Hugging Face's logo

keras
/
qwen2.5_7b_en

Text Generation
KerasHub
Model card Files Community
Divyasreepat commited on
Commit
b235646
·
verified ·
1 Parent(s): dab26db

Upload folder using huggingface_hub

Browse files
Files changed (12) hide show
  1. README.md +25 -0
  2. assets/tokenizer/merges.txt +0 -0
  3. assets/tokenizer/vocabulary.json +0 -0
  4. config.json +22 -0
  5. metadata.json +9 -0
  6. model.weights.json +261 -0
  7. model_00000.weights.h5 +3 -0
  8. model_00001.weights.h5 +3 -0
  9. model_00002.weights.h5 +3 -0
  10. preprocessor.json +65 -0
  11. task.json +95 -0
  12. tokenizer.json +44 -0
README.md ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: keras-hub
3
+ pipeline_tag: text-generation
4
+ ---
5
+ This is a [`Qwen` model](https://keras.io/api/keras_hub/models/qwen) uploaded using the KerasHub library and can be used with JAX, TensorFlow, and PyTorch backends.
6
+ This model is related to a `CausalLM` task.
7
+
8
+ Model config:
9
+ * **name:** qwen_backbone
10
+ * **trainable:** True
11
+ * **vocabulary_size:** 152064
12
+ * **num_layers:** 28
13
+ * **num_query_heads:** 28
14
+ * **hidden_dim:** 3584
15
+ * **intermediate_dim:** 18944
16
+ * **rope_max_wavelength:** 1000000.0
17
+ * **rope_scaling_factor:** 1.0
18
+ * **num_key_value_heads:** 4
19
+ * **layer_norm_epsilon:** 1e-06
20
+ * **dropout:** 0
21
+ * **tie_word_embeddings:** False
22
+ * **use_sliding_window_attention:** False
23
+ * **sliding_window_size:** 131072
24
+
25
+ This model card has been generated automatically and should be completed by the model author. See [Model Cards documentation](https://huggingface.co/docs/hub/model-cards) for more information.
assets/tokenizer/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
assets/tokenizer/vocabulary.json ADDED
The diff for this file is too large to render. See raw diff
 
config.json ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.qwen.qwen_backbone",
3
+ "class_name": "QwenBackbone",
4
+ "config": {
5
+ "name": "qwen_backbone",
6
+ "trainable": true,
7
+ "vocabulary_size": 152064,
8
+ "num_layers": 28,
9
+ "num_query_heads": 28,
10
+ "hidden_dim": 3584,
11
+ "intermediate_dim": 18944,
12
+ "rope_max_wavelength": 1000000.0,
13
+ "rope_scaling_factor": 1.0,
14
+ "num_key_value_heads": 4,
15
+ "layer_norm_epsilon": 1e-06,
16
+ "dropout": 0,
17
+ "tie_word_embeddings": false,
18
+ "use_sliding_window_attention": false,
19
+ "sliding_window_size": 131072
20
+ },
21
+ "registered_name": "keras_hub>QwenBackbone"
22
+ }
metadata.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "keras_version": "3.10.0.dev2025051003",
3
+ "keras_hub_version": "0.21.0.dev202505090407",
4
+ "parameter_count": 7615616512,
5
+ "date_saved": "2025-05-11@00:34:37",
6
+ "tasks": [
7
+ "CausalLM"
8
+ ]
9
+ }
model.weights.json ADDED
@@ -0,0 +1,261 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 30462466048.0
4
+ },
5
+ "weight_map": {
6
+ "/layers/reversible_embedding/vars": "model_00000.weights.h5",
7
+ "/layers/qwen_transformer_decoder/_feedforward_gate_dense/vars": "model_00000.weights.h5",
8
+ "/layers/qwen_transformer_decoder/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
9
+ "/layers/qwen_transformer_decoder/_feedforward_layernorm/vars": "model_00000.weights.h5",
10
+ "/layers/qwen_transformer_decoder/_feedforward_output_dense/vars": "model_00000.weights.h5",
11
+ "/layers/qwen_transformer_decoder/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
12
+ "/layers/qwen_transformer_decoder/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
13
+ "/layers/qwen_transformer_decoder/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
14
+ "/layers/qwen_transformer_decoder/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
15
+ "/layers/qwen_transformer_decoder/_self_attention_layernorm/vars": "model_00000.weights.h5",
16
+ "/layers/qwen_transformer_decoder_1/_feedforward_gate_dense/vars": "model_00000.weights.h5",
17
