+ <edge from-layer="1005" from-port="0" to-layer="1006" to-port="1" />
+ <edge from-layer="1006" from-port="2" to-layer="1007" to-port="0" />
+ <edge from-layer="1006" from-port="2" to-layer="1011" to-port="0" />
+ <edge from-layer="1006" from-port="2" to-layer="1017" to-port="0" />
+ <edge from-layer="1007" from-port="2" to-layer="1020" to-port="0" />
+ <edge from-layer="1008" from-port="0" to-layer="1011" to-port="1" />
+ <edge from-layer="1009" from-port="0" to-layer="1011" to-port="2" />
+ <edge from-layer="1010" from-port="0" to-layer="1011" to-port="3" />
+ <edge from-layer="1011" from-port="5" to-layer="1013" to-port="0" />
+ <edge from-layer="1012" from-port="0" to-layer="1013" to-port="1" />
+ <edge from-layer="1013" from-port="2" to-layer="1018" to-port="0" />
+ <edge from-layer="1014" from-port="0" to-layer="1017" to-port="1" />
+ <edge from-layer="1015" from-port="0" to-layer="1017" to-port="2" />
+ <edge from-layer="1016" from-port="0" to-layer="1017" to-port="3" />
+ <edge from-layer="1017" from-port="5" to-layer="1018" to-port="1" />
+ <edge from-layer="1018" from-port="2" to-layer="1019" to-port="0" />
+ <edge from-layer="1019" from-port="2" to-layer="1020" to-port="1" />
+ <edge from-layer="1020" from-port="2" to-layer="1081" to-port="0" />
+ <edge from-layer="1021" from-port="1" to-layer="1023" to-port="0" />
+ <edge from-layer="1022" from-port="0" to-layer="1023" to-port="2" />
+ <edge from-layer="1023" from-port="3" to-layer="1053" to-port="0" />
+ <edge from-layer="1024" from-port="0" to-layer="1025" to-port="0" />
+ <edge from-layer="1025" from-port="1" to-layer="1028" to-port="0" />
+ <edge from-layer="1026" from-port="0" to-layer="1027" to-port="0" />
+ <edge from-layer="1027" from-port="1" to-layer="1028" to-port="1" />
+ <edge from-layer="1028" from-port="2" to-layer="1030" to-port="0" />
+ <edge from-layer="1029" from-port="0" to-layer="1030" to-port="1" />
+ <edge from-layer="1030" from-port="2" to-layer="1032" to-port="0" />
+ <edge from-layer="1031" from-port="0" to-layer="1032" to-port="1" />
+ <edge from-layer="1032" from-port="2" to-layer="1033" to-port="0" />
+ <edge from-layer="1033" from-port="1" to-layer="1034" to-port="1" />
+ <edge from-layer="1034" from-port="2" to-layer="1036" to-port="0" />
+ <edge from-layer="1035" from-port="0" to-layer="1036" to-port="1" />
+ <edge from-layer="1036" from-port="2" to-layer="1038" to-port="0" />
+ <edge from-layer="1037" from-port="0" to-layer="1038" to-port="1" />
+ <edge from-layer="1038" from-port="2" to-layer="1039" to-port="0" />
+ <edge from-layer="1038" from-port="2" to-layer="1043" to-port="0" />
+ <edge from-layer="1038" from-port="2" to-layer="1049" to-port="0" />
+ <edge from-layer="1039" from-port="2" to-layer="1052" to-port="0" />
+ <edge from-layer="1040" from-port="0" to-layer="1043" to-port="1" />
+ <edge from-layer="1041" from-port="0" to-layer="1043" to-port="2" />
+ <edge from-layer="1042" from-port="0" to-layer="1043" to-port="3" />
+ <edge from-layer="1043" from-port="5" to-layer="1045" to-port="0" />
+ <edge from-layer="1044" from-port="0" to-layer="1045" to-port="1" />
+ <edge from-layer="1045" from-port="2" to-layer="1050" to-port="0" />
+ <edge from-layer="1046" from-port="0" to-layer="1049" to-port="1" />
+ <edge from-layer="1047" from-port="0" to-layer="1049" to-port="2" />
+ <edge from-layer="1048" from-port="0" to-layer="1049" to-port="3" />
+ <edge from-layer="1049" from-port="5" to-layer="1050" to-port="1" />
+ <edge from-layer="1050" from-port="2" to-layer="1051" to-port="0" />
+ <edge from-layer="1051" from-port="2" to-layer="1052" to-port="1" />
+ <edge from-layer="1052" from-port="2" to-layer="1053" to-port="1" />
+ <edge from-layer="1053" from-port="2" to-layer="1709" to-port="0" />
+ <edge from-layer="1053" from-port="2" to-layer="1054" to-port="0" />
+ <edge from-layer="1054" from-port="2" to-layer="1055" to-port="0" />
+ <edge from-layer="1055" from-port="2" to-layer="1057" to-port="0" />
+ <edge from-layer="1056" from-port="0" to-layer="1057" to-port="1" />
+ <edge from-layer="1057" from-port="2" to-layer="1081" to-port="1" />
+ <edge from-layer="1058" from-port="1" to-layer="1060" to-port="0" />
+ <edge from-layer="1059" from-port="0" to-layer="1060" to-port="2" />
+ <edge from-layer="1060" from-port="3" to-layer="1076" to-port="0" />
+ <edge from-layer="1061" from-port="0" to-layer="1062" to-port="0" />
+ <edge from-layer="1062" from-port="1" to-layer="1065" to-port="0" />
+ <edge from-layer="1063" from-port="0" to-layer="1064" to-port="0" />
+ <edge from-layer="1064" from-port="1" to-layer="1065" to-port="1" />
+ <edge from-layer="1065" from-port="2" to-layer="1067" to-port="0" />
+ <edge from-layer="1066" from-port="0" to-layer="1067" to-port="1" />
+ <edge from-layer="1067" from-port="2" to-layer="1069" to-port="0" />
+ <edge from-layer="1068" from-port="0" to-layer="1069" to-port="1" />
+ <edge from-layer="1069" from-port="2" to-layer="1070" to-port="0" />
+ <edge from-layer="1070" from-port="1" to-layer="1071" to-port="1" />
+ <edge from-layer="1071" from-port="2" to-layer="1073" to-port="0" />
+ <edge from-layer="1072" from-port="0" to-layer="1073" to-port="1" />
+ <edge from-layer="1073" from-port="2" to-layer="1075" to-port="0" />
+ <edge from-layer="1074" from-port="0" to-layer="1075" to-port="1" />
+ <edge from-layer="1075" from-port="2" to-layer="1076" to-port="1" />
+ <edge from-layer="1076" from-port="2" to-layer="1708" to-port="0" />
+ <edge from-layer="1076" from-port="2" to-layer="1077" to-port="0" />
+ <edge from-layer="1077" from-port="2" to-layer="1078" to-port="0" />
+ <edge from-layer="1078" from-port="2" to-layer="1080" to-port="0" />
+ <edge from-layer="1079" from-port="0" to-layer="1080" to-port="1" />
+ <edge from-layer="1080" from-port="2" to-layer="1081" to-port="2" />
+ <edge from-layer="1081" from-port="4" to-layer="1083" to-port="0" />
+ <edge from-layer="1082" from-port="0" to-layer="1083" to-port="1" />
+ <edge from-layer="1083" from-port="2" to-layer="1084" to-port="0" />
+ <edge from-layer="1084" from-port="2" to-layer="1095" to-port="0" />
+ <edge from-layer="1085" from-port="0" to-layer="1086" to-port="0" />
+ <edge from-layer="1086" from-port="1" to-layer="1089" to-port="0" />
+ <edge from-layer="1087" from-port="0" to-layer="1088" to-port="0" />
+ <edge from-layer="1088" from-port="1" to-layer="1089" to-port="1" />
+ <edge from-layer="1089" from-port="2" to-layer="1091" to-port="0" />
+ <edge from-layer="1090" from-port="0" to-layer="1091" to-port="1" />
+ <edge from-layer="1091" from-port="2" to-layer="1093" to-port="0" />
+ <edge from-layer="1092" from-port="0" to-layer="1093" to-port="1" />
+ <edge from-layer="1093" from-port="2" to-layer="1094" to-port="0" />
+ <edge from-layer="1094" from-port="1" to-layer="1095" to-port="1" />
+ <edge from-layer="1095" from-port="2" to-layer="1096" to-port="1" />
+ <edge from-layer="1096" from-port="2" to-layer="1100" to-port="0" />
+ <edge from-layer="1096" from-port="2" to-layer="1144" to-port="0" />
+ <edge from-layer="1096" from-port="2" to-layer="1107" to-port="0" />
+ <edge from-layer="1097" from-port="0" to-layer="1108" to-port="0" />
+ <edge from-layer="1098" from-port="0" to-layer="1106" to-port="0" />
+ <edge from-layer="1099" from-port="0" to-layer="1100" to-port="1" />
+ <edge from-layer="1100" from-port="2" to-layer="1102" to-port="0" />
+ <edge from-layer="1101" from-port="0" to-layer="1102" to-port="1" />
+ <edge from-layer="1102" from-port="2" to-layer="1104" to-port="0" />
+ <edge from-layer="1103" from-port="0" to-layer="1104" to-port="1" />
+ <edge from-layer="1104" from-port="2" to-layer="1105" to-port="0" />
+ <edge from-layer="1105" from-port="1" to-layer="1106" to-port="1" />
+ <edge from-layer="1106" from-port="2" to-layer="1107" to-port="1" />
+ <edge from-layer="1107" from-port="2" to-layer="1108" to-port="1" />
+ <edge from-layer="1108" from-port="2" to-layer="1119" to-port="0" />
+ <edge from-layer="1108" from-port="2" to-layer="1131" to-port="0" />
+ <edge from-layer="1109" from-port="0" to-layer="1110" to-port="0" />
+ <edge from-layer="1110" from-port="1" to-layer="1113" to-port="0" />
+ <edge from-layer="1111" from-port="0" to-layer="1112" to-port="0" />
+ <edge from-layer="1112" from-port="1" to-layer="1113" to-port="1" />
+ <edge from-layer="1113" from-port="2" to-layer="1115" to-port="0" />
+ <edge from-layer="1114" from-port="0" to-layer="1115" to-port="1" />
+ <edge from-layer="1115" from-port="2" to-layer="1117" to-port="0" />
+ <edge from-layer="1116" from-port="0" to-layer="1117" to-port="1" />
+ <edge from-layer="1117" from-port="2" to-layer="1118" to-port="0" />
+ <edge from-layer="1118" from-port="1" to-layer="1119" to-port="1" />
+ <edge from-layer="1119" from-port="2" to-layer="1120" to-port="0" />
+ <edge from-layer="1120" from-port="1" to-layer="1132" to-port="0" />
+ <edge from-layer="1121" from-port="0" to-layer="1122" to-port="0" />
+ <edge from-layer="1122" from-port="1" to-layer="1125" to-port="0" />
+ <edge from-layer="1123" from-port="0" to-layer="1124" to-port="0" />
+ <edge from-layer="1124" from-port="1" to-layer="1125" to-port="1" />
+ <edge from-layer="1125" from-port="2" to-layer="1127" to-port="0" />
+ <edge from-layer="1126" from-port="0" to-layer="1127" to-port="1" />
+ <edge from-layer="1127" from-port="2" to-layer="1129" to-port="0" />
+ <edge from-layer="1128" from-port="0" to-layer="1129" to-port="1" />
+ <edge from-layer="1129" from-port="2" to-layer="1130" to-port="0" />
+ <edge from-layer="1130" from-port="1" to-layer="1131" to-port="1" />
+ <edge from-layer="1131" from-port="2" to-layer="1132" to-port="1" />
+ <edge from-layer="1132" from-port="2" to-layer="1143" to-port="0" />
+ <edge from-layer="1133" from-port="0" to-layer="1134" to-port="0" />
+ <edge from-layer="1134" from-port="1" to-layer="1137" to-port="0" />
+ <edge from-layer="1135" from-port="0" to-layer="1136" to-port="0" />
+ <edge from-layer="1136" from-port="1" to-layer="1137" to-port="1" />
+ <edge from-layer="1137" from-port="2" to-layer="1139" to-port="0" />
+ <edge from-layer="1138" from-port="0" to-layer="1139" to-port="1" />
+ <edge from-layer="1139" from-port="2" to-layer="1141" to-port="0" />
+ <edge from-layer="1140" from-port="0" to-layer="1141" to-port="1" />
+ <edge from-layer="1141" from-port="2" to-layer="1142" to-port="0" />
+ <edge from-layer="1142" from-port="1" to-layer="1143" to-port="1" />
+ <edge from-layer="1143" from-port="2" to-layer="1144" to-port="1" />
+ <edge from-layer="1144" from-port="2" to-layer="1261" to-port="0" />
+ <edge from-layer="1144" from-port="2" to-layer="1148" to-port="0" />
+ <edge from-layer="1144" from-port="2" to-layer="1155" to-port="0" />
+ <edge from-layer="1145" from-port="0" to-layer="1156" to-port="0" />
+ <edge from-layer="1146" from-port="0" to-layer="1154" to-port="0" />
+ <edge from-layer="1147" from-port="0" to-layer="1148" to-port="1" />
+ <edge from-layer="1148" from-port="2" to-layer="1150" to-port="0" />
+ <edge from-layer="1149" from-port="0" to-layer="1150" to-port="1" />
+ <edge from-layer="1150" from-port="2" to-layer="1152" to-port="0" />
+ <edge from-layer="1151" from-port="0" to-layer="1152" to-port="1" />
+ <edge from-layer="1152" from-port="2" to-layer="1153" to-port="0" />
+ <edge from-layer="1153" from-port="1" to-layer="1154" to-port="1" />
+ <edge from-layer="1154" from-port="2" to-layer="1155" to-port="1" />
+ <edge from-layer="1155" from-port="2" to-layer="1156" to-port="1" />
+ <edge from-layer="1156" from-port="2" to-layer="1199" to-port="0" />
+ <edge from-layer="1156" from-port="2" to-layer="1236" to-port="0" />
+ <edge from-layer="1156" from-port="2" to-layer="1167" to-port="0" />
+ <edge from-layer="1157" from-port="0" to-layer="1158" to-port="0" />
+ <edge from-layer="1158" from-port="1" to-layer="1161" to-port="0" />
+ <edge from-layer="1159" from-port="0" to-layer="1160" to-port="0" />
+ <edge from-layer="1160" from-port="1" to-layer="1161" to-port="1" />
+ <edge from-layer="1161" from-port="2" to-layer="1163" to-port="0" />
+ <edge from-layer="1162" from-port="0" to-layer="1163" to-port="1" />
+ <edge from-layer="1163" from-port="2" to-layer="1165" to-port="0" />
+ <edge from-layer="1164" from-port="0" to-layer="1165" to-port="1" />
+ <edge from-layer="1165" from-port="2" to-layer="1166" to-port="0" />
+ <edge from-layer="1166" from-port="1" to-layer="1167" to-port="1" />
+ <edge from-layer="1167" from-port="2" to-layer="1169" to-port="0" />
+ <edge from-layer="1168" from-port="0" to-layer="1169" to-port="1" />
+ <edge from-layer="1169" from-port="2" to-layer="1171" to-port="0" />
+ <edge from-layer="1170" from-port="0" to-layer="1171" to-port="1" />
+ <edge from-layer="1171" from-port="2" to-layer="1172" to-port="0" />
+ <edge from-layer="1171" from-port="2" to-layer="1182" to-port="0" />
+ <edge from-layer="1171" from-port="2" to-layer="1176" to-port="0" />
+ <edge from-layer="1172" from-port="2" to-layer="1185" to-port="0" />
+ <edge from-layer="1173" from-port="0" to-layer="1176" to-port="1" />
+ <edge from-layer="1174" from-port="0" to-layer="1176" to-port="2" />
+ <edge from-layer="1175" from-port="0" to-layer="1176" to-port="3" />
+ <edge from-layer="1176" from-port="5" to-layer="1178" to-port="0" />
+ <edge from-layer="1177" from-port="0" to-layer="1178" to-port="1" />
+ <edge from-layer="1178" from-port="2" to-layer="1183" to-port="0" />
+ <edge from-layer="1179" from-port="0" to-layer="1182" to-port="1" />
+ <edge from-layer="1180" from-port="0" to-layer="1182" to-port="2" />
+ <edge from-layer="1181" from-port="0" to-layer="1182" to-port="3" />
+ <edge from-layer="1182" from-port="5" to-layer="1183" to-port="1" />
+ <edge from-layer="1183" from-port="2" to-layer="1184" to-port="0" />
+ <edge from-layer="1184" from-port="2" to-layer="1185" to-port="1" />
+ <edge from-layer="1185" from-port="2" to-layer="1246" to-port="0" />
+ <edge from-layer="1186" from-port="1" to-layer="1188" to-port="0" />
+ <edge from-layer="1187" from-port="0" to-layer="1188" to-port="2" />
+ <edge from-layer="1188" from-port="3" to-layer="1218" to-port="0" />
+ <edge from-layer="1189" from-port="0" to-layer="1190" to-port="0" />
+ <edge from-layer="1190" from-port="1" to-layer="1193" to-port="0" />
+ <edge from-layer="1191" from-port="0" to-layer="1192" to-port="0" />
+ <edge from-layer="1192" from-port="1" to-layer="1193" to-port="1" />
+ <edge from-layer="1193" from-port="2" to-layer="1195" to-port="0" />
+ <edge from-layer="1194" from-port="0" to-layer="1195" to-port="1" />
+ <edge from-layer="1195" from-port="2" to-layer="1197" to-port="0" />
+ <edge from-layer="1196" from-port="0" to-layer="1197" to-port="1" />
+ <edge from-layer="1197" from-port="2" to-layer="1198" to-port="0" />
+ <edge from-layer="1198" from-port="1" to-layer="1199" to-port="1" />
+ <edge from-layer="1199" from-port="2" to-layer="1201" to-port="0" />
+ <edge from-layer="1200" from-port="0" to-layer="1201" to-port="1" />
+ <edge from-layer="1201" from-port="2" to-layer="1203" to-port="0" />
+ <edge from-layer="1202" from-port="0" to-layer="1203" to-port="1" />
+ <edge from-layer="1203" from-port="2" to-layer="1204" to-port="0" />
+ <edge from-layer="1203" from-port="2" to-layer="1214" to-port="0" />
+ <edge from-layer="1203" from-port="2" to-layer="1208" to-port="0" />
+ <edge from-layer="1204" from-port="2" to-layer="1217" to-port="0" />
+ <edge from-layer="1205" from-port="0" to-layer="1208" to-port="1" />
+ <edge from-layer="1206" from-port="0" to-layer="1208" to-port="2" />
+ <edge from-layer="1207" from-port="0" to-layer="1208" to-port="3" />
+ <edge from-layer="1208" from-port="5" to-layer="1210" to-port="0" />
+ <edge from-layer="1209" from-port="0" to-layer="1210" to-port="1" />
+ <edge from-layer="1210" from-port="2" to-layer="1215" to-port="0" />
+ <edge from-layer="1211" from-port="0" to-layer="1214" to-port="1" />
+ <edge from-layer="1212" from-port="0" to-layer="1214" to-port="2" />
+ <edge from-layer="1213" from-port="0" to-layer="1214" to-port="3" />
+ <edge from-layer="1214" from-port="5" to-layer="1215" to-port="1" />
+ <edge from-layer="1215" from-port="2" to-layer="1216" to-port="0" />
+ <edge from-layer="1216" from-port="2" to-layer="1217" to-port="1" />
+ <edge from-layer="1217" from-port="2" to-layer="1218" to-port="1" />
+ <edge from-layer="1218" from-port="2" to-layer="1707" to-port="0" />
+ <edge from-layer="1218" from-port="2" to-layer="1219" to-port="0" />
+ <edge from-layer="1219" from-port="2" to-layer="1220" to-port="0" />
+ <edge from-layer="1220" from-port="2" to-layer="1222" to-port="0" />
+ <edge from-layer="1221" from-port="0" to-layer="1222" to-port="1" />
+ <edge from-layer="1222" from-port="2" to-layer="1246" to-port="1" />
+ <edge from-layer="1223" from-port="1" to-layer="1225" to-port="0" />
+ <edge from-layer="1224" from-port="0" to-layer="1225" to-port="2" />
+ <edge from-layer="1225" from-port="3" to-layer="1241" to-port="0" />
+ <edge from-layer="1226" from-port="0" to-layer="1227" to-port="0" />
+ <edge from-layer="1227" from-port="1" to-layer="1230" to-port="0" />
+ <edge from-layer="1228" from-port="0" to-layer="1229" to-port="0" />
+ <edge from-layer="1229" from-port="1" to-layer="1230" to-port="1" />
+ <edge from-layer="1230" from-port="2" to-layer="1232" to-port="0" />
+ <edge from-layer="1231" from-port="0" to-layer="1232" to-port="1" />
+ <edge from-layer="1232" from-port="2" to-layer="1234" to-port="0" />