+ "/layers/qwen_transformer_decoder_1/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
18
+ "/layers/qwen_transformer_decoder_1/_feedforward_layernorm/vars": "model_00000.weights.h5",
19
+ "/layers/qwen_transformer_decoder_1/_feedforward_output_dense/vars": "model_00000.weights.h5",
20
+ "/layers/qwen_transformer_decoder_1/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
21
+ "/layers/qwen_transformer_decoder_1/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
22
+ "/layers/qwen_transformer_decoder_1/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
23
+ "/layers/qwen_transformer_decoder_1/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
24
+ "/layers/qwen_transformer_decoder_1/_self_attention_layernorm/vars": "model_00000.weights.h5",
25
+ "/layers/qwen_transformer_decoder_2/_feedforward_gate_dense/vars": "model_00000.weights.h5",
26
+ "/layers/qwen_transformer_decoder_2/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
27
+ "/layers/qwen_transformer_decoder_2/_feedforward_layernorm/vars": "model_00000.weights.h5",
28
+ "/layers/qwen_transformer_decoder_2/_feedforward_output_dense/vars": "model_00000.weights.h5",
29
+ "/layers/qwen_transformer_decoder_2/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
30
+ "/layers/qwen_transformer_decoder_2/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
31
+ "/layers/qwen_transformer_decoder_2/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
32
+ "/layers/qwen_transformer_decoder_2/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
33
+ "/layers/qwen_transformer_decoder_2/_self_attention_layernorm/vars": "model_00000.weights.h5",
34
+ "/layers/qwen_transformer_decoder_3/_feedforward_gate_dense/vars": "model_00000.weights.h5",
35
+ "/layers/qwen_transformer_decoder_3/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
36
+ "/layers/qwen_transformer_decoder_3/_feedforward_layernorm/vars": "model_00000.weights.h5",
37
+ "/layers/qwen_transformer_decoder_3/_feedforward_output_dense/vars": "model_00000.weights.h5",
38
+ "/layers/qwen_transformer_decoder_3/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
39
+ "/layers/qwen_transformer_decoder_3/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
40
+ "/layers/qwen_transformer_decoder_3/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
41
+ "/layers/qwen_transformer_decoder_3/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
42
+ "/layers/qwen_transformer_decoder_3/_self_attention_layernorm/vars": "model_00000.weights.h5",
43
+ "/layers/qwen_transformer_decoder_4/_feedforward_gate_dense/vars": "model_00000.weights.h5",
44
+ "/layers/qwen_transformer_decoder_4/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
45
+ "/layers/qwen_transformer_decoder_4/_feedforward_layernorm/vars": "model_00000.weights.h5",
46
+ "/layers/qwen_transformer_decoder_4/_feedforward_output_dense/vars": "model_00000.weights.h5",
47
+ "/layers/qwen_transformer_decoder_4/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
48
+ "/layers/qwen_transformer_decoder_4/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
49
+ "/layers/qwen_transformer_decoder_4/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
50
+ "/layers/qwen_transformer_decoder_4/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
51
+ "/layers/qwen_transformer_decoder_4/_self_attention_layernorm/vars": "model_00000.weights.h5",
52
+ "/layers/qwen_transformer_decoder_5/_feedforward_gate_dense/vars": "model_00000.weights.h5",
53
+ "/layers/qwen_transformer_decoder_5/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
54
+ "/layers/qwen_transformer_decoder_5/_feedforward_layernorm/vars": "model_00000.weights.h5",
55
+ "/layers/qwen_transformer_decoder_5/_feedforward_output_dense/vars": "model_00000.weights.h5",
56
+ "/layers/qwen_transformer_decoder_5/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
57
+ "/layers/qwen_transformer_decoder_5/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
58
+ "/layers/qwen_transformer_decoder_5/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
59
+ "/layers/qwen_transformer_decoder_5/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
60
+ "/layers/qwen_transformer_decoder_5/_self_attention_layernorm/vars": "model_00000.weights.h5",
61
+ "/layers/qwen_transformer_decoder_6/_feedforward_gate_dense/vars": "model_00000.weights.h5",
62
+ "/layers/qwen_transformer_decoder_6/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
63
+ "/layers/qwen_transformer_decoder_6/_feedforward_layernorm/vars": "model_00000.weights.h5",
64
+ "/layers/qwen_transformer_decoder_6/_feedforward_output_dense/vars": "model_00001.weights.h5",
65
+ "/layers/qwen_transformer_decoder_6/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
66
+ "/layers/qwen_transformer_decoder_6/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
67
+ "/layers/qwen_transformer_decoder_6/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
68
+ "/layers/qwen_transformer_decoder_6/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
69
+ "/layers/qwen_transformer_decoder_6/_self_attention_layernorm/vars": "model_00001.weights.h5",
70
+ "/layers/qwen_transformer_decoder_7/_feedforward_gate_dense/vars": "model_00001.weights.h5",
71
+ "/layers/qwen_transformer_decoder_7/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
72
+ "/layers/qwen_transformer_decoder_7/_feedforward_layernorm/vars": "model_00001.weights.h5",
73
+ "/layers/qwen_transformer_decoder_7/_feedforward_output_dense/vars": "model_00001.weights.h5",
74
+ "/layers/qwen_transformer_decoder_7/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
75
+ "/layers/qwen_transformer_decoder_7/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
76
+ "/layers/qwen_transformer_decoder_7/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
77
+ "/layers/qwen_transformer_decoder_7/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
78
+ "/layers/qwen_transformer_decoder_7/_self_attention_layernorm/vars": "model_00001.weights.h5",
79
+ "/layers/qwen_transformer_decoder_8/_feedforward_gate_dense/vars": "model_00001.weights.h5",
80
+ "/layers/qwen_transformer_decoder_8/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
81
+ "/layers/qwen_transformer_decoder_8/_feedforward_layernorm/vars": "model_00001.weights.h5",
82
+ "/layers/qwen_transformer_decoder_8/_feedforward_output_dense/vars": "model_00001.weights.h5",
83
+ "/layers/qwen_transformer_decoder_8/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
84
+ "/layers/qwen_transformer_decoder_8/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
85
+ "/layers/qwen_transformer_decoder_8/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
86
+ "/layers/qwen_transformer_decoder_8/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
87
+ "/layers/qwen_transformer_decoder_8/_self_attention_layernorm/vars": "model_00001.weights.h5",
88
+ "/layers/qwen_transformer_decoder_9/_feedforward_gate_dense/vars": "model_00001.weights.h5",
89
+ "/layers/qwen_transformer_decoder_9/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
90
+ "/layers/qwen_transformer_decoder_9/_feedforward_layernorm/vars": "model_00001.weights.h5",
91
+ "/layers/qwen_transformer_decoder_9/_feedforward_output_dense/vars": "model_00001.weights.h5",
92
+ "/layers/qwen_transformer_decoder_9/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
93
+ "/layers/qwen_transformer_decoder_9/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
94
+ "/layers/qwen_transformer_decoder_9/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
95
+ "/layers/qwen_transformer_decoder_9/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
96
+ "/layers/qwen_transformer_decoder_9/_self_attention_layernorm/vars": "model_00001.weights.h5",
97
+ "/layers/qwen_transformer_decoder_10/_feedforward_gate_dense/vars": "model_00001.weights.h5",
98
+ "/layers/qwen_transformer_decoder_10/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
99
+ "/layers/qwen_transformer_decoder_10/_feedforward_layernorm/vars": "model_00001.weights.h5",