+ <edge from-layer="1233" from-port="0" to-layer="1234" to-port="1" />
+ <edge from-layer="1234" from-port="2" to-layer="1235" to-port="0" />
+ <edge from-layer="1235" from-port="1" to-layer="1236" to-port="1" />
+ <edge from-layer="1236" from-port="2" to-layer="1238" to-port="0" />
+ <edge from-layer="1237" from-port="0" to-layer="1238" to-port="1" />
+ <edge from-layer="1238" from-port="2" to-layer="1240" to-port="0" />
+ <edge from-layer="1239" from-port="0" to-layer="1240" to-port="1" />
+ <edge from-layer="1240" from-port="2" to-layer="1241" to-port="1" />
+ <edge from-layer="1241" from-port="2" to-layer="1242" to-port="0" />
+ <edge from-layer="1241" from-port="2" to-layer="1706" to-port="0" />
+ <edge from-layer="1242" from-port="2" to-layer="1243" to-port="0" />
+ <edge from-layer="1243" from-port="2" to-layer="1245" to-port="0" />
+ <edge from-layer="1244" from-port="0" to-layer="1245" to-port="1" />
+ <edge from-layer="1245" from-port="2" to-layer="1246" to-port="2" />
+ <edge from-layer="1246" from-port="4" to-layer="1248" to-port="0" />
+ <edge from-layer="1247" from-port="0" to-layer="1248" to-port="1" />
+ <edge from-layer="1248" from-port="2" to-layer="1249" to-port="0" />
+ <edge from-layer="1249" from-port="2" to-layer="1260" to-port="0" />
+ <edge from-layer="1250" from-port="0" to-layer="1251" to-port="0" />
+ <edge from-layer="1251" from-port="1" to-layer="1254" to-port="0" />
+ <edge from-layer="1252" from-port="0" to-layer="1253" to-port="0" />
+ <edge from-layer="1253" from-port="1" to-layer="1254" to-port="1" />
+ <edge from-layer="1254" from-port="2" to-layer="1256" to-port="0" />
+ <edge from-layer="1255" from-port="0" to-layer="1256" to-port="1" />
+ <edge from-layer="1256" from-port="2" to-layer="1258" to-port="0" />
+ <edge from-layer="1257" from-port="0" to-layer="1258" to-port="1" />
+ <edge from-layer="1258" from-port="2" to-layer="1259" to-port="0" />
+ <edge from-layer="1259" from-port="1" to-layer="1260" to-port="1" />
+ <edge from-layer="1260" from-port="2" to-layer="1261" to-port="1" />
+ <edge from-layer="1261" from-port="2" to-layer="1265" to-port="0" />
+ <edge from-layer="1261" from-port="2" to-layer="1309" to-port="0" />
+ <edge from-layer="1261" from-port="2" to-layer="1272" to-port="0" />
+ <edge from-layer="1262" from-port="0" to-layer="1273" to-port="0" />
+ <edge from-layer="1263" from-port="0" to-layer="1271" to-port="0" />
+ <edge from-layer="1264" from-port="0" to-layer="1265" to-port="1" />
+ <edge from-layer="1265" from-port="2" to-layer="1267" to-port="0" />
+ <edge from-layer="1266" from-port="0" to-layer="1267" to-port="1" />
+ <edge from-layer="1267" from-port="2" to-layer="1269" to-port="0" />
+ <edge from-layer="1268" from-port="0" to-layer="1269" to-port="1" />
+ <edge from-layer="1269" from-port="2" to-layer="1270" to-port="0" />
+ <edge from-layer="1270" from-port="1" to-layer="1271" to-port="1" />
+ <edge from-layer="1271" from-port="2" to-layer="1272" to-port="1" />
+ <edge from-layer="1272" from-port="2" to-layer="1273" to-port="1" />
+ <edge from-layer="1273" from-port="2" to-layer="1284" to-port="0" />
+ <edge from-layer="1273" from-port="2" to-layer="1296" to-port="0" />
+ <edge from-layer="1274" from-port="0" to-layer="1275" to-port="0" />
+ <edge from-layer="1275" from-port="1" to-layer="1278" to-port="0" />
+ <edge from-layer="1276" from-port="0" to-layer="1277" to-port="0" />
+ <edge from-layer="1277" from-port="1" to-layer="1278" to-port="1" />
+ <edge from-layer="1278" from-port="2" to-layer="1280" to-port="0" />
+ <edge from-layer="1279" from-port="0" to-layer="1280" to-port="1" />
+ <edge from-layer="1280" from-port="2" to-layer="1282" to-port="0" />
+ <edge from-layer="1281" from-port="0" to-layer="1282" to-port="1" />
+ <edge from-layer="1282" from-port="2" to-layer="1283" to-port="0" />
+ <edge from-layer="1283" from-port="1" to-layer="1284" to-port="1" />
+ <edge from-layer="1284" from-port="2" to-layer="1285" to-port="0" />
+ <edge from-layer="1285" from-port="1" to-layer="1297" to-port="0" />
+ <edge from-layer="1286" from-port="0" to-layer="1287" to-port="0" />
+ <edge from-layer="1287" from-port="1" to-layer="1290" to-port="0" />
+ <edge from-layer="1288" from-port="0" to-layer="1289" to-port="0" />
+ <edge from-layer="1289" from-port="1" to-layer="1290" to-port="1" />
+ <edge from-layer="1290" from-port="2" to-layer="1292" to-port="0" />
+ <edge from-layer="1291" from-port="0" to-layer="1292" to-port="1" />
+ <edge from-layer="1292" from-port="2" to-layer="1294" to-port="0" />
+ <edge from-layer="1293" from-port="0" to-layer="1294" to-port="1" />
+ <edge from-layer="1294" from-port="2" to-layer="1295" to-port="0" />
+ <edge from-layer="1295" from-port="1" to-layer="1296" to-port="1" />
+ <edge from-layer="1296" from-port="2" to-layer="1297" to-port="1" />
+ <edge from-layer="1297" from-port="2" to-layer="1308" to-port="0" />
+ <edge from-layer="1298" from-port="0" to-layer="1299" to-port="0" />
+ <edge from-layer="1299" from-port="1" to-layer="1302" to-port="0" />
+ <edge from-layer="1300" from-port="0" to-layer="1301" to-port="0" />
+ <edge from-layer="1301" from-port="1" to-layer="1302" to-port="1" />
+ <edge from-layer="1302" from-port="2" to-layer="1304" to-port="0" />
+ <edge from-layer="1303" from-port="0" to-layer="1304" to-port="1" />
+ <edge from-layer="1304" from-port="2" to-layer="1306" to-port="0" />
+ <edge from-layer="1305" from-port="0" to-layer="1306" to-port="1" />
+ <edge from-layer="1306" from-port="2" to-layer="1307" to-port="0" />
+ <edge from-layer="1307" from-port="1" to-layer="1308" to-port="1" />
+ <edge from-layer="1308" from-port="2" to-layer="1309" to-port="1" />
+ <edge from-layer="1309" from-port="2" to-layer="1320" to-port="0" />
+ <edge from-layer="1309" from-port="2" to-layer="1313" to-port="0" />
+ <edge from-layer="1309" from-port="2" to-layer="1426" to-port="0" />
+ <edge from-layer="1310" from-port="0" to-layer="1321" to-port="0" />
+ <edge from-layer="1311" from-port="0" to-layer="1319" to-port="0" />
+ <edge from-layer="1312" from-port="0" to-layer="1313" to-port="1" />
+ <edge from-layer="1313" from-port="2" to-layer="1315" to-port="0" />
+ <edge from-layer="1314" from-port="0" to-layer="1315" to-port="1" />
+ <edge from-layer="1315" from-port="2" to-layer="1317" to-port="0" />
+ <edge from-layer="1316" from-port="0" to-layer="1317" to-port="1" />
+ <edge from-layer="1317" from-port="2" to-layer="1318" to-port="0" />
+ <edge from-layer="1318" from-port="1" to-layer="1319" to-port="1" />
+ <edge from-layer="1319" from-port="2" to-layer="1320" to-port="1" />
+ <edge from-layer="1320" from-port="2" to-layer="1321" to-port="1" />
+ <edge from-layer="1321" from-port="2" to-layer="1332" to-port="0" />
+ <edge from-layer="1321" from-port="2" to-layer="1364" to-port="0" />
+ <edge from-layer="1321" from-port="2" to-layer="1401" to-port="0" />
+ <edge from-layer="1322" from-port="0" to-layer="1323" to-port="0" />
+ <edge from-layer="1323" from-port="1" to-layer="1326" to-port="0" />
+ <edge from-layer="1324" from-port="0" to-layer="1325" to-port="0" />
+ <edge from-layer="1325" from-port="1" to-layer="1326" to-port="1" />
+ <edge from-layer="1326" from-port="2" to-layer="1328" to-port="0" />
+ <edge from-layer="1327" from-port="0" to-layer="1328" to-port="1" />
+ <edge from-layer="1328" from-port="2" to-layer="1330" to-port="0" />
+ <edge from-layer="1329" from-port="0" to-layer="1330" to-port="1" />
+ <edge from-layer="1330" from-port="2" to-layer="1331" to-port="0" />
+ <edge from-layer="1331" from-port="1" to-layer="1332" to-port="1" />
+ <edge from-layer="1332" from-port="2" to-layer="1334" to-port="0" />
+ <edge from-layer="1333" from-port="0" to-layer="1334" to-port="1" />
+ <edge from-layer="1334" from-port="2" to-layer="1336" to-port="0" />
+ <edge from-layer="1335" from-port="0" to-layer="1336" to-port="1" />
+ <edge from-layer="1336" from-port="2" to-layer="1337" to-port="0" />
+ <edge from-layer="1336" from-port="2" to-layer="1341" to-port="0" />
+ <edge from-layer="1336" from-port="2" to-layer="1347" to-port="0" />
+ <edge from-layer="1337" from-port="2" to-layer="1350" to-port="0" />
+ <edge from-layer="1338" from-port="0" to-layer="1341" to-port="1" />
+ <edge from-layer="1339" from-port="0" to-layer="1341" to-port="2" />
+ <edge from-layer="1340" from-port="0" to-layer="1341" to-port="3" />
+ <edge from-layer="1341" from-port="5" to-layer="1343" to-port="0" />
+ <edge from-layer="1342" from-port="0" to-layer="1343" to-port="1" />
+ <edge from-layer="1343" from-port="2" to-layer="1348" to-port="0" />
+ <edge from-layer="1344" from-port="0" to-layer="1347" to-port="1" />
+ <edge from-layer="1345" from-port="0" to-layer="1347" to-port="2" />
+ <edge from-layer="1346" from-port="0" to-layer="1347" to-port="3" />
+ <edge from-layer="1347" from-port="5" to-layer="1348" to-port="1" />
+ <edge from-layer="1348" from-port="2" to-layer="1349" to-port="0" />
+ <edge from-layer="1349" from-port="2" to-layer="1350" to-port="1" />
+ <edge from-layer="1350" from-port="2" to-layer="1411" to-port="0" />
+ <edge from-layer="1351" from-port="1" to-layer="1353" to-port="0" />
+ <edge from-layer="1352" from-port="0" to-layer="1353" to-port="2" />
+ <edge from-layer="1353" from-port="3" to-layer="1383" to-port="0" />
+ <edge from-layer="1354" from-port="0" to-layer="1355" to-port="0" />
+ <edge from-layer="1355" from-port="1" to-layer="1358" to-port="0" />
+ <edge from-layer="1356" from-port="0" to-layer="1357" to-port="0" />
+ <edge from-layer="1357" from-port="1" to-layer="1358" to-port="1" />
+ <edge from-layer="1358" from-port="2" to-layer="1360" to-port="0" />
+ <edge from-layer="1359" from-port="0" to-layer="1360" to-port="1" />
+ <edge from-layer="1360" from-port="2" to-layer="1362" to-port="0" />
+ <edge from-layer="1361" from-port="0" to-layer="1362" to-port="1" />
+ <edge from-layer="1362" from-port="2" to-layer="1363" to-port="0" />
+ <edge from-layer="1363" from-port="1" to-layer="1364" to-port="1" />
+ <edge from-layer="1364" from-port="2" to-layer="1366" to-port="0" />
+ <edge from-layer="1365" from-port="0" to-layer="1366" to-port="1" />
+ <edge from-layer="1366" from-port="2" to-layer="1368" to-port="0" />
+ <edge from-layer="1367" from-port="0" to-layer="1368" to-port="1" />
+ <edge from-layer="1368" from-port="2" to-layer="1373" to-port="0" />
+ <edge from-layer="1368" from-port="2" to-layer="1379" to-port="0" />
+ <edge from-layer="1368" from-port="2" to-layer="1369" to-port="0" />
+ <edge from-layer="1369" from-port="2" to-layer="1382" to-port="0" />
+ <edge from-layer="1370" from-port="0" to-layer="1373" to-port="1" />
+ <edge from-layer="1371" from-port="0" to-layer="1373" to-port="2" />
+ <edge from-layer="1372" from-port="0" to-layer="1373" to-port="3" />
+ <edge from-layer="1373" from-port="5" to-layer="1375" to-port="0" />
+ <edge from-layer="1374" from-port="0" to-layer="1375" to-port="1" />
+ <edge from-layer="1375" from-port="2" to-layer="1380" to-port="0" />
+ <edge from-layer="1376" from-port="0" to-layer="1379" to-port="1" />
+ <edge from-layer="1377" from-port="0" to-layer="1379" to-port="2" />
+ <edge from-layer="1378" from-port="0" to-layer="1379" to-port="3" />
+ <edge from-layer="1379" from-port="5" to-layer="1380" to-port="1" />
+ <edge from-layer="1380" from-port="2" to-layer="1381" to-port="0" />
+ <edge from-layer="1381" from-port="2" to-layer="1382" to-port="1" />
+ <edge from-layer="1382" from-port="2" to-layer="1383" to-port="1" />
+ <edge from-layer="1383" from-port="2" to-layer="1384" to-port="0" />
+ <edge from-layer="1383" from-port="2" to-layer="1705" to-port="0" />
+ <edge from-layer="1384" from-port="2" to-layer="1385" to-port="0" />
+ <edge from-layer="1385" from-port="2" to-layer="1387" to-port="0" />
+ <edge from-layer="1386" from-port="0" to-layer="1387" to-port="1" />
+ <edge from-layer="1387" from-port="2" to-layer="1411" to-port="1" />
+ <edge from-layer="1388" from-port="1" to-layer="1390" to-port="0" />
+ <edge from-layer="1389" from-port="0" to-layer="1390" to-port="2" />
+ <edge from-layer="1390" from-port="3" to-layer="1406" to-port="0" />
+ <edge from-layer="1391" from-port="0" to-layer="1392" to-port="0" />
+ <edge from-layer="1392" from-port="1" to-layer="1395" to-port="0" />
+ <edge from-layer="1393" from-port="0" to-layer="1394" to-port="0" />
+ <edge from-layer="1394" from-port="1" to-layer="1395" to-port="1" />
+ <edge from-layer="1395" from-port="2" to-layer="1397" to-port="0" />
+ <edge from-layer="1396" from-port="0" to-layer="1397" to-port="1" />
+ <edge from-layer="1397" from-port="2" to-layer="1399" to-port="0" />
+ <edge from-layer="1398" from-port="0" to-layer="1399" to-port="1" />
+ <edge from-layer="1399" from-port="2" to-layer="1400" to-port="0" />
+ <edge from-layer="1400" from-port="1" to-layer="1401" to-port="1" />
+ <edge from-layer="1401" from-port="2" to-layer="1403" to-port="0" />
+ <edge from-layer="1402" from-port="0" to-layer="1403" to-port="1" />
+ <edge from-layer="1403" from-port="2" to-layer="1405" to-port="0" />
+ <edge from-layer="1404" from-port="0" to-layer="1405" to-port="1" />
+ <edge from-layer="1405" from-port="2" to-layer="1406" to-port="1" />
+ <edge from-layer="1406" from-port="2" to-layer="1407" to-port="0" />
+ <edge from-layer="1406" from-port="2" to-layer="1704" to-port="0" />
+ <edge from-layer="1407" from-port="2" to-layer="1408" to-port="0" />
+ <edge from-layer="1408" from-port="2" to-layer="1410" to-port="0" />
+ <edge from-layer="1409" from-port="0" to-layer="1410" to-port="1" />
+ <edge from-layer="1410" from-port="2" to-layer="1411" to-port="2" />
+ <edge from-layer="1411" from-port="4" to-layer="1413" to-port="0" />
+ <edge from-layer="1412" from-port="0" to-layer="1413" to-port="1" />
+ <edge from-layer="1413" from-port="2" to-layer="1414" to-port="0" />
+ <edge from-layer="1414" from-port="2" to-layer="1425" to-port="0" />
+ <edge from-layer="1415" from-port="0" to-layer="1416" to-port="0" />
+ <edge from-layer="1416" from-port="1" to-layer="1419" to-port="0" />
+ <edge from-layer="1417" from-port="0" to-layer="1418" to-port="0" />
+ <edge from-layer="1418" from-port="1" to-layer="1419" to-port="1" />
+ <edge from-layer="1419" from-port="2" to-layer="1421" to-port="0" />
+ <edge from-layer="1420" from-port="0" to-layer="1421" to-port="1" />
+ <edge from-layer="1421" from-port="2" to-layer="1423" to-port="0" />
+ <edge from-layer="1422" from-port="0" to-layer="1423" to-port="1" />
+ <edge from-layer="1423" from-port="2" to-layer="1424" to-port="0" />
+ <edge from-layer="1424" from-port="1" to-layer="1425" to-port="1" />
+ <edge from-layer="1425" from-port="2" to-layer="1426" to-port="1" />
+ <edge from-layer="1426" from-port="2" to-layer="1437" to-port="0" />
+ <edge from-layer="1426" from-port="2" to-layer="1474" to-port="0" />
+ <edge from-layer="1426" from-port="2" to-layer="1430" to-port="0" />
+ <edge from-layer="1427" from-port="0" to-layer="1438" to-port="0" />
+ <edge from-layer="1428" from-port="0" to-layer="1436" to-port="0" />
+ <edge from-layer="1429" from-port="0" to-layer="1430" to-port="1" />
+ <edge from-layer="1430" from-port="2" to-layer="1432" to-port="0" />
+ <edge from-layer="1431" from-port="0" to-layer="1432" to-port="1" />
+ <edge from-layer="1432" from-port="2" to-layer="1434" to-port="0" />
+ <edge from-layer="1433" from-port="0" to-layer="1434" to-port="1" />
+ <edge from-layer="1434" from-port="2" to-layer="1435" to-port="0" />
+ <edge from-layer="1435" from-port="1" to-layer="1436" to-port="1" />
+ <edge from-layer="1436" from-port="2" to-layer="1437" to-port="1" />
+ <edge from-layer="1437" from-port="2" to-layer="1438" to-port="1" />
+ <edge from-layer="1438" from-port="2" to-layer="1461" to-port="0" />
+ <edge from-layer="1438" from-port="2" to-layer="1449" to-port="0" />
+ <edge from-layer="1439" from-port="0" to-layer="1440" to-port="0" />
+ <edge from-layer="1440" from-port="1" to-layer="1443" to-port="0" />
+ <edge from-layer="1441" from-port="0" to-layer="1442" to-port="0" />
+ <edge from-layer="1442" from-port="1" to-layer="1443" to-port="1" />
+ <edge from-layer="1443" from-port="2" to-layer="1445" to-port="0" />
+ <edge from-layer="1444" from-port="0" to-layer="1445" to-port="1" />
+ <edge from-layer="1445" from-port="2" to-layer="1447" to-port="0" />
+ <edge from-layer="1446" from-port="0" to-layer="1447" to-port="1" />
+ <edge from-layer="1447" from-port="2" to-layer="1448" to-port="0" />
+ <edge from-layer="1448" from-port="1" to-layer="1449" to-port="1" />
+ <edge from-layer="1449" from-port="2" to-layer="1450" to-port="0" />
+ <edge from-layer="1450" from-port="1" to-layer="1462" to-port="0" />
+ <edge from-layer="1451" from-port="0" to-layer="1452" to-port="0" />
+ <edge from-layer="1452" from-port="1" to-layer="1455" to-port="0" />
+ <edge from-layer="1453" from-port="0" to-layer="1454" to-port="0" />
+ <edge from-layer="1454" from-port="1" to-layer="1455" to-port="1" />
+ <edge from-layer="1455" from-port="2" to-layer="1457" to-port="0" />
+ <edge from-layer="1456" from-port="0" to-layer="1457" to-port="1" />
+ <edge from-layer="1457" from-port="2" to-layer="1459" to-port="0" />
+ <edge from-layer="1458" from-port="0" to-layer="1459" to-port="1" />
+ <edge from-layer="1459" from-port="2" to-layer="1460" to-port="0" />
+ <edge from-layer="1460" from-port="1" to-layer="1461" to-port="1" />
+ <edge from-layer="1461" from-port="2" to-layer="1462" to-port="1" />