100
+ "/layers/qwen_transformer_decoder_10/_feedforward_output_dense/vars": "model_00001.weights.h5",
101
+ "/layers/qwen_transformer_decoder_10/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
102
+ "/layers/qwen_transformer_decoder_10/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
103
+ "/layers/qwen_transformer_decoder_10/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
104
+ "/layers/qwen_transformer_decoder_10/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
105
+ "/layers/qwen_transformer_decoder_10/_self_attention_layernorm/vars": "model_00001.weights.h5",
106
+ "/layers/qwen_transformer_decoder_11/_feedforward_gate_dense/vars": "model_00001.weights.h5",
107
+ "/layers/qwen_transformer_decoder_11/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
108
+ "/layers/qwen_transformer_decoder_11/_feedforward_layernorm/vars": "model_00001.weights.h5",
109
+ "/layers/qwen_transformer_decoder_11/_feedforward_output_dense/vars": "model_00001.weights.h5",
110
+ "/layers/qwen_transformer_decoder_11/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
111
+ "/layers/qwen_transformer_decoder_11/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
112
+ "/layers/qwen_transformer_decoder_11/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
113
+ "/layers/qwen_transformer_decoder_11/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
114
+ "/layers/qwen_transformer_decoder_11/_self_attention_layernorm/vars": "model_00001.weights.h5",
115
+ "/layers/qwen_transformer_decoder_12/_feedforward_gate_dense/vars": "model_00001.weights.h5",
116
+ "/layers/qwen_transformer_decoder_12/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
117
+ "/layers/qwen_transformer_decoder_12/_feedforward_layernorm/vars": "model_00001.weights.h5",
118
+ "/layers/qwen_transformer_decoder_12/_feedforward_output_dense/vars": "model_00001.weights.h5",
119
+ "/layers/qwen_transformer_decoder_12/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
120
+ "/layers/qwen_transformer_decoder_12/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
121
+ "/layers/qwen_transformer_decoder_12/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
122
+ "/layers/qwen_transformer_decoder_12/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
123
+ "/layers/qwen_transformer_decoder_12/_self_attention_layernorm/vars": "model_00001.weights.h5",
124
+ "/layers/qwen_transformer_decoder_13/_feedforward_gate_dense/vars": "model_00001.weights.h5",
125
+ "/layers/qwen_transformer_decoder_13/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
126
+ "/layers/qwen_transformer_decoder_13/_feedforward_layernorm/vars": "model_00001.weights.h5",
127
+ "/layers/qwen_transformer_decoder_13/_feedforward_output_dense/vars": "model_00001.weights.h5",
128
+ "/layers/qwen_transformer_decoder_13/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
129
+ "/layers/qwen_transformer_decoder_13/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
130
+ "/layers/qwen_transformer_decoder_13/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
131
+ "/layers/qwen_transformer_decoder_13/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
132
+ "/layers/qwen_transformer_decoder_13/_self_attention_layernorm/vars": "model_00001.weights.h5",
133
+ "/layers/qwen_transformer_decoder_14/_feedforward_gate_dense/vars": "model_00001.weights.h5",
134
+ "/layers/qwen_transformer_decoder_14/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
135
+ "/layers/qwen_transformer_decoder_14/_feedforward_layernorm/vars": "model_00001.weights.h5",
136
+ "/layers/qwen_transformer_decoder_14/_feedforward_output_dense/vars": "model_00001.weights.h5",
137
+ "/layers/qwen_transformer_decoder_14/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
138
+ "/layers/qwen_transformer_decoder_14/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
139
+ "/layers/qwen_transformer_decoder_14/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
140