+ <edge from-layer="1462" from-port="2" to-layer="1473" to-port="0" />
+ <edge from-layer="1463" from-port="0" to-layer="1464" to-port="0" />
+ <edge from-layer="1464" from-port="1" to-layer="1467" to-port="0" />
+ <edge from-layer="1465" from-port="0" to-layer="1466" to-port="0" />
+ <edge from-layer="1466" from-port="1" to-layer="1467" to-port="1" />
+ <edge from-layer="1467" from-port="2" to-layer="1469" to-port="0" />
+ <edge from-layer="1468" from-port="0" to-layer="1469" to-port="1" />
+ <edge from-layer="1469" from-port="2" to-layer="1471" to-port="0" />
+ <edge from-layer="1470" from-port="0" to-layer="1471" to-port="1" />
+ <edge from-layer="1471" from-port="2" to-layer="1472" to-port="0" />
+ <edge from-layer="1472" from-port="1" to-layer="1473" to-port="1" />
+ <edge from-layer="1473" from-port="2" to-layer="1474" to-port="1" />
+ <edge from-layer="1474" from-port="2" to-layer="1478" to-port="0" />
+ <edge from-layer="1474" from-port="2" to-layer="1591" to-port="0" />
+ <edge from-layer="1474" from-port="2" to-layer="1485" to-port="0" />
+ <edge from-layer="1475" from-port="0" to-layer="1486" to-port="0" />
+ <edge from-layer="1476" from-port="0" to-layer="1484" to-port="0" />
+ <edge from-layer="1477" from-port="0" to-layer="1478" to-port="1" />
+ <edge from-layer="1478" from-port="2" to-layer="1480" to-port="0" />
+ <edge from-layer="1479" from-port="0" to-layer="1480" to-port="1" />
+ <edge from-layer="1480" from-port="2" to-layer="1482" to-port="0" />
+ <edge from-layer="1481" from-port="0" to-layer="1482" to-port="1" />
+ <edge from-layer="1482" from-port="2" to-layer="1483" to-port="0" />
+ <edge from-layer="1483" from-port="1" to-layer="1484" to-port="1" />
+ <edge from-layer="1484" from-port="2" to-layer="1485" to-port="1" />
+ <edge from-layer="1485" from-port="2" to-layer="1486" to-port="1" />
+ <edge from-layer="1486" from-port="2" to-layer="1529" to-port="0" />
+ <edge from-layer="1486" from-port="2" to-layer="1497" to-port="0" />
+ <edge from-layer="1486" from-port="2" to-layer="1566" to-port="0" />
+ <edge from-layer="1487" from-port="0" to-layer="1488" to-port="0" />
+ <edge from-layer="1488" from-port="1" to-layer="1491" to-port="0" />
+ <edge from-layer="1489" from-port="0" to-layer="1490" to-port="0" />
+ <edge from-layer="1490" from-port="1" to-layer="1491" to-port="1" />
+ <edge from-layer="1491" from-port="2" to-layer="1493" to-port="0" />
+ <edge from-layer="1492" from-port="0" to-layer="1493" to-port="1" />
+ <edge from-layer="1493" from-port="2" to-layer="1495" to-port="0" />
+ <edge from-layer="1494" from-port="0" to-layer="1495" to-port="1" />
+ <edge from-layer="1495" from-port="2" to-layer="1496" to-port="0" />
+ <edge from-layer="1496" from-port="1" to-layer="1497" to-port="1" />
+ <edge from-layer="1497" from-port="2" to-layer="1499" to-port="0" />
+ <edge from-layer="1498" from-port="0" to-layer="1499" to-port="1" />
+ <edge from-layer="1499" from-port="2" to-layer="1501" to-port="0" />
+ <edge from-layer="1500" from-port="0" to-layer="1501" to-port="1" />
+ <edge from-layer="1501" from-port="2" to-layer="1506" to-port="0" />
+ <edge from-layer="1501" from-port="2" to-layer="1502" to-port="0" />
+ <edge from-layer="1501" from-port="2" to-layer="1512" to-port="0" />
+ <edge from-layer="1502" from-port="2" to-layer="1515" to-port="0" />
+ <edge from-layer="1503" from-port="0" to-layer="1506" to-port="1" />
+ <edge from-layer="1504" from-port="0" to-layer="1506" to-port="2" />
+ <edge from-layer="1505" from-port="0" to-layer="1506" to-port="3" />
+ <edge from-layer="1506" from-port="5" to-layer="1508" to-port="0" />
+ <edge from-layer="1507" from-port="0" to-layer="1508" to-port="1" />
+ <edge from-layer="1508" from-port="2" to-layer="1513" to-port="0" />
+ <edge from-layer="1509" from-port="0" to-layer="1512" to-port="1" />
+ <edge from-layer="1510" from-port="0" to-layer="1512" to-port="2" />
+ <edge from-layer="1511" from-port="0" to-layer="1512" to-port="3" />
+ <edge from-layer="1512" from-port="5" to-layer="1513" to-port="1" />
+ <edge from-layer="1513" from-port="2" to-layer="1514" to-port="0" />
+ <edge from-layer="1514" from-port="2" to-layer="1515" to-port="1" />
+ <edge from-layer="1515" from-port="2" to-layer="1576" to-port="0" />
+ <edge from-layer="1516" from-port="1" to-layer="1518" to-port="0" />
+ <edge from-layer="1517" from-port="0" to-layer="1518" to-port="2" />
+ <edge from-layer="1518" from-port="3" to-layer="1548" to-port="0" />
+ <edge from-layer="1519" from-port="0" to-layer="1520" to-port="0" />
+ <edge from-layer="1520" from-port="1" to-layer="1523" to-port="0" />
+ <edge from-layer="1521" from-port="0" to-layer="1522" to-port="0" />
+ <edge from-layer="1522" from-port="1" to-layer="1523" to-port="1" />
+ <edge from-layer="1523" from-port="2" to-layer="1525" to-port="0" />
+ <edge from-layer="1524" from-port="0" to-layer="1525" to-port="1" />
+ <edge from-layer="1525" from-port="2" to-layer="1527" to-port="0" />
+ <edge from-layer="1526" from-port="0" to-layer="1527" to-port="1" />
+ <edge from-layer="1527" from-port="2" to-layer="1528" to-port="0" />
+ <edge from-layer="1528" from-port="1" to-layer="1529" to-port="1" />
+ <edge from-layer="1529" from-port="2" to-layer="1531" to-port="0" />
+ <edge from-layer="1530" from-port="0" to-layer="1531" to-port="1" />
+ <edge from-layer="1531" from-port="2" to-layer="1533" to-port="0" />
+ <edge from-layer="1532" from-port="0" to-layer="1533" to-port="1" />
+ <edge from-layer="1533" from-port="2" to-layer="1544" to-port="0" />
+ <edge from-layer="1533" from-port="2" to-layer="1538" to-port="0" />
+ <edge from-layer="1533" from-port="2" to-layer="1534" to-port="0" />
+ <edge from-layer="1534" from-port="2" to-layer="1547" to-port="0" />
+ <edge from-layer="1535" from-port="0" to-layer="1538" to-port="1" />
+ <edge from-layer="1536" from-port="0" to-layer="1538" to-port="2" />
+ <edge from-layer="1537" from-port="0" to-layer="1538" to-port="3" />
+ <edge from-layer="1538" from-port="5" to-layer="1540" to-port="0" />
+ <edge from-layer="1539" from-port="0" to-layer="1540" to-port="1" />
+ <edge from-layer="1540" from-port="2" to-layer="1545" to-port="0" />
+ <edge from-layer="1541" from-port="0" to-layer="1544" to-port="1" />
+ <edge from-layer="1542" from-port="0" to-layer="1544" to-port="2" />
+ <edge from-layer="1543" from-port="0" to-layer="1544" to-port="3" />
+ <edge from-layer="1544" from-port="5" to-layer="1545" to-port="1" />
+ <edge from-layer="1545" from-port="2" to-layer="1546" to-port="0" />
+ <edge from-layer="1546" from-port="2" to-layer="1547" to-port="1" />
+ <edge from-layer="1547" from-port="2" to-layer="1548" to-port="1" />
+ <edge from-layer="1548" from-port="2" to-layer="1549" to-port="0" />
+ <edge from-layer="1548" from-port="2" to-layer="1703" to-port="0" />
+ <edge from-layer="1549" from-port="2" to-layer="1550" to-port="0" />
+ <edge from-layer="1550" from-port="2" to-layer="1552" to-port="0" />
+ <edge from-layer="1551" from-port="0" to-layer="1552" to-port="1" />
+ <edge from-layer="1552" from-port="2" to-layer="1576" to-port="1" />
+ <edge from-layer="1553" from-port="1" to-layer="1555" to-port="0" />
+ <edge from-layer="1554" from-port="0" to-layer="1555" to-port="2" />
+ <edge from-layer="1555" from-port="3" to-layer="1571" to-port="0" />
+ <edge from-layer="1556" from-port="0" to-layer="1557" to-port="0" />
+ <edge from-layer="1557" from-port="1" to-layer="1560" to-port="0" />
+ <edge from-layer="1558" from-port="0" to-layer="1559" to-port="0" />
+ <edge from-layer="1559" from-port="1" to-layer="1560" to-port="1" />
+ <edge from-layer="1560" from-port="2" to-layer="1562" to-port="0" />
+ <edge from-layer="1561" from-port="0" to-layer="1562" to-port="1" />
+ <edge from-layer="1562" from-port="2" to-layer="1564" to-port="0" />
+ <edge from-layer="1563" from-port="0" to-layer="1564" to-port="1" />
+ <edge from-layer="1564" from-port="2" to-layer="1565" to-port="0" />
+ <edge from-layer="1565" from-port="1" to-layer="1566" to-port="1" />
+ <edge from-layer="1566" from-port="2" to-layer="1568" to-port="0" />
+ <edge from-layer="1567" from-port="0" to-layer="1568" to-port="1" />
+ <edge from-layer="1568" from-port="2" to-layer="1570" to-port="0" />
+ <edge from-layer="1569" from-port="0" to-layer="1570" to-port="1" />
+ <edge from-layer="1570" from-port="2" to-layer="1571" to-port="1" />
+ <edge from-layer="1571" from-port="2" to-layer="1572" to-port="0" />
+ <edge from-layer="1571" from-port="2" to-layer="1702" to-port="0" />
+ <edge from-layer="1572" from-port="2" to-layer="1573" to-port="0" />
+ <edge from-layer="1573" from-port="2" to-layer="1575" to-port="0" />
+ <edge from-layer="1574" from-port="0" to-layer="1575" to-port="1" />
+ <edge from-layer="1575" from-port="2" to-layer="1576" to-port="2" />
+ <edge from-layer="1576" from-port="4" to-layer="1578" to-port="0" />
+ <edge from-layer="1577" from-port="0" to-layer="1578" to-port="1" />
+ <edge from-layer="1578" from-port="2" to-layer="1579" to-port="0" />
+ <edge from-layer="1579" from-port="2" to-layer="1590" to-port="0" />
+ <edge from-layer="1580" from-port="0" to-layer="1581" to-port="0" />
+ <edge from-layer="1581" from-port="1" to-layer="1584" to-port="0" />
+ <edge from-layer="1582" from-port="0" to-layer="1583" to-port="0" />
+ <edge from-layer="1583" from-port="1" to-layer="1584" to-port="1" />
+ <edge from-layer="1584" from-port="2" to-layer="1586" to-port="0" />
+ <edge from-layer="1585" from-port="0" to-layer="1586" to-port="1" />
+ <edge from-layer="1586" from-port="2" to-layer="1588" to-port="0" />
+ <edge from-layer="1587" from-port="0" to-layer="1588" to-port="1" />
+ <edge from-layer="1588" from-port="2" to-layer="1589" to-port="0" />
+ <edge from-layer="1589" from-port="1" to-layer="1590" to-port="1" />
+ <edge from-layer="1590" from-port="2" to-layer="1591" to-port="1" />
+ <edge from-layer="1591" from-port="2" to-layer="1639" to-port="0" />
+ <edge from-layer="1591" from-port="2" to-layer="1602" to-port="0" />
+ <edge from-layer="1591" from-port="2" to-layer="1595" to-port="0" />
+ <edge from-layer="1592" from-port="0" to-layer="1603" to-port="0" />
+ <edge from-layer="1593" from-port="0" to-layer="1601" to-port="0" />
+ <edge from-layer="1594" from-port="0" to-layer="1595" to-port="1" />
+ <edge from-layer="1595" from-port="2" to-layer="1597" to-port="0" />
+ <edge from-layer="1596" from-port="0" to-layer="1597" to-port="1" />
+ <edge from-layer="1597" from-port="2" to-layer="1599" to-port="0" />
+ <edge from-layer="1598" from-port="0" to-layer="1599" to-port="1" />
+ <edge from-layer="1599" from-port="2" to-layer="1600" to-port="0" />
+ <edge from-layer="1600" from-port="1" to-layer="1601" to-port="1" />
+ <edge from-layer="1601" from-port="2" to-layer="1602" to-port="1" />
+ <edge from-layer="1602" from-port="2" to-layer="1603" to-port="1" />
+ <edge from-layer="1603" from-port="2" to-layer="1614" to-port="0" />
+ <edge from-layer="1603" from-port="2" to-layer="1626" to-port="0" />
+ <edge from-layer="1604" from-port="0" to-layer="1605" to-port="0" />
+ <edge from-layer="1605" from-port="1" to-layer="1608" to-port="0" />
+ <edge from-layer="1606" from-port="0" to-layer="1607" to-port="0" />
+ <edge from-layer="1607" from-port="1" to-layer="1608" to-port="1" />
+ <edge from-layer="1608" from-port="2" to-layer="1610" to-port="0" />
+ <edge from-layer="1609" from-port="0" to-layer="1610" to-port="1" />
+ <edge from-layer="1610" from-port="2" to-layer="1612" to-port="0" />
+ <edge from-layer="1611" from-port="0" to-layer="1612" to-port="1" />
+ <edge from-layer="1612" from-port="2" to-layer="1613" to-port="0" />
+ <edge from-layer="1613" from-port="1" to-layer="1614" to-port="1" />
+ <edge from-layer="1614" from-port="2" to-layer="1615" to-port="0" />
+ <edge from-layer="1615" from-port="1" to-layer="1627" to-port="0" />
+ <edge from-layer="1616" from-port="0" to-layer="1617" to-port="0" />
+ <edge from-layer="1617" from-port="1" to-layer="1620" to-port="0" />
+ <edge from-layer="1618" from-port="0" to-layer="1619" to-port="0" />
+ <edge from-layer="1619" from-port="1" to-layer="1620" to-port="1" />
+ <edge from-layer="1620" from-port="2" to-layer="1622" to-port="0" />
+ <edge from-layer="1621" from-port="0" to-layer="1622" to-port="1" />
+ <edge from-layer="1622" from-port="2" to-layer="1624" to-port="0" />
+ <edge from-layer="1623" from-port="0" to-layer="1624" to-port="1" />
+ <edge from-layer="1624" from-port="2" to-layer="1625" to-port="0" />
+ <edge from-layer="1625" from-port="1" to-layer="1626" to-port="1" />
+ <edge from-layer="1626" from-port="2" to-layer="1627" to-port="1" />
+ <edge from-layer="1627" from-port="2" to-layer="1638" to-port="0" />
+ <edge from-layer="1628" from-port="0" to-layer="1629" to-port="0" />
+ <edge from-layer="1629" from-port="1" to-layer="1632" to-port="0" />
+ <edge from-layer="1630" from-port="0" to-layer="1631" to-port="0" />
+ <edge from-layer="1631" from-port="1" to-layer="1632" to-port="1" />
+ <edge from-layer="1632" from-port="2" to-layer="1634" to-port="0" />
+ <edge from-layer="1633" from-port="0" to-layer="1634" to-port="1" />
+ <edge from-layer="1634" from-port="2" to-layer="1636" to-port="0" />
+ <edge from-layer="1635" from-port="0" to-layer="1636" to-port="1" />
+ <edge from-layer="1636" from-port="2" to-layer="1637" to-port="0" />
+ <edge from-layer="1637" from-port="1" to-layer="1638" to-port="1" />
+ <edge from-layer="1638" from-port="2" to-layer="1639" to-port="1" />
+ <edge from-layer="1639" from-port="2" to-layer="1649" to-port="0" />
+ <edge from-layer="1639" from-port="2" to-layer="1642" to-port="0" />
+ <edge from-layer="1639" from-port="2" to-layer="1772" to-port="0" />
+ <edge from-layer="1640" from-port="0" to-layer="1648" to-port="0" />
+ <edge from-layer="1641" from-port="0" to-layer="1642" to-port="1" />
+ <edge from-layer="1642" from-port="2" to-layer="1644" to-port="0" />
+ <edge from-layer="1643" from-port="0" to-layer="1644" to-port="1" />
+ <edge from-layer="1644" from-port="2" to-layer="1646" to-port="0" />
+ <edge from-layer="1645" from-port="0" to-layer="1646" to-port="1" />
+ <edge from-layer="1646" from-port="2" to-layer="1647" to-port="0" />
+ <edge from-layer="1647" from-port="1" to-layer="1648" to-port="1" />
+ <edge from-layer="1648" from-port="2" to-layer="1649" to-port="1" />
+ <edge from-layer="1649" from-port="2" to-layer="1650" to-port="1" />
+ <edge from-layer="1650" from-port="2" to-layer="1661" to-port="0" />
+ <edge from-layer="1650" from-port="2" to-layer="1730" to-port="0" />
+ <edge from-layer="1650" from-port="2" to-layer="1681" to-port="0" />
+ <edge from-layer="1651" from-port="0" to-layer="1652" to-port="0" />
+ <edge from-layer="1652" from-port="1" to-layer="1655" to-port="0" />
+ <edge from-layer="1653" from-port="0" to-layer="1654" to-port="0" />
+ <edge from-layer="1654" from-port="1" to-layer="1655" to-port="1" />
+ <edge from-layer="1655" from-port="2" to-layer="1657" to-port="0" />
+ <edge from-layer="1656" from-port="0" to-layer="1657" to-port="1" />
+ <edge from-layer="1657" from-port="2" to-layer="1659" to-port="0" />
+ <edge from-layer="1658" from-port="0" to-layer="1659" to-port="1" />
+ <edge from-layer="1659" from-port="2" to-layer="1660" to-port="0" />
+ <edge from-layer="1660" from-port="1" to-layer="1661" to-port="1" />
+ <edge from-layer="1661" from-port="2" to-layer="1663" to-port="0" />
+ <edge from-layer="1662" from-port="0" to-layer="1663" to-port="1" />
+ <edge from-layer="1663" from-port="2" to-layer="1665" to-port="0" />
+ <edge from-layer="1664" from-port="0" to-layer="1665" to-port="1" />
+ <edge from-layer="1665" from-port="2" to-layer="1666" to-port="1" />
+ <edge from-layer="1666" from-port="2" to-layer="1667" to-port="0" />
+ <edge from-layer="1666" from-port="2" to-layer="1753" to-port="0" />
+ <edge from-layer="1668" from-port="1" to-layer="1670" to-port="0" />
+ <edge from-layer="1669" from-port="0" to-layer="1670" to-port="2" />
+ <edge from-layer="1670" from-port="3" to-layer="1700" to-port="0" />
+ <edge from-layer="1671" from-port="0" to-layer="1672" to-port="0" />
+ <edge from-layer="1672" from-port="1" to-layer="1675" to-port="0" />
+ <edge from-layer="1673" from-port="0" to-layer="1674" to-port="0" />
+ <edge from-layer="1674" from-port="1" to-layer="1675" to-port="1" />
+ <edge from-layer="1675" from-port="2" to-layer="1677" to-port="0" />
+ <edge from-layer="1676" from-port="0" to-layer="1677" to-port="1" />
+ <edge from-layer="1677" from-port="2" to-layer="1679" to-port="0" />
+ <edge from-layer="1678" from-port="0" to-layer="1679" to-port="1" />
+ <edge from-layer="1679" from-port="2" to-layer="1680" to-port="0" />
+ <edge from-layer="1680" from-port="1" to-layer="1681" to-port="1" />
+ <edge from-layer="1681" from-port="2" to-layer="1683" to-port="0" />
+ <edge from-layer="1682" from-port="0" to-layer="1683" to-port="1" />
+ <edge from-layer="1683" from-port="2" to-layer="1685" to-port="0" />
+ <edge from-layer="1684" from-port="0" to-layer="1685" to-port="1" />
+ <edge from-layer="1685" from-port="2" to-layer="1686" to-port="0" />
+ <edge from-layer="1685" from-port="2" to-layer="1690" to-port="0" />
+ <edge from-layer="1685" from-port="2" to-layer="1696" to-port="0" />
+ <edge from-layer="1686" from-port="2" to-layer="1699" to-port="0" />
+ <edge from-layer="1687" from-port="0" to-layer="1690" to-port="1" />
+ <edge from-layer="1688" from-port="0" to-layer="1690" to-port="2" />
+ <edge from-layer="1689" from-port="0" to-layer="1690" to-port="3" />
+ <edge from-layer="1690" from-port="5" to-layer="1692" to-port="0" />
+ <edge from-layer="1691" from-port="0" to-layer="1692" to-port="1" />
+ <edge from-layer="1692" from-port="2" to-layer="1697" to-port="0" />
+ <edge from-layer="1693" from-port="0" to-layer="1696" to-port="1" />
+ <edge from-layer="1694" from-port="0" to-layer="1696" to-port="2" />
+ <edge from-layer="1695" from-port="0" to-layer="1696" to-port="3" />
+ <edge from-layer="1696" from-port="5" to-layer="1697" to-port="1" />
+ <edge from-layer="1697" from-port="2" to-layer="1698" to-port="0" />
+ <edge from-layer="1698" from-port="2" to-layer="1699" to-port="1" />
+ <edge from-layer="1699" from-port="2" to-layer="1700" to-port="1" />
+ <edge from-layer="1700" from-port="2" to-layer="1701" to-port="0" />
+ <edge from-layer="1700" from-port="2" to-layer="1749" to-port="0" />
+ <edge from-layer="1716" from-port="1" to-layer="1718" to-port="0" />
+ <edge from-layer="1717" from-port="0" to-layer="1718" to-port="2" />
+ <edge from-layer="1718" from-port="3" to-layer="2672" to-port="0" />
+ <edge from-layer="1719" from-port="0" to-layer="2656" to-port="0" />
+ <edge from-layer="1720" from-port="0" to-layer="1721" to-port="0" />
+ <edge from-layer="1721" from-port="1" to-layer="1724" to-port="0" />
+ <edge from-layer="1722" from-port="0" to-layer="1723" to-port="0" />
+ <edge from-layer="1723" from-port="1" to-layer="1724" to-port="1" />
+ <edge from-layer="1724" from-port="2" to-layer="1726" to-port="0" />
+ <edge from-layer="1725" from-port="0" to-layer="1726" to-port="1" />
+ <edge from-layer="1726" from-port="2" to-layer="1728" to-port="0" />
+ <edge from-layer="1727" from-port="0" to-layer="1728" to-port="1" />
+ <edge from-layer="1728" from-port="2" to-layer="1729" to-port="0" />
+ <edge from-layer="1729" from-port="1" to-layer="1730" to-port="1" />
+ <edge from-layer="1730" from-port="2" to-layer="1732" to-port="0" />
+ <edge from-layer="1731" from-port="0" to-layer="1732" to-port="1" />
+ <edge from-layer="1732" from-port="2" to-layer="1734" to-port="0" />
+ <edge from-layer="1733" from-port="0" to-layer="1734" to-port="1" />
+ <edge from-layer="1734" from-port="2" to-layer="1735" to-port="0" />
+ <edge from-layer="1734" from-port="2" to-layer="1739" to-port="0" />
+ <edge from-layer="1734" from-port="2" to-layer="1745" to-port="0" />
+ <edge from-layer="1735" from-port="2" to-layer="1748" to-port="0" />
+ <edge from-layer="1736" from-port="0" to-layer="1739" to-port="1" />
+ <edge from-layer="1737" from-port="0" to-layer="1739" to-port="2" />
+ <edge from-layer="1738" from-port="0" to-layer="1739" to-port="3" />
+ <edge from-layer="1739" from-port="5" to-layer="1741" to-port="0" />
+ <edge from-layer="1740" from-port="0" to-layer="1741" to-port="1" />
+ <edge from-layer="1741" from-port="2" to-layer="1746" to-port="0" />
+ <edge from-layer="1742" from-port="0" to-layer="1745" to-port="1" />
+ <edge from-layer="1743" from-port="0" to-layer="1745" to-port="2" />
+ <edge from-layer="1744" from-port="0" to-layer="1745" to-port="3" />
+ <edge from-layer="1745" from-port="5" to-layer="1746" to-port="1" />
+ <edge from-layer="1746" from-port="2" to-layer="1747" to-port="0" />
+ <edge from-layer="1747" from-port="2" to-layer="1748" to-port="1" />
+ <edge from-layer="1748" from-port="2" to-layer="1757" to-port="0" />
+ <edge from-layer="1749" from-port="2" to-layer="1750" to-port="0" />
+ <edge from-layer="1750" from-port="2" to-layer="1752" to-port="0" />
+ <edge from-layer="1751" from-port="0" to-layer="1752" to-port="1" />
+ <edge from-layer="1752" from-port="2" to-layer="1757" to-port="1" />
+ <edge from-layer="1753" from-port="2" to-layer="1754" to-port="0" />
+ <edge from-layer="1754" from-port="2" to-layer="1756" to-port="0" />
+ <edge from-layer="1755" from-port="0" to-layer="1756" to-port="1" />
+ <edge from-layer="1756" from-port="2" to-layer="1757" to-port="2" />
+ <edge from-layer="1757" from-port="4" to-layer="1759" to-port="0" />
+ <edge from-layer="1758" from-port="0" to-layer="1759" to-port="1" />
+ <edge from-layer="1759" from-port="2" to-layer="1760" to-port="0" />
+ <edge from-layer="1760" from-port="2" to-layer="1771" to-port="0" />
+ <edge from-layer="1761" from-port="0" to-layer="1762" to-port="0" />
+ <edge from-layer="1762" from-port="1" to-layer="1765" to-port="0" />
+ <edge from-layer="1763" from-port="0" to-layer="1764" to-port="0" />
+ <edge from-layer="1764" from-port="1" to-layer="1765" to-port="1" />
+ <edge from-layer="1765" from-port="2" to-layer="1767" to-port="0" />
+ <edge from-layer="1766" from-port="0" to-layer="1767" to-port="1" />
+ <edge from-layer="1767" from-port="2" to-layer="1769" to-port="0" />
+ <edge from-layer="1768" from-port="0" to-layer="1769" to-port="1" />
+ <edge from-layer="1769" from-port="2" to-layer="1770" to-port="0" />
+ <edge from-layer="1770" from-port="1" to-layer="1771" to-port="1" />
+ <edge from-layer="1771" from-port="2" to-layer="1772" to-port="1" />
+ <edge from-layer="1772" from-port="2" to-layer="1820" to-port="0" />
+ <edge from-layer="1772" from-port="2" to-layer="1776" to-port="0" />
+ <edge from-layer="1772" from-port="2" to-layer="1783" to-port="0" />
+ <edge from-layer="1773" from-port="0" to-layer="1784" to-port="0" />
+ <edge from-layer="1774" from-port="0" to-layer="1782" to-port="0" />
+ <edge from-layer="1775" from-port="0" to-layer="1776" to-port="1" />
+ <edge from-layer="1776" from-port="2" to-layer="1778" to-port="0" />
+ <edge from-layer="1777" from-port="0" to-layer="1778" to-port="1" />
+ <edge from-layer="1778" from-port="2" to-layer="1780" to-port="0" />
+ <edge from-layer="1779" from-port="0" to-layer="1780" to-port="1" />
+ <edge from-layer="1780" from-port="2" to-layer="1781" to-port="0" />
+ <edge from-layer="1781" from-port="1" to-layer="1782" to-port="1" />
+ <edge from-layer="1782" from-port="2" to-layer="1783" to-port="1" />
+ <edge from-layer="1783" from-port="2" to-layer="1784" to-port="1" />
+ <edge from-layer="1784" from-port="2" to-layer="1795" to-port="0" />
+ <edge from-layer="1784" from-port="2" to-layer="1807" to-port="0" />
+ <edge from-layer="1785" from-port="0" to-layer="1786" to-port="0" />
+ <edge from-layer="1786" from-port="1" to-layer="1789" to-port="0" />
+ <edge from-layer="1787" from-port="0" to-layer="1788" to-port="0" />
+ <edge from-layer="1788" from-port="1" to-layer="1789" to-port="1" />
+ <edge from-layer="1789" from-port="2" to-layer="1791" to-port="0" />
+ <edge from-layer="1790" from-port="0" to-layer="1791" to-port="1" />
+ <edge from-layer="1791" from-port="2" to-layer="1793" to-port="0" />
+ <edge from-layer="1792" from-port="0" to-layer="1793" to-port="1" />
+ <edge from-layer="1793" from-port="2" to-layer="1794" to-port="0" />
+ <edge from-layer="1794" from-port="1" to-layer="1795" to-port="1" />
+ <edge from-layer="1795" from-port="2" to-layer="1796" to-port="0" />
+ <edge from-layer="1796" from-port="1" to-layer="1808" to-port="0" />
+ <edge from-layer="1797" from-port="0" to-layer="1798" to-port="0" />
+ <edge from-layer="1798" from-port="1" to-layer="1801" to-port="0" />
+ <edge from-layer="1799" from-port="0" to-layer="1800" to-port="0" />
+ <edge from-layer="1800" from-port="1" to-layer="1801" to-port="1" />
+ <edge from-layer="1801" from-port="2" to-layer="1803" to-port="0" />
+ <edge from-layer="1802" from-port="0" to-layer="1803" to-port="1" />
+ <edge from-layer="1803" from-port="2" to-layer="1805" to-port="0" />
+ <edge from-layer="1804" from-port="0" to-layer="1805" to-port="1" />
+ <edge from-layer="1805" from-port="2" to-layer="1806" to-port="0" />
+ <edge from-layer="1806" from-port="1" to-layer="1807" to-port="1" />
+ <edge from-layer="1807" from-port="2" to-layer="1808" to-port="1" />
+ <edge from-layer="1808" from-port="2" to-layer="1819" to-port="0" />
+ <edge from-layer="1809" from-port="0" to-layer="1810" to-port="0" />
+ <edge from-layer="1810" from-port="1" to-layer="1813" to-port="0" />
+ <edge from-layer="1811" from-port="0" to-layer="1812" to-port="0" />
+ <edge from-layer="1812" from-port="1" to-layer="1813" to-port="1" />
+ <edge from-layer="1813" from-port="2" to-layer="1815" to-port="0" />
+ <edge from-layer="1814" from-port="0" to-layer="1815" to-port="1" />
+ <edge from-layer="1815" from-port="2" to-layer="1817" to-port="0" />
+ <edge from-layer="1816" from-port="0" to-layer="1817" to-port="1" />
+ <edge from-layer="1817" from-port="2" to-layer="1818" to-port="0" />
+ <edge from-layer="1818" from-port="1" to-layer="1819" to-port="1" />
+ <edge from-layer="1819" from-port="2" to-layer="1820" to-port="1" />
+ <edge from-layer="1820" from-port="2" to-layer="1831" to-port="0" />
+ <edge from-layer="1820" from-port="2" to-layer="1937" to-port="0" />
+ <edge from-layer="1820" from-port="2" to-layer="1824" to-port="0" />
+ <edge from-layer="1821" from-port="0" to-layer="1832" to-port="0" />
+ <edge from-layer="1822" from-port="0" to-layer="1830" to-port="0" />
+ <edge from-layer="1823" from-port="0" to-layer="1824" to-port="1" />
+ <edge from-layer="1824" from-port="2" to-layer="1826" to-port="0" />
+ <edge from-layer="1825" from-port="0" to-layer="1826" to-port="1" />
+ <edge from-layer="1826" from-port="2" to-layer="1828" to-port="0" />
+ <edge from-layer="1827" from-port="0" to-layer="1828" to-port="1" />
+ <edge from-layer="1828" from-port="2" to-layer="1829" to-port="0" />
+ <edge from-layer="1829" from-port="1" to-layer="1830" to-port="1" />
+ <edge from-layer="1830" from-port="2" to-layer="1831" to-port="1" />
+ <edge from-layer="1831" from-port="2" to-layer="1832" to-port="1" />
+ <edge from-layer="1832" from-port="2" to-layer="1912" to-port="0" />
+ <edge from-layer="1832" from-port="2" to-layer="1875" to-port="0" />
+ <edge from-layer="1832" from-port="2" to-layer="1843" to-port="0" />
+ <edge from-layer="1833" from-port="0" to-layer="1834" to-port="0" />
+ <edge from-layer="1834" from-port="1" to-layer="1837" to-port="0" />
+ <edge from-layer="1835" from-port="0" to-layer="1836" to-port="0" />
+ <edge from-layer="1836" from-port="1" to-layer="1837" to-port="1" />
+ <edge from-layer="1837" from-port="2" to-layer="1839" to-port="0" />
+ <edge from-layer="1838" from-port="0" to-layer="1839" to-port="1" />
+ <edge from-layer="1839" from-port="2" to-layer="1841" to-port="0" />
+ <edge from-layer="1840" from-port="0" to-layer="1841" to-port="1" />
+ <edge from-layer="1841" from-port="2" to-layer="1842" to-port="0" />
+ <edge from-layer="1842" from-port="1" to-layer="1843" to-port="1" />
+ <edge from-layer="1843" from-port="2" to-layer="1845" to-port="0" />
+ <edge from-layer="1844" from-port="0" to-layer="1845" to-port="1" />
+ <edge from-layer="1845" from-port="2" to-layer="1847" to-port="0" />
+ <edge from-layer="1846" from-port="0" to-layer="1847" to-port="1" />
+ <edge from-layer="1847" from-port="2" to-layer="1848" to-port="0" />
+ <edge from-layer="1847" from-port="2" to-layer="1852" to-port="0" />
+ <edge from-layer="1847" from-port="2" to-layer="1858" to-port="0" />
+ <edge from-layer="1848" from-port="2" to-layer="1861" to-port="0" />
+ <edge from-layer="1849" from-port="0" to-layer="1852" to-port="1" />
+ <edge from-layer="1850" from-port="0" to-layer="1852" to-port="2" />
+ <edge from-layer="1851" from-port="0" to-layer="1852" to-port="3" />
+ <edge from-layer="1852" from-port="5" to-layer="1854" to-port="0" />
+ <edge from-layer="1853" from-port="0" to-layer="1854" to-port="1" />
+ <edge from-layer="1854" from-port="2" to-layer="1859" to-port="0" />
+ <edge from-layer="1855" from-port="0" to-layer="1858" to-port="1" />
+ <edge from-layer="1856" from-port="0" to-layer="1858" to-port="2" />
+ <edge from-layer="1857" from-port="0" to-layer="1858" to-port="3" />
+ <edge from-layer="1858" from-port="5" to-layer="1859" to-port="1" />
+ <edge from-layer="1859" from-port="2" to-layer="1860" to-port="0" />
+ <edge from-layer="1860" from-port="2" to-layer="1861" to-port="1" />
+ <edge from-layer="1861" from-port="2" to-layer="1922" to-port="0" />
+ <edge from-layer="1862" from-port="1" to-layer="1864" to-port="0" />
+ <edge from-layer="1863" from-port="0" to-layer="1864" to-port="2" />
+ <edge from-layer="1864" from-port="3" to-layer="1894" to-port="0" />
+ <edge from-layer="1865" from-port="0" to-layer="1866" to-port="0" />
+ <edge from-layer="1866" from-port="1" to-layer="1869" to-port="0" />
+ <edge from-layer="1867" from-port="0" to-layer="1868" to-port="0" />
+ <edge from-layer="1868" from-port="1" to-layer="1869" to-port="1" />
+ <edge from-layer="1869" from-port="2" to-layer="1871" to-port="0" />
+ <edge from-layer="1870" from-port="0" to-layer="1871" to-port="1" />
+ <edge from-layer="1871" from-port="2" to-layer="1873" to-port="0" />
+ <edge from-layer="1872" from-port="0" to-layer="1873" to-port="1" />
+ <edge from-layer="1873" from-port="2" to-layer="1874" to-port="0" />
+ <edge from-layer="1874" from-port="1" to-layer="1875" to-port="1" />
+ <edge from-layer="1875" from-port="2" to-layer="1877" to-port="0" />
+ <edge from-layer="1876" from-port="0" to-layer="1877" to-port="1" />
+ <edge from-layer="1877" from-port="2" to-layer="1879" to-port="0" />
+ <edge from-layer="1878" from-port="0" to-layer="1879" to-port="1" />
+ <edge from-layer="1879" from-port="2" to-layer="1880" to-port="0" />
+ <edge from-layer="1879" from-port="2" to-layer="1884" to-port="0" />
+ <edge from-layer="1879" from-port="2" to-layer="1890" to-port="0" />
+ <edge from-layer="1880" from-port="2" to-layer="1893" to-port="0" />
+ <edge from-layer="1881" from-port="0" to-layer="1884" to-port="1" />
+ <edge from-layer="1882" from-port="0" to-layer="1884" to-port="2" />
+ <edge from-layer="1883" from-port="0" to-layer="1884" to-port="3" />
+ <edge from-layer="1884" from-port="5" to-layer="1886" to-port="0" />
+ <edge from-layer="1885" from-port="0" to-layer="1886" to-port="1" />
+ <edge from-layer="1886" from-port="2" to-layer="1891" to-port="0" />
+ <edge from-layer="1887" from-port="0" to-layer="1890" to-port="1" />
+ <edge from-layer="1888" from-port="0" to-layer="1890" to-port="2" />
+ <edge from-layer="1889" from-port="0" to-layer="1890" to-port="3" />
+ <edge from-layer="1890" from-port="5" to-layer="1891" to-port="1" />
+ <edge from-layer="1891" from-port="2" to-layer="1892" to-port="0" />
+ <edge from-layer="1892" from-port="2" to-layer="1893" to-port="1" />
+ <edge from-layer="1893" from-port="2" to-layer="1894" to-port="1" />
+ <edge from-layer="1894" from-port="2" to-layer="2717" to-port="0" />
+ <edge from-layer="1894" from-port="2" to-layer="1895" to-port="0" />
+ <edge from-layer="1895" from-port="2" to-layer="1896" to-port="0" />
+ <edge from-layer="1896" from-port="2" to-layer="1898" to-port="0" />
+ <edge from-layer="1897" from-port="0" to-layer="1898" to-port="1" />
+ <edge from-layer="1898" from-port="2" to-layer="1922" to-port="1" />
+ <edge from-layer="1899" from-port="1" to-layer="1901" to-port="0" />
+ <edge from-layer="1900" from-port="0" to-layer="1901" to-port="2" />
+ <edge from-layer="1901" from-port="3" to-layer="1917" to-port="0" />
+ <edge from-layer="1902" from-port="0" to-layer="1903" to-port="0" />
+ <edge from-layer="1903" from-port="1" to-layer="1906" to-port="0" />
+ <edge from-layer="1904" from-port="0" to-layer="1905" to-port="0" />
+ <edge from-layer="1905" from-port="1" to-layer="1906" to-port="1" />
+ <edge from-layer="1906" from-port="2" to-layer="1908" to-port="0" />
+ <edge from-layer="1907" from-port="0" to-layer="1908" to-port="1" />
+ <edge from-layer="1908" from-port="2" to-layer="1910" to-port="0" />
+ <edge from-layer="1909" from-port="0" to-layer="1910" to-port="1" />
+ <edge from-layer="1910" from-port="2" to-layer="1911" to-port="0" />
+ <edge from-layer="1911" from-port="1" to-layer="1912" to-port="1" />
+ <edge from-layer="1912" from-port="2" to-layer="1914" to-port="0" />
+ <edge from-layer="1913" from-port="0" to-layer="1914" to-port="1" />
+ <edge from-layer="1914" from-port="2" to-layer="1916" to-port="0" />
+ <edge from-layer="1915" from-port="0" to-layer="1916" to-port="1" />
+ <edge from-layer="1916" from-port="2" to-layer="1917" to-port="1" />
+ <edge from-layer="1917" from-port="2" to-layer="2716" to-port="0" />
+ <edge from-layer="1917" from-port="2" to-layer="1918" to-port="0" />
+ <edge from-layer="1918" from-port="2" to-layer="1919" to-port="0" />
+ <edge from-layer="1919" from-port="2" to-layer="1921" to-port="0" />
+ <edge from-layer="1920" from-port="0" to-layer="1921" to-port="1" />
+ <edge from-layer="1921" from-port="2" to-layer="1922" to-port="2" />
+ <edge from-layer="1922" from-port="4" to-layer="1924" to-port="0" />
+ <edge from-layer="1923" from-port="0" to-layer="1924" to-port="1" />
+ <edge from-layer="1924" from-port="2" to-layer="1925" to-port="0" />
+ <edge from-layer="1925" from-port="2" to-layer="1936" to-port="0" />
+ <edge from-layer="1926" from-port="0" to-layer="1927" to-port="0" />
+ <edge from-layer="1927" from-port="1" to-layer="1930" to-port="0" />
+ <edge from-layer="1928" from-port="0" to-layer="1929" to-port="0" />
+ <edge from-layer="1929" from-port="1" to-layer="1930" to-port="1" />
+ <edge from-layer="1930" from-port="2" to-layer="1932" to-port="0" />
+ <edge from-layer="1931" from-port="0" to-layer="1932" to-port="1" />
+ <edge from-layer="1932" from-port="2" to-layer="1934" to-port="0" />
+ <edge from-layer="1933" from-port="0" to-layer="1934" to-port="1" />