+ "/layers/qwen_transformer_decoder_14/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
141
+ "/layers/qwen_transformer_decoder_14/_self_attention_layernorm/vars": "model_00001.weights.h5",
142
+ "/layers/qwen_transformer_decoder_15/_feedforward_gate_dense/vars": "model_00001.weights.h5",
143
+ "/layers/qwen_transformer_decoder_15/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
144
+ "/layers/qwen_transformer_decoder_15/_feedforward_layernorm/vars": "model_00001.weights.h5",
145
+ "/layers/qwen_transformer_decoder_15/_feedforward_output_dense/vars": "model_00001.weights.h5",
146
+ "/layers/qwen_transformer_decoder_15/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
147
+ "/layers/qwen_transformer_decoder_15/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
148
+ "/layers/qwen_transformer_decoder_15/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
149
+ "/layers/qwen_transformer_decoder_15/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
150
+ "/layers/qwen_transformer_decoder_15/_self_attention_layernorm/vars": "model_00001.weights.h5",
151
+ "/layers/qwen_transformer_decoder_16/_feedforward_gate_dense/vars": "model_00001.weights.h5",
152
+ "/layers/qwen_transformer_decoder_16/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
153
+ "/layers/qwen_transformer_decoder_16/_feedforward_layernorm/vars": "model_00001.weights.h5",
154
+ "/layers/qwen_transformer_decoder_16/_feedforward_output_dense/vars": "model_00001.weights.h5",
155
+ "/layers/qwen_transformer_decoder_16/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
156
+ "/layers/qwen_transformer_decoder_16/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
157
+ "/layers/qwen_transformer_decoder_16/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
158
+ "/layers/qwen_transformer_decoder_16/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
159
+ "/layers/qwen_transformer_decoder_16/_self_attention_layernorm/vars": "model_00001.weights.h5",
160
+ "/layers/qwen_transformer_decoder_17/_feedforward_gate_dense/vars": "model_00001.weights.h5",
161
+ "/layers/qwen_transformer_decoder_17/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
162
+ "/layers/qwen_transformer_decoder_17/_feedforward_layernorm/vars": "model_00001.weights.h5",
163
+ "/layers/qwen_transformer_decoder_17/_feedforward_output_dense/vars": "model_00001.weights.h5",
164
+ "/layers/qwen_transformer_decoder_17/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
165
+ "/layers/qwen_transformer_decoder_17/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
166
+ "/layers/qwen_transformer_decoder_17/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
167
+ "/layers/qwen_transformer_decoder_17/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
168
+ "/layers/qwen_transformer_decoder_17/_self_attention_layernorm/vars": "model_00001.weights.h5",
169
+ "/layers/qwen_transformer_decoder_18/_feedforward_gate_dense/vars": "model_00002.weights.h5",
170
+ "/layers/qwen_transformer_decoder_18/_feedforward_intermediate_dense/vars": "model_00002.weights.h5",
171
+ "/layers/qwen_transformer_decoder_18/_feedforward_layernorm/vars": "model_00002.weights.h5",
172
+ "/layers/qwen_transformer_decoder_18/_feedforward_output_dense/vars": "model_00002.weights.h5",
173
+ "/layers/qwen_transformer_decoder_18/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5",
174
+ "/layers/qwen_transformer_decoder_18/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5",
175
+ "/layers/qwen_transformer_decoder_18/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5",
176
+ "/layers/qwen_transformer_decoder_18/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5",
177
+ "/layers/qwen_transformer_decoder_18/_self_attention_layernorm/vars": "model_00002.weights.h5",
178
+ "/layers/qwen_transformer_decoder_19/_feedforward_gate_dense/vars": "model_00002.weights.h5",
179
+ "/layers/qwen_transformer_decoder_19/_feedforward_intermediate_dense/vars": "model_00002.weights.h5",
180