+ <edge from-layer="1934" from-port="2" to-layer="1935" to-port="0" />
+ <edge from-layer="1935" from-port="1" to-layer="1936" to-port="1" />
+ <edge from-layer="1936" from-port="2" to-layer="1937" to-port="1" />
+ <edge from-layer="1937" from-port="2" to-layer="1941" to-port="0" />
+ <edge from-layer="1937" from-port="2" to-layer="1948" to-port="0" />
+ <edge from-layer="1937" from-port="2" to-layer="1985" to-port="0" />
+ <edge from-layer="1938" from-port="0" to-layer="1949" to-port="0" />
+ <edge from-layer="1939" from-port="0" to-layer="1947" to-port="0" />
+ <edge from-layer="1940" from-port="0" to-layer="1941" to-port="1" />
+ <edge from-layer="1941" from-port="2" to-layer="1943" to-port="0" />
+ <edge from-layer="1942" from-port="0" to-layer="1943" to-port="1" />
+ <edge from-layer="1943" from-port="2" to-layer="1945" to-port="0" />
+ <edge from-layer="1944" from-port="0" to-layer="1945" to-port="1" />
+ <edge from-layer="1945" from-port="2" to-layer="1946" to-port="0" />
+ <edge from-layer="1946" from-port="1" to-layer="1947" to-port="1" />
+ <edge from-layer="1947" from-port="2" to-layer="1948" to-port="1" />
+ <edge from-layer="1948" from-port="2" to-layer="1949" to-port="1" />
+ <edge from-layer="1949" from-port="2" to-layer="1960" to-port="0" />
+ <edge from-layer="1949" from-port="2" to-layer="1972" to-port="0" />
+ <edge from-layer="1950" from-port="0" to-layer="1951" to-port="0" />
+ <edge from-layer="1951" from-port="1" to-layer="1954" to-port="0" />
+ <edge from-layer="1952" from-port="0" to-layer="1953" to-port="0" />
+ <edge from-layer="1953" from-port="1" to-layer="1954" to-port="1" />
+ <edge from-layer="1954" from-port="2" to-layer="1956" to-port="0" />
+ <edge from-layer="1955" from-port="0" to-layer="1956" to-port="1" />
+ <edge from-layer="1956" from-port="2" to-layer="1958" to-port="0" />
+ <edge from-layer="1957" from-port="0" to-layer="1958" to-port="1" />
+ <edge from-layer="1958" from-port="2" to-layer="1959" to-port="0" />
+ <edge from-layer="1959" from-port="1" to-layer="1960" to-port="1" />
+ <edge from-layer="1960" from-port="2" to-layer="1961" to-port="0" />
+ <edge from-layer="1961" from-port="1" to-layer="1973" to-port="0" />
+ <edge from-layer="1962" from-port="0" to-layer="1963" to-port="0" />
+ <edge from-layer="1963" from-port="1" to-layer="1966" to-port="0" />
+ <edge from-layer="1964" from-port="0" to-layer="1965" to-port="0" />
+ <edge from-layer="1965" from-port="1" to-layer="1966" to-port="1" />
+ <edge from-layer="1966" from-port="2" to-layer="1968" to-port="0" />
+ <edge from-layer="1967" from-port="0" to-layer="1968" to-port="1" />
+ <edge from-layer="1968" from-port="2" to-layer="1970" to-port="0" />
+ <edge from-layer="1969" from-port="0" to-layer="1970" to-port="1" />
+ <edge from-layer="1970" from-port="2" to-layer="1971" to-port="0" />
+ <edge from-layer="1971" from-port="1" to-layer="1972" to-port="1" />
+ <edge from-layer="1972" from-port="2" to-layer="1973" to-port="1" />
+ <edge from-layer="1973" from-port="2" to-layer="1984" to-port="0" />
+ <edge from-layer="1974" from-port="0" to-layer="1975" to-port="0" />
+ <edge from-layer="1975" from-port="1" to-layer="1978" to-port="0" />
+ <edge from-layer="1976" from-port="0" to-layer="1977" to-port="0" />
+ <edge from-layer="1977" from-port="1" to-layer="1978" to-port="1" />
+ <edge from-layer="1978" from-port="2" to-layer="1980" to-port="0" />
+ <edge from-layer="1979" from-port="0" to-layer="1980" to-port="1" />
+ <edge from-layer="1980" from-port="2" to-layer="1982" to-port="0" />
+ <edge from-layer="1981" from-port="0" to-layer="1982" to-port="1" />
+ <edge from-layer="1982" from-port="2" to-layer="1983" to-port="0" />
+ <edge from-layer="1983" from-port="1" to-layer="1984" to-port="1" />
+ <edge from-layer="1984" from-port="2" to-layer="1985" to-port="1" />
+ <edge from-layer="1985" from-port="2" to-layer="1996" to-port="0" />
+ <edge from-layer="1985" from-port="2" to-layer="1989" to-port="0" />
+ <edge from-layer="1985" from-port="2" to-layer="2102" to-port="0" />
+ <edge from-layer="1986" from-port="0" to-layer="1997" to-port="0" />
+ <edge from-layer="1987" from-port="0" to-layer="1995" to-port="0" />
+ <edge from-layer="1988" from-port="0" to-layer="1989" to-port="1" />
+ <edge from-layer="1989" from-port="2" to-layer="1991" to-port="0" />
+ <edge from-layer="1990" from-port="0" to-layer="1991" to-port="1" />
+ <edge from-layer="1991" from-port="2" to-layer="1993" to-port="0" />
+ <edge from-layer="1992" from-port="0" to-layer="1993" to-port="1" />
+ <edge from-layer="1993" from-port="2" to-layer="1994" to-port="0" />
+ <edge from-layer="1994" from-port="1" to-layer="1995" to-port="1" />
+ <edge from-layer="1995" from-port="2" to-layer="1996" to-port="1" />
+ <edge from-layer="1996" from-port="2" to-layer="1997" to-port="1" />
+ <edge from-layer="1997" from-port="2" to-layer="2077" to-port="0" />
+ <edge from-layer="1997" from-port="2" to-layer="2008" to-port="0" />
+ <edge from-layer="1997" from-port="2" to-layer="2040" to-port="0" />
+ <edge from-layer="1998" from-port="0" to-layer="1999" to-port="0" />
+ <edge from-layer="1999" from-port="1" to-layer="2002" to-port="0" />
+ <edge from-layer="2000" from-port="0" to-layer="2001" to-port="0" />
+ <edge from-layer="2001" from-port="1" to-layer="2002" to-port="1" />
+ <edge from-layer="2002" from-port="2" to-layer="2004" to-port="0" />
+ <edge from-layer="2003" from-port="0" to-layer="2004" to-port="1" />
+ <edge from-layer="2004" from-port="2" to-layer="2006" to-port="0" />
+ <edge from-layer="2005" from-port="0" to-layer="2006" to-port="1" />
+ <edge from-layer="2006" from-port="2" to-layer="2007" to-port="0" />
+ <edge from-layer="2007" from-port="1" to-layer="2008" to-port="1" />
+ <edge from-layer="2008" from-port="2" to-layer="2010" to-port="0" />
+ <edge from-layer="2009" from-port="0" to-layer="2010" to-port="1" />
+ <edge from-layer="2010" from-port="2" to-layer="2012" to-port="0" />
+ <edge from-layer="2011" from-port="0" to-layer="2012" to-port="1" />
+ <edge from-layer="2012" from-port="2" to-layer="2017" to-port="0" />
+ <edge from-layer="2012" from-port="2" to-layer="2013" to-port="0" />
+ <edge from-layer="2012" from-port="2" to-layer="2023" to-port="0" />
+ <edge from-layer="2013" from-port="2" to-layer="2026" to-port="0" />
+ <edge from-layer="2014" from-port="0" to-layer="2017" to-port="1" />
+ <edge from-layer="2015" from-port="0" to-layer="2017" to-port="2" />
+ <edge from-layer="2016" from-port="0" to-layer="2017" to-port="3" />
+ <edge from-layer="2017" from-port="5" to-layer="2019" to-port="0" />
+ <edge from-layer="2018" from-port="0" to-layer="2019" to-port="1" />
+ <edge from-layer="2019" from-port="2" to-layer="2024" to-port="0" />
+ <edge from-layer="2020" from-port="0" to-layer="2023" to-port="1" />
+ <edge from-layer="2021" from-port="0" to-layer="2023" to-port="2" />
+ <edge from-layer="2022" from-port="0" to-layer="2023" to-port="3" />
+ <edge from-layer="2023" from-port="5" to-layer="2024" to-port="1" />
+ <edge from-layer="2024" from-port="2" to-layer="2025" to-port="0" />
+ <edge from-layer="2025" from-port="2" to-layer="2026" to-port="1" />
+ <edge from-layer="2026" from-port="2" to-layer="2087" to-port="0" />
+ <edge from-layer="2027" from-port="1" to-layer="2029" to-port="0" />
+ <edge from-layer="2028" from-port="0" to-layer="2029" to-port="2" />
+ <edge from-layer="2029" from-port="3" to-layer="2059" to-port="0" />
+ <edge from-layer="2030" from-port="0" to-layer="2031" to-port="0" />
+ <edge from-layer="2031" from-port="1" to-layer="2034" to-port="0" />
+ <edge from-layer="2032" from-port="0" to-layer="2033" to-port="0" />
+ <edge from-layer="2033" from-port="1" to-layer="2034" to-port="1" />
+ <edge from-layer="2034" from-port="2" to-layer="2036" to-port="0" />
+ <edge from-layer="2035" from-port="0" to-layer="2036" to-port="1" />
+ <edge from-layer="2036" from-port="2" to-layer="2038" to-port="0" />
+ <edge from-layer="2037" from-port="0" to-layer="2038" to-port="1" />
+ <edge from-layer="2038" from-port="2" to-layer="2039" to-port="0" />
+ <edge from-layer="2039" from-port="1" to-layer="2040" to-port="1" />
+ <edge from-layer="2040" from-port="2" to-layer="2042" to-port="0" />
+ <edge from-layer="2041" from-port="0" to-layer="2042" to-port="1" />
+ <edge from-layer="2042" from-port="2" to-layer="2044" to-port="0" />
+ <edge from-layer="2043" from-port="0" to-layer="2044" to-port="1" />
+ <edge from-layer="2044" from-port="2" to-layer="2049" to-port="0" />
+ <edge from-layer="2044" from-port="2" to-layer="2055" to-port="0" />
+ <edge from-layer="2044" from-port="2" to-layer="2045" to-port="0" />
+ <edge from-layer="2045" from-port="2" to-layer="2058" to-port="0" />
+ <edge from-layer="2046" from-port="0" to-layer="2049" to-port="1" />
+ <edge from-layer="2047" from-port="0" to-layer="2049" to-port="2" />
+ <edge from-layer="2048" from-port="0" to-layer="2049" to-port="3" />
+ <edge from-layer="2049" from-port="5" to-layer="2051" to-port="0" />
+ <edge from-layer="2050" from-port="0" to-layer="2051" to-port="1" />
+ <edge from-layer="2051" from-port="2" to-layer="2056" to-port="0" />
+ <edge from-layer="2052" from-port="0" to-layer="2055" to-port="1" />
+ <edge from-layer="2053" from-port="0" to-layer="2055" to-port="2" />
+ <edge from-layer="2054" from-port="0" to-layer="2055" to-port="3" />
+ <edge from-layer="2055" from-port="5" to-layer="2056" to-port="1" />
+ <edge from-layer="2056" from-port="2" to-layer="2057" to-port="0" />
+ <edge from-layer="2057" from-port="2" to-layer="2058" to-port="1" />
+ <edge from-layer="2058" from-port="2" to-layer="2059" to-port="1" />
+ <edge from-layer="2059" from-port="2" to-layer="2060" to-port="0" />
+ <edge from-layer="2059" from-port="2" to-layer="2715" to-port="0" />
+ <edge from-layer="2060" from-port="2" to-layer="2061" to-port="0" />
+ <edge from-layer="2061" from-port="2" to-layer="2063" to-port="0" />
+ <edge from-layer="2062" from-port="0" to-layer="2063" to-port="1" />
+ <edge from-layer="2063" from-port="2" to-layer="2087" to-port="1" />
+ <edge from-layer="2064" from-port="1" to-layer="2066" to-port="0" />
+ <edge from-layer="2065" from-port="0" to-layer="2066" to-port="2" />
+ <edge from-layer="2066" from-port="3" to-layer="2082" to-port="0" />
+ <edge from-layer="2067" from-port="0" to-layer="2068" to-port="0" />
+ <edge from-layer="2068" from-port="1" to-layer="2071" to-port="0" />
+ <edge from-layer="2069" from-port="0" to-layer="2070" to-port="0" />
+ <edge from-layer="2070" from-port="1" to-layer="2071" to-port="1" />
+ <edge from-layer="2071" from-port="2" to-layer="2073" to-port="0" />
+ <edge from-layer="2072" from-port="0" to-layer="2073" to-port="1" />
+ <edge from-layer="2073" from-port="2" to-layer="2075" to-port="0" />
+ <edge from-layer="2074" from-port="0" to-layer="2075" to-port="1" />
+ <edge from-layer="2075" from-port="2" to-layer="2076" to-port="0" />
+ <edge from-layer="2076" from-port="1" to-layer="2077" to-port="1" />
+ <edge from-layer="2077" from-port="2" to-layer="2079" to-port="0" />
+ <edge from-layer="2078" from-port="0" to-layer="2079" to-port="1" />
+ <edge from-layer="2079" from-port="2" to-layer="2081" to-port="0" />
+ <edge from-layer="2080" from-port="0" to-layer="2081" to-port="1" />
+ <edge from-layer="2081" from-port="2" to-layer="2082" to-port="1" />
+ <edge from-layer="2082" from-port="2" to-layer="2083" to-port="0" />
+ <edge from-layer="2082" from-port="2" to-layer="2714" to-port="0" />
+ <edge from-layer="2083" from-port="2" to-layer="2084" to-port="0" />
+ <edge from-layer="2084" from-port="2" to-layer="2086" to-port="0" />
+ <edge from-layer="2085" from-port="0" to-layer="2086" to-port="1" />
+ <edge from-layer="2086" from-port="2" to-layer="2087" to-port="2" />
+ <edge from-layer="2087" from-port="4" to-layer="2089" to-port="0" />
+ <edge from-layer="2088" from-port="0" to-layer="2089" to-port="1" />
+ <edge from-layer="2089" from-port="2" to-layer="2090" to-port="0" />
+ <edge from-layer="2090" from-port="2" to-layer="2101" to-port="0" />
+ <edge from-layer="2091" from-port="0" to-layer="2092" to-port="0" />
+ <edge from-layer="2092" from-port="1" to-layer="2095" to-port="0" />
+ <edge from-layer="2093" from-port="0" to-layer="2094" to-port="0" />
+ <edge from-layer="2094" from-port="1" to-layer="2095" to-port="1" />
+ <edge from-layer="2095" from-port="2" to-layer="2097" to-port="0" />
+ <edge from-layer="2096" from-port="0" to-layer="2097" to-port="1" />
+ <edge from-layer="2097" from-port="2" to-layer="2099" to-port="0" />
+ <edge from-layer="2098" from-port="0" to-layer="2099" to-port="1" />
+ <edge from-layer="2099" from-port="2" to-layer="2100" to-port="0" />
+ <edge from-layer="2100" from-port="1" to-layer="2101" to-port="1" />
+ <edge from-layer="2101" from-port="2" to-layer="2102" to-port="1" />
+ <edge from-layer="2102" from-port="2" to-layer="2113" to-port="0" />
+ <edge from-layer="2102" from-port="2" to-layer="2150" to-port="0" />
+ <edge from-layer="2102" from-port="2" to-layer="2106" to-port="0" />
+ <edge from-layer="2103" from-port="0" to-layer="2114" to-port="0" />
+ <edge from-layer="2104" from-port="0" to-layer="2112" to-port="0" />
+ <edge from-layer="2105" from-port="0" to-layer="2106" to-port="1" />
+ <edge from-layer="2106" from-port="2" to-layer="2108" to-port="0" />
+ <edge from-layer="2107" from-port="0" to-layer="2108" to-port="1" />
+ <edge from-layer="2108" from-port="2" to-layer="2110" to-port="0" />
+ <edge from-layer="2109" from-port="0" to-layer="2110" to-port="1" />
+ <edge from-layer="2110" from-port="2" to-layer="2111" to-port="0" />
+ <edge from-layer="2111" from-port="1" to-layer="2112" to-port="1" />
+ <edge from-layer="2112" from-port="2" to-layer="2113" to-port="1" />
+ <edge from-layer="2113" from-port="2" to-layer="2114" to-port="1" />
+ <edge from-layer="2114" from-port="2" to-layer="2137" to-port="0" />
+ <edge from-layer="2114" from-port="2" to-layer="2125" to-port="0" />
+ <edge from-layer="2115" from-port="0" to-layer="2116" to-port="0" />
+ <edge from-layer="2116" from-port="1" to-layer="2119" to-port="0" />
+ <edge from-layer="2117" from-port="0" to-layer="2118" to-port="0" />
+ <edge from-layer="2118" from-port="1" to-layer="2119" to-port="1" />
+ <edge from-layer="2119" from-port="2" to-layer="2121" to-port="0" />
+ <edge from-layer="2120" from-port="0" to-layer="2121" to-port="1" />
+ <edge from-layer="2121" from-port="2" to-layer="2123" to-port="0" />
+ <edge from-layer="2122" from-port="0" to-layer="2123" to-port="1" />
+ <edge from-layer="2123" from-port="2" to-layer="2124" to-port="0" />
+ <edge from-layer="2124" from-port="1" to-layer="2125" to-port="1" />
+ <edge from-layer="2125" from-port="2" to-layer="2126" to-port="0" />
+ <edge from-layer="2126" from-port="1" to-layer="2138" to-port="0" />
+ <edge from-layer="2127" from-port="0" to-layer="2128" to-port="0" />
+ <edge from-layer="2128" from-port="1" to-layer="2131" to-port="0" />
+ <edge from-layer="2129" from-port="0" to-layer="2130" to-port="0" />
+ <edge from-layer="2130" from-port="1" to-layer="2131" to-port="1" />
+ <edge from-layer="2131" from-port="2" to-layer="2133" to-port="0" />
+ <edge from-layer="2132" from-port="0" to-layer="2133" to-port="1" />
+ <edge from-layer="2133" from-port="2" to-layer="2135" to-port="0" />
+ <edge from-layer="2134" from-port="0" to-layer="2135" to-port="1" />
+ <edge from-layer="2135" from-port="2" to-layer="2136" to-port="0" />
+ <edge from-layer="2136" from-port="1" to-layer="2137" to-port="1" />
+ <edge from-layer="2137" from-port="2" to-layer="2138" to-port="1" />
+ <edge from-layer="2138" from-port="2" to-layer="2149" to-port="0" />
+ <edge from-layer="2139" from-port="0" to-layer="2140" to-port="0" />
+ <edge from-layer="2140" from-port="1" to-layer="2143" to-port="0" />
+ <edge from-layer="2141" from-port="0" to-layer="2142" to-port="0" />
+ <edge from-layer="2142" from-port="1" to-layer="2143" to-port="1" />
+ <edge from-layer="2143" from-port="2" to-layer="2145" to-port="0" />
+ <edge from-layer="2144" from-port="0" to-layer="2145" to-port="1" />
+ <edge from-layer="2145" from-port="2" to-layer="2147" to-port="0" />
+ <edge from-layer="2146" from-port="0" to-layer="2147" to-port="1" />
+ <edge from-layer="2147" from-port="2" to-layer="2148" to-port="0" />
+ <edge from-layer="2148" from-port="1" to-layer="2149" to-port="1" />
+ <edge from-layer="2149" from-port="2" to-layer="2150" to-port="1" />
+ <edge from-layer="2150" from-port="2" to-layer="2267" to-port="0" />
+ <edge from-layer="2150" from-port="2" to-layer="2154" to-port="0" />
+ <edge from-layer="2150" from-port="2" to-layer="2161" to-port="0" />
+ <edge from-layer="2151" from-port="0" to-layer="2162" to-port="0" />
+ <edge from-layer="2152" from-port="0" to-layer="2160" to-port="0" />
+ <edge from-layer="2153" from-port="0" to-layer="2154" to-port="1" />
+ <edge from-layer="2154" from-port="2" to-layer="2156" to-port="0" />
+ <edge from-layer="2155" from-port="0" to-layer="2156" to-port="1" />
+ <edge from-layer="2156" from-port="2" to-layer="2158" to-port="0" />
+ <edge from-layer="2157" from-port="0" to-layer="2158" to-port="1" />
+ <edge from-layer="2158" from-port="2" to-layer="2159" to-port="0" />
+ <edge from-layer="2159" from-port="1" to-layer="2160" to-port="1" />
+ <edge from-layer="2160" from-port="2" to-layer="2161" to-port="1" />
+ <edge from-layer="2161" from-port="2" to-layer="2162" to-port="1" />
+ <edge from-layer="2162" from-port="2" to-layer="2173" to-port="0" />
+ <edge from-layer="2162" from-port="2" to-layer="2205" to-port="0" />
+ <edge from-layer="2162" from-port="2" to-layer="2242" to-port="0" />
+ <edge from-layer="2163" from-port="0" to-layer="2164" to-port="0" />
+ <edge from-layer="2164" from-port="1" to-layer="2167" to-port="0" />
+ <edge from-layer="2165" from-port="0" to-layer="2166" to-port="0" />
+ <edge from-layer="2166" from-port="1" to-layer="2167" to-port="1" />
+ <edge from-layer="2167" from-port="2" to-layer="2169" to-port="0" />
+ <edge from-layer="2168" from-port="0" to-layer="2169" to-port="1" />
+ <edge from-layer="2169" from-port="2" to-layer="2171" to-port="0" />
+ <edge from-layer="2170" from-port="0" to-layer="2171" to-port="1" />
+ <edge from-layer="2171" from-port="2" to-layer="2172" to-port="0" />
+ <edge from-layer="2172" from-port="1" to-layer="2173" to-port="1" />
+ <edge from-layer="2173" from-port="2" to-layer="2175" to-port="0" />
+ <edge from-layer="2174" from-port="0" to-layer="2175" to-port="1" />
+ <edge from-layer="2175" from-port="2" to-layer="2177" to-port="0" />
+ <edge from-layer="2176" from-port="0" to-layer="2177" to-port="1" />
+ <edge from-layer="2177" from-port="2" to-layer="2178" to-port="0" />
+ <edge from-layer="2177" from-port="2" to-layer="2182" to-port="0" />
+ <edge from-layer="2177" from-port="2" to-layer="2188" to-port="0" />
+ <edge from-layer="2178" from-port="2" to-layer="2191" to-port="0" />
+ <edge from-layer="2179" from-port="0" to-layer="2182" to-port="1" />
+ <edge from-layer="2180" from-port="0" to-layer="2182" to-port="2" />
+ <edge from-layer="2181" from-port="0" to-layer="2182" to-port="3" />
+ <edge from-layer="2182" from-port="5" to-layer="2184" to-port="0" />
+ <edge from-layer="2183" from-port="0" to-layer="2184" to-port="1" />
+ <edge from-layer="2184" from-port="2" to-layer="2189" to-port="0" />
+ <edge from-layer="2185" from-port="0" to-layer="2188" to-port="1" />
+ <edge from-layer="2186" from-port="0" to-layer="2188" to-port="2" />
+ <edge from-layer="2187" from-port="0" to-layer="2188" to-port="3" />
+ <edge from-layer="2188" from-port="5" to-layer="2189" to-port="1" />
+ <edge from-layer="2189" from-port="2" to-layer="2190" to-port="0" />
+ <edge from-layer="2190" from-port="2" to-layer="2191" to-port="1" />
+ <edge from-layer="2191" from-port="2" to-layer="2252" to-port="0" />
+ <edge from-layer="2192" from-port="1" to-layer="2194" to-port="0" />
+ <edge from-layer="2193" from-port="0" to-layer="2194" to-port="2" />
+ <edge from-layer="2194" from-port="3" to-layer="2224" to-port="0" />
+ <edge from-layer="2195" from-port="0" to-layer="2196" to-port="0" />
+ <edge from-layer="2196" from-port="1" to-layer="2199" to-port="0" />
+ <edge from-layer="2197" from-port="0" to-layer="2198" to-port="0" />
+ <edge from-layer="2198" from-port="1" to-layer="2199" to-port="1" />
+ <edge from-layer="2199" from-port="2" to-layer="2201" to-port="0" />
+ <edge from-layer="2200" from-port="0" to-layer="2201" to-port="1" />
+ <edge from-layer="2201" from-port="2" to-layer="2203" to-port="0" />
+ <edge from-layer="2202" from-port="0" to-layer="2203" to-port="1" />
+ <edge from-layer="2203" from-port="2" to-layer="2204" to-port="0" />
+ <edge from-layer="2204" from-port="1" to-layer="2205" to-port="1" />
+ <edge from-layer="2205" from-port="2" to-layer="2207" to-port="0" />
+ <edge from-layer="2206" from-port="0" to-layer="2207" to-port="1" />
+ <edge from-layer="2207" from-port="2" to-layer="2209" to-port="0" />
+ <edge from-layer="2208" from-port="0" to-layer="2209" to-port="1" />
+ <edge from-layer="2209" from-port="2" to-layer="2210" to-port="0" />
+ <edge from-layer="2209" from-port="2" to-layer="2214" to-port="0" />
+ <edge from-layer="2209" from-port="2" to-layer="2220" to-port="0" />
+ <edge from-layer="2210" from-port="2" to-layer="2223" to-port="0" />
+ <edge from-layer="2211" from-port="0" to-layer="2214" to-port="1" />
+ <edge from-layer="2212" from-port="0" to-layer="2214" to-port="2" />
+ <edge from-layer="2213" from-port="0" to-layer="2214" to-port="3" />
+ <edge from-layer="2214" from-port="5" to-layer="2216" to-port="0" />
+ <edge from-layer="2215" from-port="0" to-layer="2216" to-port="1" />
+ <edge from-layer="2216" from-port="2" to-layer="2221" to-port="0" />
+ <edge from-layer="2217" from-port="0" to-layer="2220" to-port="1" />
+ <edge from-layer="2218" from-port="0" to-layer="2220" to-port="2" />
+ <edge from-layer="2219" from-port="0" to-layer="2220" to-port="3" />
+ <edge from-layer="2220" from-port="5" to-layer="2221" to-port="1" />
+ <edge from-layer="2221" from-port="2" to-layer="2222" to-port="0" />
+ <edge from-layer="2222" from-port="2" to-layer="2223" to-port="1" />
+ <edge from-layer="2223" from-port="2" to-layer="2224" to-port="1" />
+ <edge from-layer="2224" from-port="2" to-layer="2713" to-port="0" />
+ <edge from-layer="2224" from-port="2" to-layer="2225" to-port="0" />
+ <edge from-layer="2225" from-port="2" to-layer="2226" to-port="0" />
+ <edge from-layer="2226" from-port="2" to-layer="2228" to-port="0" />
+ <edge from-layer="2227" from-port="0" to-layer="2228" to-port="1" />
+ <edge from-layer="2228" from-port="2" to-layer="2252" to-port="1" />
+ <edge from-layer="2229" from-port="1" to-layer="2231" to-port="0" />
+ <edge from-layer="2230" from-port="0" to-layer="2231" to-port="2" />
+ <edge from-layer="2231" from-port="3" to-layer="2247" to-port="0" />
+ <edge from-layer="2232" from-port="0" to-layer="2233" to-port="0" />
+ <edge from-layer="2233" from-port="1" to-layer="2236" to-port="0" />
+ <edge from-layer="2234" from-port="0" to-layer="2235" to-port="0" />
+ <edge from-layer="2235" from-port="1" to-layer="2236" to-port="1" />
+ <edge from-layer="2236" from-port="2" to-layer="2238" to-port="0" />
+ <edge from-layer="2237" from-port="0" to-layer="2238" to-port="1" />
+ <edge from-layer="2238" from-port="2" to-layer="2240" to-port="0" />
+ <edge from-layer="2239" from-port="0" to-layer="2240" to-port="1" />
+ <edge from-layer="2240" from-port="2" to-layer="2241" to-port="0" />
+ <edge from-layer="2241" from-port="1" to-layer="2242" to-port="1" />
+ <edge from-layer="2242" from-port="2" to-layer="2244" to-port="0" />
+ <edge from-layer="2243" from-port="0" to-layer="2244" to-port="1" />
+ <edge from-layer="2244" from-port="2" to-layer="2246" to-port="0" />
+ <edge from-layer="2245" from-port="0" to-layer="2246" to-port="1" />
+ <edge from-layer="2246" from-port="2" to-layer="2247" to-port="1" />
+ <edge from-layer="2247" from-port="2" to-layer="2712" to-port="0" />
+ <edge from-layer="2247" from-port="2" to-layer="2248" to-port="0" />
+ <edge from-layer="2248" from-port="2" to-layer="2249" to-port="0" />
+ <edge from-layer="2249" from-port="2" to-layer="2251" to-port="0" />
+ <edge from-layer="2250" from-port="0" to-layer="2251" to-port="1" />
+ <edge from-layer="2251" from-port="2" to-layer="2252" to-port="2" />
+ <edge from-layer="2252" from-port="4" to-layer="2254" to-port="0" />
+ <edge from-layer="2253" from-port="0" to-layer="2254" to-port="1" />
+ <edge from-layer="2254" from-port="2" to-layer="2255" to-port="0" />
+ <edge from-layer="2255" from-port="2" to-layer="2266" to-port="0" />
+ <edge from-layer="2256" from-port="0" to-layer="2257" to-port="0" />
+ <edge from-layer="2257" from-port="1" to-layer="2260" to-port="0" />
+ <edge from-layer="2258" from-port="0" to-layer="2259" to-port="0" />
+ <edge from-layer="2259" from-port="1" to-layer="2260" to-port="1" />
+ <edge from-layer="2260" from-port="2" to-layer="2262" to-port="0" />
+ <edge from-layer="2261" from-port="0" to-layer="2262" to-port="1" />
+ <edge from-layer="2262" from-port="2" to-layer="2264" to-port="0" />
+ <edge from-layer="2263" from-port="0" to-layer="2264" to-port="1" />
+ <edge from-layer="2264" from-port="2" to-layer="2265" to-port="0" />
+ <edge from-layer="2265" from-port="1" to-layer="2266" to-port="1" />
+ <edge from-layer="2266" from-port="2" to-layer="2267" to-port="1" />
+ <edge from-layer="2267" from-port="2" to-layer="2315" to-port="0" />
+ <edge from-layer="2267" from-port="2" to-layer="2278" to-port="0" />
+ <edge from-layer="2267" from-port="2" to-layer="2271" to-port="0" />
+ <edge from-layer="2268" from-port="0" to-layer="2279" to-port="0" />
+ <edge from-layer="2269" from-port="0" to-layer="2277" to-port="0" />
+ <edge from-layer="2270" from-port="0" to-layer="2271" to-port="1" />
+ <edge from-layer="2271" from-port="2" to-layer="2273" to-port="0" />
+ <edge from-layer="2272" from-port="0" to-layer="2273" to-port="1" />
+ <edge from-layer="2273" from-port="2" to-layer="2275" to-port="0" />
+ <edge from-layer="2274" from-port="0" to-layer="2275" to-port="1" />
+ <edge from-layer="2275" from-port="2" to-layer="2276" to-port="0" />
+ <edge from-layer="2276" from-port="1" to-layer="2277" to-port="1" />
+ <edge from-layer="2277" from-port="2" to-layer="2278" to-port="1" />
+ <edge from-layer="2278" from-port="2" to-layer="2279" to-port="1" />
+ <edge from-layer="2279" from-port="2" to-layer="2290" to-port="0" />
+ <edge from-layer="2279" from-port="2" to-layer="2302" to-port="0" />
+ <edge from-layer="2280" from-port="0" to-layer="2281" to-port="0" />
+ <edge from-layer="2281" from-port="1" to-layer="2284" to-port="0" />
+ <edge from-layer="2282" from-port="0" to-layer="2283" to-port="0" />
+ <edge from-layer="2283" from-port="1" to-layer="2284" to-port="1" />
+ <edge from-layer="2284" from-port="2" to-layer="2286" to-port="0" />
+ <edge from-layer="2285" from-port="0" to-layer="2286" to-port="1" />
+ <edge from-layer="2286" from-port="2" to-layer="2288" to-port="0" />
+ <edge from-layer="2287" from-port="0" to-layer="2288" to-port="1" />
+ <edge from-layer="2288" from-port="2" to-layer="2289" to-port="0" />
+ <edge from-layer="2289" from-port="1" to-layer="2290" to-port="1" />
+ <edge from-layer="2290" from-port="2" to-layer="2291" to-port="0" />
+ <edge from-layer="2291" from-port="1" to-layer="2303" to-port="0" />
+ <edge from-layer="2292" from-port="0" to-layer="2293" to-port="0" />
+ <edge from-layer="2293" from-port="1" to-layer="2296" to-port="0" />
+ <edge from-layer="2294" from-port="0" to-layer="2295" to-port="0" />
+ <edge from-layer="2295" from-port="1" to-layer="2296" to-port="1" />
+ <edge from-layer="2296" from-port="2" to-layer="2298" to-port="0" />
+ <edge from-layer="2297" from-port="0" to-layer="2298" to-port="1" />
+ <edge from-layer="2298" from-port="2" to-layer="2300" to-port="0" />
+ <edge from-layer="2299" from-port="0" to-layer="2300" to-port="1" />
+ <edge from-layer="2300" from-port="2" to-layer="2301" to-port="0" />
+ <edge from-layer="2301" from-port="1" to-layer="2302" to-port="1" />
+ <edge from-layer="2302" from-port="2" to-layer="2303" to-port="1" />
+ <edge from-layer="2303" from-port="2" to-layer="2314" to-port="0" />
+ <edge from-layer="2304" from-port="0" to-layer="2305" to-port="0" />
+ <edge from-layer="2305" from-port="1" to-layer="2308" to-port="0" />
+ <edge from-layer="2306" from-port="0" to-layer="2307" to-port="0" />
+ <edge from-layer="2307" from-port="1" to-layer="2308" to-port="1" />
+ <edge from-layer="2308" from-port="2" to-layer="2310" to-port="0" />
+ <edge from-layer="2309" from-port="0" to-layer="2310" to-port="1" />
+ <edge from-layer="2310" from-port="2" to-layer="2312" to-port="0" />
+ <edge from-layer="2311" from-port="0" to-layer="2312" to-port="1" />
+ <edge from-layer="2312" from-port="2" to-layer="2313" to-port="0" />
+ <edge from-layer="2313" from-port="1" to-layer="2314" to-port="1" />
+ <edge from-layer="2314" from-port="2" to-layer="2315" to-port="1" />
+ <edge from-layer="2315" from-port="2" to-layer="2326" to-port="0" />
+ <edge from-layer="2315" from-port="2" to-layer="2432" to-port="0" />
+ <edge from-layer="2315" from-port="2" to-layer="2319" to-port="0" />
+ <edge from-layer="2316" from-port="0" to-layer="2327" to-port="0" />
+ <edge from-layer="2317" from-port="0" to-layer="2325" to-port="0" />
+ <edge from-layer="2318" from-port="0" to-layer="2319" to-port="1" />
+ <edge from-layer="2319" from-port="2" to-layer="2321" to-port="0" />
+ <edge from-layer="2320" from-port="0" to-layer="2321" to-port="1" />
+ <edge from-layer="2321" from-port="2" to-layer="2323" to-port="0" />
+ <edge from-layer="2322" from-port="0" to-layer="2323" to-port="1" />
+ <edge from-layer="2323" from-port="2" to-layer="2324" to-port="0" />
+ <edge from-layer="2324" from-port="1" to-layer="2325" to-port="1" />
+ <edge from-layer="2325" from-port="2" to-layer="2326" to-port="1" />
+ <edge from-layer="2326" from-port="2" to-layer="2327" to-port="1" />
+ <edge from-layer="2327" from-port="2" to-layer="2407" to-port="0" />
+ <edge from-layer="2327" from-port="2" to-layer="2338" to-port="0" />
+ <edge from-layer="2327" from-port="2" to-layer="2370" to-port="0" />
+ <edge from-layer="2328" from-port="0" to-layer="2329" to-port="0" />
+ <edge from-layer="2329" from-port="1" to-layer="2332" to-port="0" />
+ <edge from-layer="2330" from-port="0" to-layer="2331" to-port="0" />
+ <edge from-layer="2331" from-port="1" to-layer="2332" to-port="1" />
+ <edge from-layer="2332" from-port="2" to-layer="2334" to-port="0" />
+ <edge from-layer="2333" from-port="0" to-layer="2334" to-port="1" />
+ <edge from-layer="2334" from-port="2" to-layer="2336" to-port="0" />
+ <edge from-layer="2335" from-port="0" to-layer="2336" to-port="1" />
+ <edge from-layer="2336" from-port="2" to-layer="2337" to-port="0" />
+ <edge from-layer="2337" from-port="1" to-layer="2338" to-port="1" />
+ <edge from-layer="2338" from-port="2" to-layer="2340" to-port="0" />
+ <edge from-layer="2339" from-port="0" to-layer="2340" to-port="1" />
+ <edge from-layer="2340" from-port="2" to-layer="2342" to-port="0" />
+ <edge from-layer="2341" from-port="0" to-layer="2342" to-port="1" />
+ <edge from-layer="2342" from-port="2" to-layer="2343" to-port="0" />
+ <edge from-layer="2342" from-port="2" to-layer="2353" to-port="0" />
+ <edge from-layer="2342" from-port="2" to-layer="2347" to-port="0" />
+ <edge from-layer="2343" from-port="2" to-layer="2356" to-port="0" />
+ <edge from-layer="2344" from-port="0" to-layer="2347" to-port="1" />
+ <edge from-layer="2345" from-port="0" to-layer="2347" to-port="2" />
+ <edge from-layer="2346" from-port="0" to-layer="2347" to-port="3" />
+ <edge from-layer="2347" from-port="5" to-layer="2349" to-port="0" />
+ <edge from-layer="2348" from-port="0" to-layer="2349" to-port="1" />
+ <edge from-layer="2349" from-port="2" to-layer="2354" to-port="0" />
+ <edge from-layer="2350" from-port="0" to-layer="2353" to-port="1" />
+ <edge from-layer="2351" from-port="0" to-layer="2353" to-port="2" />
+ <edge from-layer="2352" from-port="0" to-layer="2353" to-port="3" />
+ <edge from-layer="2353" from-port="5" to-layer="2354" to-port="1" />
+ <edge from-layer="2354" from-port="2" to-layer="2355" to-port="0" />
+ <edge from-layer="2355" from-port="2" to-layer="2356" to-port="1" />
+ <edge from-layer="2356" from-port="2" to-layer="2417" to-port="0" />
+ <edge from-layer="2357" from-port="1" to-layer="2359" to-port="0" />
+ <edge from-layer="2358" from-port="0" to-layer="2359" to-port="2" />
+ <edge from-layer="2359" from-port="3" to-layer="2389" to-port="0" />
+ <edge from-layer="2360" from-port="0" to-layer="2361" to-port="0" />
+ <edge from-layer="2361" from-port="1" to-layer="2364" to-port="0" />
+ <edge from-layer="2362" from-port="0" to-layer="2363" to-port="0" />
+ <edge from-layer="2363" from-port="1" to-layer="2364" to-port="1" />
+ <edge from-layer="2364" from-port="2" to-layer="2366" to-port="0" />
+ <edge from-layer="2365" from-port="0" to-layer="2366" to-port="1" />
+ <edge from-layer="2366" from-port="2" to-layer="2368" to-port="0" />
+ <edge from-layer="2367" from-port="0" to-layer="2368" to-port="1" />
+ <edge from-layer="2368" from-port="2" to-layer="2369" to-port="0" />
+ <edge from-layer="2369" from-port="1" to-layer="2370" to-port="1" />
+ <edge from-layer="2370" from-port="2" to-layer="2372" to-port="0" />
+ <edge from-layer="2371" from-port="0" to-layer="2372" to-port="1" />
+ <edge from-layer="2372" from-port="2" to-layer="2374" to-port="0" />
+ <edge from-layer="2373" from-port="0" to-layer="2374" to-port="1" />
+ <edge from-layer="2374" from-port="2" to-layer="2375" to-port="0" />
+ <edge from-layer="2374" from-port="2" to-layer="2379" to-port="0" />
+ <edge from-layer="2374" from-port="2" to-layer="2385" to-port="0" />
+ <edge from-layer="2375" from-port="2" to-layer="2388" to-port="0" />
+ <edge from-layer="2376" from-port="0" to-layer="2379" to-port="1" />
+ <edge from-layer="2377" from-port="0" to-layer="2379" to-port="2" />
+ <edge from-layer="2378" from-port="0" to-layer="2379" to-port="3" />
+ <edge from-layer="2379" from-port="5" to-layer="2381" to-port="0" />
+ <edge from-layer="2380" from-port="0" to-layer="2381" to-port="1" />
+ <edge from-layer="2381" from-port="2" to-layer="2386" to-port="0" />
+ <edge from-layer="2382" from-port="0" to-layer="2385" to-port="1" />
+ <edge from-layer="2383" from-port="0" to-layer="2385" to-port="2" />
+ <edge from-layer="2384" from-port="0" to-layer="2385" to-port="3" />
+ <edge from-layer="2385" from-port="5" to-layer="2386" to-port="1" />
+ <edge from-layer="2386" from-port="2" to-layer="2387" to-port="0" />
+ <edge from-layer="2387" from-port="2" to-layer="2388" to-port="1" />
+ <edge from-layer="2388" from-port="2" to-layer="2389" to-port="1" />
+ <edge from-layer="2389" from-port="2" to-layer="2711" to-port="0" />
+ <edge from-layer="2389" from-port="2" to-layer="2390" to-port="0" />
+ <edge from-layer="2390" from-port="2" to-layer="2391" to-port="0" />
+ <edge from-layer="2391" from-port="2" to-layer="2393" to-port="0" />
+ <edge from-layer="2392" from-port="0" to-layer="2393" to-port="1" />
+ <edge from-layer="2393" from-port="2" to-layer="2417" to-port="1" />
+ <edge from-layer="2394" from-port="1" to-layer="2396" to-port="0" />
+ <edge from-layer="2395" from-port="0" to-layer="2396" to-port="2" />
+ <edge from-layer="2396" from-port="3" to-layer="2412" to-port="0" />
+ <edge from-layer="2397" from-port="0" to-layer="2398" to-port="0" />
+ <edge from-layer="2398" from-port="1" to-layer="2401" to-port="0" />
+ <edge from-layer="2399" from-port="0" to-layer="2400" to-port="0" />
+ <edge from-layer="2400" from-port="1" to-layer="2401" to-port="1" />
+ <edge from-layer="2401" from-port="2" to-layer="2403" to-port="0" />
+ <edge from-layer="2402" from-port="0" to-layer="2403" to-port="1" />
+ <edge from-layer="2403" from-port="2" to-layer="2405" to-port="0" />
+ <edge from-layer="2404" from-port="0" to-layer="2405" to-port="1" />
+ <edge from-layer="2405" from-port="2" to-layer="2406" to-port="0" />
+ <edge from-layer="2406" from-port="1" to-layer="2407" to-port="1" />
+ <edge from-layer="2407" from-port="2" to-layer="2409" to-port="0" />
+ <edge from-layer="2408" from-port="0" to-layer="2409" to-port="1" />
+ <edge from-layer="2409" from-port="2" to-layer="2411" to-port="0" />
+ <edge from-layer="2410" from-port="0" to-layer="2411" to-port="1" />
+ <edge from-layer="2411" from-port="2" to-layer="2412" to-port="1" />
+ <edge from-layer="2412" from-port="2" to-layer="2710" to-port="0" />
+ <edge from-layer="2412" from-port="2" to-layer="2413" to-port="0" />
+ <edge from-layer="2413" from-port="2" to-layer="2414" to-port="0" />
+ <edge from-layer="2414" from-port="2" to-layer="2416" to-port="0" />
+ <edge from-layer="2415" from-port="0" to-layer="2416" to-port="1" />
+ <edge from-layer="2416" from-port="2" to-layer="2417" to-port="2" />
+ <edge from-layer="2417" from-port="4" to-layer="2419" to-port="0" />
+ <edge from-layer="2418" from-port="0" to-layer="2419" to-port="1" />
+ <edge from-layer="2419" from-port="2" to-layer="2420" to-port="0" />
+ <edge from-layer="2420" from-port="2" to-layer="2431" to-port="0" />
+ <edge from-layer="2421" from-port="0" to-layer="2422" to-port="0" />
+ <edge from-layer="2422" from-port="1" to-layer="2425" to-port="0" />
+ <edge from-layer="2423" from-port="0" to-layer="2424" to-port="0" />
+ <edge from-layer="2424" from-port="1" to-layer="2425" to-port="1" />
+ <edge from-layer="2425" from-port="2" to-layer="2427" to-port="0" />
+ <edge from-layer="2426" from-port="0" to-layer="2427" to-port="1" />
+ <edge from-layer="2427" from-port="2" to-layer="2429" to-port="0" />
+ <edge from-layer="2428" from-port="0" to-layer="2429" to-port="1" />
+ <edge from-layer="2429" from-port="2" to-layer="2430" to-port="0" />
+ <edge from-layer="2430" from-port="1" to-layer="2431" to-port="1" />
+ <edge from-layer="2431" from-port="2" to-layer="2432" to-port="1" />
+ <edge from-layer="2432" from-port="2" to-layer="2436" to-port="0" />
+ <edge from-layer="2432" from-port="2" to-layer="2480" to-port="0" />
+ <edge from-layer="2432" from-port="2" to-layer="2443" to-port="0" />
+ <edge from-layer="2433" from-port="0" to-layer="2444" to-port="0" />
+ <edge from-layer="2434" from-port="0" to-layer="2442" to-port="0" />
+ <edge from-layer="2435" from-port="0" to-layer="2436" to-port="1" />
+ <edge from-layer="2436" from-port="2" to-layer="2438" to-port="0" />
+ <edge from-layer="2437" from-port="0" to-layer="2438" to-port="1" />
+ <edge from-layer="2438" from-port="2" to-layer="2440" to-port="0" />
+ <edge from-layer="2439" from-port="0" to-layer="2440" to-port="1" />
+ <edge from-layer="2440" from-port="2" to-layer="2441" to-port="0" />
+ <edge from-layer="2441" from-port="1" to-layer="2442" to-port="1" />
+ <edge from-layer="2442" from-port="2" to-layer="2443" to-port="1" />
+ <edge from-layer="2443" from-port="2" to-layer="2444" to-port="1" />
+ <edge from-layer="2444" from-port="2" to-layer="2467" to-port="0" />
+ <edge from-layer="2444" from-port="2" to-layer="2455" to-port="0" />
+ <edge from-layer="2445" from-port="0" to-layer="2446" to-port="0" />
+ <edge from-layer="2446" from-port="1" to-layer="2449" to-port="0" />
+ <edge from-layer="2447" from-port="0" to-layer="2448" to-port="0" />
+ <edge from-layer="2448" from-port="1" to-layer="2449" to-port="1" />
+ <edge from-layer="2449" from-port="2" to-layer="2451" to-port="0" />
+ <edge from-layer="2450" from-port="0" to-layer="2451" to-port="1" />
+ <edge from-layer="2451" from-port="2" to-layer="2453" to-port="0" />
+ <edge from-layer="2452" from-port="0" to-layer="2453" to-port="1" />
+ <edge from-layer="2453" from-port="2" to-layer="2454" to-port="0" />
+ <edge from-layer="2454" from-port="1" to-layer="2455" to-port="1" />
+ <edge from-layer="2455" from-port="2" to-layer="2456" to-port="0" />
+ <edge from-layer="2456" from-port="1" to-layer="2468" to-port="0" />
+ <edge from-layer="2457" from-port="0" to-layer="2458" to-port="0" />
+ <edge from-layer="2458" from-port="1" to-layer="2461" to-port="0" />
+ <edge from-layer="2459" from-port="0" to-layer="2460" to-port="0" />
+ <edge from-layer="2460" from-port="1" to-layer="2461" to-port="1" />
+ <edge from-layer="2461" from-port="2" to-layer="2463" to-port="0" />
+ <edge from-layer="2462" from-port="0" to-layer="2463" to-port="1" />
+ <edge from-layer="2463" from-port="2" to-layer="2465" to-port="0" />
+ <edge from-layer="2464" from-port="0" to-layer="2465" to-port="1" />
+ <edge from-layer="2465" from-port="2" to-layer="2466" to-port="0" />
+ <edge from-layer="2466" from-port="1" to-layer="2467" to-port="1" />
+ <edge from-layer="2467" from-port="2" to-layer="2468" to-port="1" />
+ <edge from-layer="2468" from-port="2" to-layer="2479" to-port="0" />
+ <edge from-layer="2469" from-port="0" to-layer="2470" to-port="0" />
+ <edge from-layer="2470" from-port="1" to-layer="2473" to-port="0" />
+ <edge from-layer="2471" from-port="0" to-layer="2472" to-port="0" />
+ <edge from-layer="2472" from-port="1" to-layer="2473" to-port="1" />
+ <edge from-layer="2473" from-port="2" to-layer="2475" to-port="0" />
+ <edge from-layer="2474" from-port="0" to-layer="2475" to-port="1" />
+ <edge from-layer="2475" from-port="2" to-layer="2477" to-port="0" />
+ <edge from-layer="2476" from-port="0" to-layer="2477" to-port="1" />
+ <edge from-layer="2477" from-port="2" to-layer="2478" to-port="0" />
+ <edge from-layer="2478" from-port="1" to-layer="2479" to-port="1" />
+ <edge from-layer="2479" from-port="2" to-layer="2480" to-port="1" />
+ <edge from-layer="2480" from-port="2" to-layer="2491" to-port="0" />
+ <edge from-layer="2480" from-port="2" to-layer="2597" to-port="0" />
+ <edge from-layer="2480" from-port="2" to-layer="2484" to-port="0" />
+ <edge from-layer="2481" from-port="0" to-layer="2492" to-port="0" />
+ <edge from-layer="2482" from-port="0" to-layer="2490" to-port="0" />
+ <edge from-layer="2483" from-port="0" to-layer="2484" to-port="1" />
+ <edge from-layer="2484" from-port="2" to-layer="2486" to-port="0" />
+ <edge from-layer="2485" from-port="0" to-layer="2486" to-port="1" />
+ <edge from-layer="2486" from-port="2" to-layer="2488" to-port="0" />
+ <edge from-layer="2487" from-port="0" to-layer="2488" to-port="1" />
+ <edge from-layer="2488" from-port="2" to-layer="2489" to-port="0" />
+ <edge from-layer="2489" from-port="1" to-layer="2490" to-port="1" />
+ <edge from-layer="2490" from-port="2" to-layer="2491" to-port="1" />
+ <edge from-layer="2491" from-port="2" to-layer="2492" to-port="1" />
+ <edge from-layer="2492" from-port="2" to-layer="2535" to-port="0" />
+ <edge from-layer="2492" from-port="2" to-layer="2572" to-port="0" />
+ <edge from-layer="2492" from-port="2" to-layer="2503" to-port="0" />
+ <edge from-layer="2493" from-port="0" to-layer="2494" to-port="0" />
+ <edge from-layer="2494" from-port="1" to-layer="2497" to-port="0" />
+ <edge from-layer="2495" from-port="0" to-layer="2496" to-port="0" />
+ <edge from-layer="2496" from-port="1" to-layer="2497" to-port="1" />
+ <edge from-layer="2497" from-port="2" to-layer="2499" to-port="0" />
+ <edge from-layer="2498" from-port="0" to-layer="2499" to-port="1" />
+ <edge from-layer="2499" from-port="2" to-layer="2501" to-port="0" />
+ <edge from-layer="2500" from-port="0" to-layer="2501" to-port="1" />
+ <edge from-layer="2501" from-port="2" to-layer="2502" to-port="0" />
+ <edge from-layer="2502" from-port="1" to-layer="2503" to-port="1" />
+ <edge from-layer="2503" from-port="2" to-layer="2505" to-port="0" />
+ <edge from-layer="2504" from-port="0" to-layer="2505" to-port="1" />
+ <edge from-layer="2505" from-port="2" to-layer="2507" to-port="0" />
+ <edge from-layer="2506" from-port="0" to-layer="2507" to-port="1" />
+ <edge from-layer="2507" from-port="2" to-layer="2512" to-port="0" />
+ <edge from-layer="2507" from-port="2" to-layer="2508" to-port="0" />
+ <edge from-layer="2507" from-port="2" to-layer="2518" to-port="0" />
+ <edge from-layer="2508" from-port="2" to-layer="2521" to-port="0" />
+ <edge from-layer="2509" from-port="0" to-layer="2512" to-port="1" />
+ <edge from-layer="2510" from-port="0" to-layer="2512" to-port="2" />
+ <edge from-layer="2511" from-port="0" to-layer="2512" to-port="3" />
+ <edge from-layer="2512" from-port="5" to-layer="2514" to-port="0" />
+ <edge from-layer="2513" from-port="0" to-layer="2514" to-port="1" />
+ <edge from-layer="2514" from-port="2" to-layer="2519" to-port="0" />
+ <edge from-layer="2515" from-port="0" to-layer="2518" to-port="1" />
+ <edge from-layer="2516" from-port="0" to-layer="2518" to-port="2" />
+ <edge from-layer="2517" from-port="0" to-layer="2518" to-port="3" />
+ <edge from-layer="2518" from-port="5" to-layer="2519" to-port="1" />
+ <edge from-layer="2519" from-port="2" to-layer="2520" to-port="0" />
+ <edge from-layer="2520" from-port="2" to-layer="2521" to-port="1" />
+ <edge from-layer="2521" from-port="2" to-layer="2582" to-port="0" />
+ <edge from-layer="2522" from-port="1" to-layer="2524" to-port="0" />
+ <edge from-layer="2523" from-port="0" to-layer="2524" to-port="2" />
+ <edge from-layer="2524" from-port="3" to-layer="2554" to-port="0" />
+ <edge from-layer="2525" from-port="0" to-layer="2526" to-port="0" />
+ <edge from-layer="2526" from-port="1" to-layer="2529" to-port="0" />
+ <edge from-layer="2527" from-port="0" to-layer="2528" to-port="0" />
+ <edge from-layer="2528" from-port="1" to-layer="2529" to-port="1" />
+ <edge from-layer="2529" from-port="2" to-layer="2531" to-port="0" />
+ <edge from-layer="2530" from-port="0" to-layer="2531" to-port="1" />
+ <edge from-layer="2531" from-port="2" to-layer="2533" to-port="0" />
+ <edge from-layer="2532" from-port="0" to-layer="2533" to-port="1" />
+ <edge from-layer="2533" from-port="2" to-layer="2534" to-port="0" />
+ <edge from-layer="2534" from-port="1" to-layer="2535" to-port="1" />
+ <edge from-layer="2535" from-port="2" to-layer="2537" to-port="0" />
+ <edge from-layer="2536" from-port="0" to-layer="2537" to-port="1" />
+ <edge from-layer="2537" from-port="2" to-layer="2539" to-port="0" />
+ <edge from-layer="2538" from-port="0" to-layer="2539" to-port="1" />
+ <edge from-layer="2539" from-port="2" to-layer="2550" to-port="0" />
+ <edge from-layer="2539" from-port="2" to-layer="2544" to-port="0" />
+ <edge from-layer="2539" from-port="2" to-layer="2540" to-port="0" />
+ <edge from-layer="2540" from-port="2" to-layer="2553" to-port="0" />
+ <edge from-layer="2541" from-port="0" to-layer="2544" to-port="1" />
+ <edge from-layer="2542" from-port="0" to-layer="2544" to-port="2" />
+ <edge from-layer="2543" from-port="0" to-layer="2544" to-port="3" />
+ <edge from-layer="2544" from-port="5" to-layer="2546" to-port="0" />
+ <edge from-layer="2545" from-port="0" to-layer="2546" to-port="1" />
+ <edge from-layer="2546" from-port="2" to-layer="2551" to-port="0" />
+ <edge from-layer="2547" from-port="0" to-layer="2550" to-port="1" />
+ <edge from-layer="2548" from-port="0" to-layer="2550" to-port="2" />
+ <edge from-layer="2549" from-port="0" to-layer="2550" to-port="3" />
+ <edge from-layer="2550" from-port="5" to-layer="2551" to-port="1" />
+ <edge from-layer="2551" from-port="2" to-layer="2552" to-port="0" />
+ <edge from-layer="2552" from-port="2" to-layer="2553" to-port="1" />
+ <edge from-layer="2553" from-port="2" to-layer="2554" to-port="1" />
+ <edge from-layer="2554" from-port="2" to-layer="2709" to-port="0" />
+ <edge from-layer="2554" from-port="2" to-layer="2555" to-port="0" />
+ <edge from-layer="2555" from-port="2" to-layer="2556" to-port="0" />
+ <edge from-layer="2556" from-port="2" to-layer="2558" to-port="0" />
+ <edge from-layer="2557" from-port="0" to-layer="2558" to-port="1" />
+ <edge from-layer="2558" from-port="2" to-layer="2582" to-port="1" />
+ <edge from-layer="2559" from-port="1" to-layer="2561" to-port="0" />
+ <edge from-layer="2560" from-port="0" to-layer="2561" to-port="2" />
+ <edge from-layer="2561" from-port="3" to-layer="2577" to-port="0" />
+ <edge from-layer="2562" from-port="0" to-layer="2563" to-port="0" />
+ <edge from-layer="2563" from-port="1" to-layer="2566" to-port="0" />
+ <edge from-layer="2564" from-port="0" to-layer="2565" to-port="0" />
+ <edge from-layer="2565" from-port="1" to-layer="2566" to-port="1" />
+ <edge from-layer="2566" from-port="2" to-layer="2568" to-port="0" />
+ <edge from-layer="2567" from-port="0" to-layer="2568" to-port="1" />
+ <edge from-layer="2568" from-port="2" to-layer="2570" to-port="0" />
+ <edge from-layer="2569" from-port="0" to-layer="2570" to-port="1" />
+ <edge from-layer="2570" from-port="2" to-layer="2571" to-port="0" />
+ <edge from-layer="2571" from-port="1" to-layer="2572" to-port="1" />
+ <edge from-layer="2572" from-port="2" to-layer="2574" to-port="0" />
+ <edge from-layer="2573" from-port="0" to-layer="2574" to-port="1" />
+ <edge from-layer="2574" from-port="2" to-layer="2576" to-port="0" />
+ <edge from-layer="2575" from-port="0" to-layer="2576" to-port="1" />
+ <edge from-layer="2576" from-port="2" to-layer="2577" to-port="1" />
+ <edge from-layer="2577" from-port="2" to-layer="2578" to-port="0" />
+ <edge from-layer="2577" from-port="2" to-layer="2708" to-port="0" />
+ <edge from-layer="2578" from-port="2" to-layer="2579" to-port="0" />
+ <edge from-layer="2579" from-port="2" to-layer="2581" to-port="0" />
+ <edge from-layer="2580" from-port="0" to-layer="2581" to-port="1" />
+ <edge from-layer="2581" from-port="2" to-layer="2582" to-port="2" />
+ <edge from-layer="2582" from-port="4" to-layer="2584" to-port="0" />
+ <edge from-layer="2583" from-port="0" to-layer="2584" to-port="1" />
+ <edge from-layer="2584" from-port="2" to-layer="2585" to-port="0" />
+ <edge from-layer="2585" from-port="2" to-layer="2596" to-port="0" />
+ <edge from-layer="2586" from-port="0" to-layer="2587" to-port="0" />
+ <edge from-layer="2587" from-port="1" to-layer="2590" to-port="0" />
+ <edge from-layer="2588" from-port="0" to-layer="2589" to-port="0" />
+ <edge from-layer="2589" from-port="1" to-layer="2590" to-port="1" />
+ <edge from-layer="2590" from-port="2" to-layer="2592" to-port="0" />
+ <edge from-layer="2591" from-port="0" to-layer="2592" to-port="1" />
+ <edge from-layer="2592" from-port="2" to-layer="2594" to-port="0" />
+ <edge from-layer="2593" from-port="0" to-layer="2594" to-port="1" />
+ <edge from-layer="2594" from-port="2" to-layer="2595" to-port="0" />
+ <edge from-layer="2595" from-port="1" to-layer="2596" to-port="1" />
+ <edge from-layer="2596" from-port="2" to-layer="2597" to-port="1" />
+ <edge from-layer="2597" from-port="2" to-layer="2645" to-port="0" />
+ <edge from-layer="2597" from-port="2" to-layer="2601" to-port="0" />
+ <edge from-layer="2597" from-port="2" to-layer="2608" to-port="0" />
+ <edge from-layer="2598" from-port="0" to-layer="2609" to-port="0" />
+ <edge from-layer="2599" from-port="0" to-layer="2607" to-port="0" />
+ <edge from-layer="2600" from-port="0" to-layer="2601" to-port="1" />
+ <edge from-layer="2601" from-port="2" to-layer="2603" to-port="0" />
+ <edge from-layer="2602" from-port="0" to-layer="2603" to-port="1" />
+ <edge from-layer="2603" from-port="2" to-layer="2605" to-port="0" />
+ <edge from-layer="2604" from-port="0" to-layer="2605" to-port="1" />
+ <edge from-layer="2605" from-port="2" to-layer="2606" to-port="0" />
+ <edge from-layer="2606" from-port="1" to-layer="2607" to-port="1" />
+ <edge from-layer="2607" from-port="2" to-layer="2608" to-port="1" />
+ <edge from-layer="2608" from-port="2" to-layer="2609" to-port="1" />
+ <edge from-layer="2609" from-port="2" to-layer="2632" to-port="0" />
+ <edge from-layer="2609" from-port="2" to-layer="2620" to-port="0" />
+ <edge from-layer="2610" from-port="0" to-layer="2611" to-port="0" />
+ <edge from-layer="2611" from-port="1" to-layer="2614" to-port="0" />
+ <edge from-layer="2612" from-port="0" to-layer="2613" to-port="0" />
+ <edge from-layer="2613" from-port="1" to-layer="2614" to-port="1" />
+ <edge from-layer="2614" from-port="2" to-layer="2616" to-port="0" />
+ <edge from-layer="2615" from-port="0" to-layer="2616" to-port="1" />
+ <edge from-layer="2616" from-port="2" to-layer="2618" to-port="0" />
+ <edge from-layer="2617" from-port="0" to-layer="2618" to-port="1" />
+ <edge from-layer="2618" from-port="2" to-layer="2619" to-port="0" />
+ <edge from-layer="2619" from-port="1" to-layer="2620" to-port="1" />
+ <edge from-layer="2620" from-port="2" to-layer="2621" to-port="0" />
+ <edge from-layer="2621" from-port="1" to-layer="2633" to-port="0" />
+ <edge from-layer="2622" from-port="0" to-layer="2623" to-port="0" />
+ <edge from-layer="2623" from-port="1" to-layer="2626" to-port="0" />
+ <edge from-layer="2624" from-port="0" to-layer="2625" to-port="0" />
+ <edge from-layer="2625" from-port="1" to-layer="2626" to-port="1" />
+ <edge from-layer="2626" from-port="2" to-layer="2628" to-port="0" />
+ <edge from-layer="2627" from-port="0" to-layer="2628" to-port="1" />
+ <edge from-layer="2628" from-port="2" to-layer="2630" to-port="0" />
+ <edge from-layer="2629" from-port="0" to-layer="2630" to-port="1" />
+ <edge from-layer="2630" from-port="2" to-layer="2631" to-port="0" />
+ <edge from-layer="2631" from-port="1" to-layer="2632" to-port="1" />
+ <edge from-layer="2632" from-port="2" to-layer="2633" to-port="1" />
+ <edge from-layer="2633" from-port="2" to-layer="2644" to-port="0" />
+ <edge from-layer="2634" from-port="0" to-layer="2635" to-port="0" />
+ <edge from-layer="2635" from-port="1" to-layer="2638" to-port="0" />
+ <edge from-layer="2636" from-port="0" to-layer="2637" to-port="0" />
+ <edge from-layer="2637" from-port="1" to-layer="2638" to-port="1" />
+ <edge from-layer="2638" from-port="2" to-layer="2640" to-port="0" />
+ <edge from-layer="2639" from-port="0" to-layer="2640" to-port="1" />
+ <edge from-layer="2640" from-port="2" to-layer="2642" to-port="0" />
+ <edge from-layer="2641" from-port="0" to-layer="2642" to-port="1" />
+ <edge from-layer="2642" from-port="2" to-layer="2643" to-port="0" />
+ <edge from-layer="2643" from-port="1" to-layer="2644" to-port="1" />
+ <edge from-layer="2644" from-port="2" to-layer="2645" to-port="1" />
+ <edge from-layer="2645" from-port="2" to-layer="2775" to-port="0" />
+ <edge from-layer="2645" from-port="2" to-layer="2648" to-port="0" />
+ <edge from-layer="2645" from-port="2" to-layer="2655" to-port="0" />
+ <edge from-layer="2646" from-port="0" to-layer="2654" to-port="0" />
+ <edge from-layer="2647" from-port="0" to-layer="2648" to-port="1" />
+ <edge from-layer="2648" from-port="2" to-layer="2650" to-port="0" />
+ <edge from-layer="2649" from-port="0" to-layer="2650" to-port="1" />
+ <edge from-layer="2650" from-port="2" to-layer="2652" to-port="0" />
+ <edge from-layer="2651" from-port="0" to-layer="2652" to-port="1" />
+ <edge from-layer="2652" from-port="2" to-layer="2653" to-port="0" />
+ <edge from-layer="2653" from-port="1" to-layer="2654" to-port="1" />
+ <edge from-layer="2654" from-port="2" to-layer="2655" to-port="1" />
+ <edge from-layer="2655" from-port="2" to-layer="2656" to-port="1" />
+ <edge from-layer="2656" from-port="2" to-layer="2733" to-port="0" />
+ <edge from-layer="2656" from-port="2" to-layer="2687" to-port="0" />
+ <edge from-layer="2656" from-port="2" to-layer="2667" to-port="0" />
+ <edge from-layer="2657" from-port="0" to-layer="2658" to-port="0" />
+ <edge from-layer="2658" from-port="1" to-layer="2661" to-port="0" />
+ <edge from-layer="2659" from-port="0" to-layer="2660" to-port="0" />
+ <edge from-layer="2660" from-port="1" to-layer="2661" to-port="1" />
+ <edge from-layer="2661" from-port="2" to-layer="2663" to-port="0" />
+ <edge from-layer="2662" from-port="0" to-layer="2663" to-port="1" />
+ <edge from-layer="2663" from-port="2" to-layer="2665" to-port="0" />
+ <edge from-layer="2664" from-port="0" to-layer="2665" to-port="1" />
+ <edge from-layer="2665" from-port="2" to-layer="2666" to-port="0" />
+ <edge from-layer="2666" from-port="1" to-layer="2667" to-port="1" />
+ <edge from-layer="2667" from-port="2" to-layer="2669" to-port="0" />
+ <edge from-layer="2668" from-port="0" to-layer="2669" to-port="1" />
+ <edge from-layer="2669" from-port="2" to-layer="2671" to-port="0" />
+ <edge from-layer="2670" from-port="0" to-layer="2671" to-port="1" />
+ <edge from-layer="2671" from-port="2" to-layer="2672" to-port="1" />
+ <edge from-layer="2672" from-port="2" to-layer="2673" to-port="0" />
+ <edge from-layer="2672" from-port="2" to-layer="2756" to-port="0" />
+ <edge from-layer="2674" from-port="1" to-layer="2676" to-port="0" />
+ <edge from-layer="2675" from-port="0" to-layer="2676" to-port="2" />
+ <edge from-layer="2676" from-port="3" to-layer="2706" to-port="0" />
+ <edge from-layer="2677" from-port="0" to-layer="2678" to-port="0" />
+ <edge from-layer="2678" from-port="1" to-layer="2681" to-port="0" />
+ <edge from-layer="2679" from-port="0" to-layer="2680" to-port="0" />
+ <edge from-layer="2680" from-port="1" to-layer="2681" to-port="1" />
+ <edge from-layer="2681" from-port="2" to-layer="2683" to-port="0" />
+ <edge from-layer="2682" from-port="0" to-layer="2683" to-port="1" />
+ <edge from-layer="2683" from-port="2" to-layer="2685" to-port="0" />
+ <edge from-layer="2684" from-port="0" to-layer="2685" to-port="1" />
+ <edge from-layer="2685" from-port="2" to-layer="2686" to-port="0" />
+ <edge from-layer="2686" from-port="1" to-layer="2687" to-port="1" />
+ <edge from-layer="2687" from-port="2" to-layer="2689" to-port="0" />
+ <edge from-layer="2688" from-port="0" to-layer="2689" to-port="1" />
+ <edge from-layer="2689" from-port="2" to-layer="2691" to-port="0" />
+ <edge from-layer="2690" from-port="0" to-layer="2691" to-port="1" />
+ <edge from-layer="2691" from-port="2" to-layer="2702" to-port="0" />
+ <edge from-layer="2691" from-port="2" to-layer="2696" to-port="0" />
+ <edge from-layer="2691" from-port="2" to-layer="2692" to-port="0" />
+ <edge from-layer="2692" from-port="2" to-layer="2705" to-port="0" />
+ <edge from-layer="2693" from-port="0" to-layer="2696" to-port="1" />
+ <edge from-layer="2694" from-port="0" to-layer="2696" to-port="2" />
+ <edge from-layer="2695" from-port="0" to-layer="2696" to-port="3" />
+ <edge from-layer="2696" from-port="5" to-layer="2698" to-port="0" />
+ <edge from-layer="2697" from-port="0" to-layer="2698" to-port="1" />
+ <edge from-layer="2698" from-port="2" to-layer="2703" to-port="0" />
+ <edge from-layer="2699" from-port="0" to-layer="2702" to-port="1" />
+ <edge from-layer="2700" from-port="0" to-layer="2702" to-port="2" />
+ <edge from-layer="2701" from-port="0" to-layer="2702" to-port="3" />
+ <edge from-layer="2702" from-port="5" to-layer="2703" to-port="1" />
+ <edge from-layer="2703" from-port="2" to-layer="2704" to-port="0" />
+ <edge from-layer="2704" from-port="2" to-layer="2705" to-port="1" />
+ <edge from-layer="2705" from-port="2" to-layer="2706" to-port="1" />
+ <edge from-layer="2706" from-port="2" to-layer="2752" to-port="0" />
+ <edge from-layer="2706" from-port="2" to-layer="2707" to-port="0" />
+ <edge from-layer="2722" from-port="0" to-layer="2834" to-port="0" />
+ <edge from-layer="2723" from-port="0" to-layer="2724" to-port="0" />
+ <edge from-layer="2724" from-port="1" to-layer="2727" to-port="0" />
+ <edge from-layer="2725" from-port="0" to-layer="2726" to-port="0" />
+ <edge from-layer="2726" from-port="1" to-layer="2727" to-port="1" />
+ <edge from-layer="2727" from-port="2" to-layer="2729" to-port="0" />
+ <edge from-layer="2728" from-port="0" to-layer="2729" to-port="1" />
+ <edge from-layer="2729" from-port="2" to-layer="2731" to-port="0" />
+ <edge from-layer="2730" from-port="0" to-layer="2731" to-port="1" />
+ <edge from-layer="2731" from-port="2" to-layer="2732" to-port="0" />
+ <edge from-layer="2732" from-port="1" to-layer="2733" to-port="1" />
+ <edge from-layer="2733" from-port="2" to-layer="2735" to-port="0" />
+ <edge from-layer="2734" from-port="0" to-layer="2735" to-port="1" />
+ <edge from-layer="2735" from-port="2" to-layer="2737" to-port="0" />
+ <edge from-layer="2736" from-port="0" to-layer="2737" to-port="1" />
+ <edge from-layer="2737" from-port="2" to-layer="2742" to-port="0" />
+ <edge from-layer="2737" from-port="2" to-layer="2738" to-port="0" />
+ <edge from-layer="2737" from-port="2" to-layer="2748" to-port="0" />
+ <edge from-layer="2738" from-port="2" to-layer="2751" to-port="0" />
+ <edge from-layer="2739" from-port="0" to-layer="2742" to-port="1" />
+ <edge from-layer="2740" from-port="0" to-layer="2742" to-port="2" />
+ <edge from-layer="2741" from-port="0" to-layer="2742" to-port="3" />
+ <edge from-layer="2742" from-port="5" to-layer="2744" to-port="0" />
+ <edge from-layer="2743" from-port="0" to-layer="2744" to-port="1" />
+ <edge from-layer="2744" from-port="2" to-layer="2749" to-port="0" />
+ <edge from-layer="2745" from-port="0" to-layer="2748" to-port="1" />
+ <edge from-layer="2746" from-port="0" to-layer="2748" to-port="2" />
+ <edge from-layer="2747" from-port="0" to-layer="2748" to-port="3" />
+ <edge from-layer="2748" from-port="5" to-layer="2749" to-port="1" />
+ <edge from-layer="2749" from-port="2" to-layer="2750" to-port="0" />
+ <edge from-layer="2750" from-port="2" to-layer="2751" to-port="1" />
+ <edge from-layer="2751" from-port="2" to-layer="2760" to-port="0" />
+ <edge from-layer="2752" from-port="2" to-layer="2753" to-port="0" />
+ <edge from-layer="2753" from-port="2" to-layer="2755" to-port="0" />
+ <edge from-layer="2754" from-port="0" to-layer="2755" to-port="1" />
+ <edge from-layer="2755" from-port="2" to-layer="2760" to-port="1" />
+ <edge from-layer="2756" from-port="2" to-layer="2757" to-port="0" />
+ <edge from-layer="2757" from-port="2" to-layer="2759" to-port="0" />
+ <edge from-layer="2758" from-port="0" to-layer="2759" to-port="1" />
+ <edge from-layer="2759" from-port="2" to-layer="2760" to-port="2" />
+ <edge from-layer="2760" from-port="4" to-layer="2762" to-port="0" />
+ <edge from-layer="2761" from-port="0" to-layer="2762" to-port="1" />
+ <edge from-layer="2762" from-port="2" to-layer="2763" to-port="0" />
+ <edge from-layer="2763" from-port="2" to-layer="2774" to-port="0" />
+ <edge from-layer="2764" from-port="0" to-layer="2765" to-port="0" />
+ <edge from-layer="2765" from-port="1" to-layer="2768" to-port="0" />
+ <edge from-layer="2766" from-port="0" to-layer="2767" to-port="0" />
+ <edge from-layer="2767" from-port="1" to-layer="2768" to-port="1" />
+ <edge from-layer="2768" from-port="2" to-layer="2770" to-port="0" />
+ <edge from-layer="2769" from-port="0" to-layer="2770" to-port="1" />
+ <edge from-layer="2770" from-port="2" to-layer="2772" to-port="0" />
+ <edge from-layer="2771" from-port="0" to-layer="2772" to-port="1" />
+ <edge from-layer="2772" from-port="2" to-layer="2773" to-port="0" />
+ <edge from-layer="2773" from-port="1" to-layer="2774" to-port="1" />
+ <edge from-layer="2774" from-port="2" to-layer="2775" to-port="1" />
+ <edge from-layer="2775" from-port="2" to-layer="2779" to-port="0" />
+ <edge from-layer="2775" from-port="2" to-layer="2786" to-port="0" />
+ <edge from-layer="2775" from-port="2" to-layer="2823" to-port="0" />
+ <edge from-layer="2776" from-port="0" to-layer="2787" to-port="0" />
+ <edge from-layer="2777" from-port="0" to-layer="2785" to-port="0" />
+ <edge from-layer="2778" from-port="0" to-layer="2779" to-port="1" />
+ <edge from-layer="2779" from-port="2" to-layer="2781" to-port="0" />
+ <edge from-layer="2780" from-port="0" to-layer="2781" to-port="1" />
+ <edge from-layer="2781" from-port="2" to-layer="2783" to-port="0" />
+ <edge from-layer="2782" from-port="0" to-layer="2783" to-port="1" />
+ <edge from-layer="2783" from-port="2" to-layer="2784" to-port="0" />
+ <edge from-layer="2784" from-port="1" to-layer="2785" to-port="1" />
+ <edge from-layer="2785" from-port="2" to-layer="2786" to-port="1" />
+ <edge from-layer="2786" from-port="2" to-layer="2787" to-port="1" />
+ <edge from-layer="2787" from-port="2" to-layer="2810" to-port="0" />
+ <edge from-layer="2787" from-port="2" to-layer="2798" to-port="0" />
+ <edge from-layer="2788" from-port="0" to-layer="2789" to-port="0" />
+ <edge from-layer="2789" from-port="1" to-layer="2792" to-port="0" />
+ <edge from-layer="2790" from-port="0" to-layer="2791" to-port="0" />
+ <edge from-layer="2791" from-port="1" to-layer="2792" to-port="1" />
+ <edge from-layer="2792" from-port="2" to-layer="2794" to-port="0" />
+ <edge from-layer="2793" from-port="0" to-layer="2794" to-port="1" />
+ <edge from-layer="2794" from-port="2" to-layer="2796" to-port="0" />
+ <edge from-layer="2795" from-port="0" to-layer="2796" to-port="1" />
+ <edge from-layer="2796" from-port="2" to-layer="2797" to-port="0" />
+ <edge from-layer="2797" from-port="1" to-layer="2798" to-port="1" />
+ <edge from-layer="2798" from-port="2" to-layer="2799" to-port="0" />
+ <edge from-layer="2799" from-port="1" to-layer="2811" to-port="0" />
+ <edge from-layer="2800" from-port="0" to-layer="2801" to-port="0" />
+ <edge from-layer="2801" from-port="1" to-layer="2804" to-port="0" />
+ <edge from-layer="2802" from-port="0" to-layer="2803" to-port="0" />
+ <edge from-layer="2803" from-port="1" to-layer="2804" to-port="1" />
+ <edge from-layer="2804" from-port="2" to-layer="2806" to-port="0" />
+ <edge from-layer="2805" from-port="0" to-layer="2806" to-port="1" />
+ <edge from-layer="2806" from-port="2" to-layer="2808" to-port="0" />
+ <edge from-layer="2807" from-port="0" to-layer="2808" to-port="1" />
+ <edge from-layer="2808" from-port="2" to-layer="2809" to-port="0" />
+ <edge from-layer="2809" from-port="1" to-layer="2810" to-port="1" />
+ <edge from-layer="2810" from-port="2" to-layer="2811" to-port="1" />
+ <edge from-layer="2811" from-port="2" to-layer="2822" to-port="0" />
+ <edge from-layer="2812" from-port="0" to-layer="2813" to-port="0" />
+ <edge from-layer="2813" from-port="1" to-layer="2816" to-port="0" />
+ <edge from-layer="2814" from-port="0" to-layer="2815" to-port="0" />
+ <edge from-layer="2815" from-port="1" to-layer="2816" to-port="1" />
+ <edge from-layer="2816" from-port="2" to-layer="2818" to-port="0" />
+ <edge from-layer="2817" from-port="0" to-layer="2818" to-port="1" />
+ <edge from-layer="2818" from-port="2" to-layer="2820" to-port="0" />
+ <edge from-layer="2819" from-port="0" to-layer="2820" to-port="1" />
+ <edge from-layer="2820" from-port="2" to-layer="2821" to-port="0" />
+ <edge from-layer="2821" from-port="1" to-layer="2822" to-port="1" />
+ <edge from-layer="2822" from-port="2" to-layer="2823" to-port="1" />
+ <edge from-layer="2823" from-port="2" to-layer="2826" to-port="0" />
+ <edge from-layer="2823" from-port="2" to-layer="2833" to-port="0" />
+ <edge from-layer="2824" from-port="0" to-layer="2832" to-port="0" />
+ <edge from-layer="2825" from-port="0" to-layer="2826" to-port="1" />
+ <edge from-layer="2826" from-port="2" to-layer="2828" to-port="0" />
+ <edge from-layer="2827" from-port="0" to-layer="2828" to-port="1" />
+ <edge from-layer="2828" from-port="2" to-layer="2830" to-port="0" />
+ <edge from-layer="2829" from-port="0" to-layer="2830" to-port="1" />
+ <edge from-layer="2830" from-port="2" to-layer="2831" to-port="0" />
+ <edge from-layer="2831" from-port="1" to-layer="2832" to-port="1" />
+ <edge from-layer="2832" from-port="2" to-layer="2833" to-port="1" />
+ <edge from-layer="2833" from-port="2" to-layer="2834" to-port="1" />
+ <edge from-layer="2834" from-port="2" to-layer="2843" to-port="0" />
+ <edge from-layer="2835" from-port="0" to-layer="2836" to-port="0" />
+ <edge from-layer="2836" from-port="1" to-layer="2839" to-port="0" />
+ <edge from-layer="2837" from-port="0" to-layer="2838" to-port="0" />
+ <edge from-layer="2838" from-port="1" to-layer="2839" to-port="1" />
+ <edge from-layer="2839" from-port="2" to-layer="2841" to-port="0" />
+ <edge from-layer="2840" from-port="0" to-layer="2841" to-port="1" />
+ <edge from-layer="2841" from-port="2" to-layer="2842" to-port="0" />
+ <edge from-layer="2842" from-port="1" to-layer="2843" to-port="1" />
+ <edge from-layer="2843" from-port="2" to-layer="2844" to-port="0" />
+ </edges>
+ <rt_info>
+ <Runtime_version value="2024.5.0-17288-7975fa5da0c-refs/pull/3856/head" />
+ <conversion_parameters>
+ <framework value="pytorch" />
+ <is_python_object value="True" />
+ </conversion_parameters>
+ <nncf>
+ <friendly_names_were_updated value="True" />
+ <weight_compression>
+ <advanced_parameters value="{'statistics_path': None, 'awq_params': {'subset_size': 32, 'percent_to_apply': 0.002, 'alpha_min': 0.0, 'alpha_max': 1.0, 'steps': 100}, 'scale_estimation_params': {'subset_size': 64, 'initial_steps': 5, 'scale_steps': 5, 'weight_penalty': -1.0}, 'gptq_params': {'damp_percent': 0.1, 'block_size': 128, 'subset_size': 128}, 'lora_correction_params': {'adapter_rank': 8, 'num_iterations': 3, 'apply_regularization': True, 'subset_size': 128, 'use_int8_adapters': True}}" />
+ <all_layers value="False" />
+ <awq value="False" />
+ <backup_mode value="int8_asym" />
+ <gptq value="False" />
+ <group_size value="64" />
+ <ignored_scope value="[]" />
+ <lora_correction value="False" />
+ <mode value="int4_asym" />
+ <ratio value="1.0" />
+ <scale_estimation value="False" />
+ <sensitivity_metric value="max_activation_variance" />
+ </weight_compression>
+ </nncf>
+ <optimum>
+ <optimum_intel_version value="1.21.0.dev0+c94b3f5" />
+ <optimum_version value="1.23.3" />
+ <pytorch_version value="2.4.1+cpu" />
+ <transformers_version value="4.46.3" />
+ </optimum>
+ <runtime_options>
+ <ACTIVATIONS_SCALE_FACTOR value="8.0" />
+ <KV_CACHE_PRECISION value="f16" />
+ </runtime_options>
+ </rt_info>
+</net>