+ "/layers/qwen_transformer_decoder_19/_feedforward_layernorm/vars": "model_00002.weights.h5",
181
+ "/layers/qwen_transformer_decoder_19/_feedforward_output_dense/vars": "model_00002.weights.h5",
182
+ "/layers/qwen_transformer_decoder_19/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5",
183
+ "/layers/qwen_transformer_decoder_19/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5",
184
+ "/layers/qwen_transformer_decoder_19/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5",
185
+ "/layers/qwen_transformer_decoder_19/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5",
186
+ "/layers/qwen_transformer_decoder_19/_self_attention_layernorm/vars": "model_00002.weights.h5",
187
+ "/layers/qwen_transformer_decoder_20/_feedforward_gate_dense/vars": "model_00002.weights.h5",
188
+ "/layers/qwen_transformer_decoder_20/_feedforward_intermediate_dense/vars": "model_00002.weights.h5",
189
+ "/layers/qwen_transformer_decoder_20/_feedforward_layernorm/vars": "model_00002.weights.h5",
190
+ "/layers/qwen_transformer_decoder_20/_feedforward_output_dense/vars": "model_00002.weights.h5",
191
+ "/layers/qwen_transformer_decoder_20/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5",
192
+ "/layers/qwen_transformer_decoder_20/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5",
193
+ "/layers/qwen_transformer_decoder_20/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5",
194
+ "/layers/qwen_transformer_decoder_20/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5",
195
+ "/layers/qwen_transformer_decoder_20/_self_attention_layernorm/vars": "model_00002.weights.h5",
196
+ "/layers/qwen_transformer_decoder_21/_feedforward_gate_dense/vars": "model_00002.weights.h5",
197
+ "/layers/qwen_transformer_decoder_21/_feedforward_intermediate_dense/vars": "model_00002.weights.h5",
198
+ "/layers/qwen_transformer_decoder_21/_feedforward_layernorm/vars": "model_00002.weights.h5",
199
+ "/layers/qwen_transformer_decoder_21/_feedforward_output_dense/vars": "model_00002.weights.h5",
200
+ "/layers/qwen_transformer_decoder_21/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5",
201
+ "/layers/qwen_transformer_decoder_21/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5",
202
+ "/layers/qwen_transformer_decoder_21/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5",
203
+ "/layers/qwen_transformer_decoder_21/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5",
204
+ "/layers/qwen_transformer_decoder_21/_self_attention_layernorm/vars": "model_00002.weights.h5",
205
+ "/layers/qwen_transformer_decoder_22/_feedforward_gate_dense/vars": "model_00002.weights.h5",
206
+ "/layers/qwen_transformer_decoder_22/_feedforward_intermediate_dense/vars": "model_00002.weights.h5",
207
+ "/layers/qwen_transformer_decoder_22/_feedforward_layernorm/vars": "model_00002.weights.h5",
208
+ "/layers/qwen_transformer_decoder_22/_feedforward_output_dense/vars": "model_00002.weights.h5",
209
+ "/layers/qwen_transformer_decoder_22/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5",
210
+ "/layers/qwen_transformer_decoder_22/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5",
211
+ "/layers/qwen_transformer_decoder_22/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5",
212
+ "/layers/qwen_transformer_decoder_22/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5",
213
+ "/layers/qwen_transformer_decoder_22/_self_attention_layernorm/vars": "model_00002.weights.h5",
214
+ "/layers/qwen_transformer_decoder_23/_feedforward_gate_dense/vars": "model_00002.weights.h5",
215
+ "/layers/qwen_transformer_decoder_23/_feedforward_intermediate_dense/vars": "model_00002.weights.h5",
216
+ "/layers/qwen_transformer_decoder_23/_feedforward_layernorm/vars": "model_00002.weights.h5",
217
+ "/layers/qwen_transformer_decoder_23/_feedforward_output_dense/vars": "model_00002.weights.h5",
218
+ "/layers/qwen_transformer_decoder_23/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5",
219
+ "/layers/qwen_transformer_decoder_23/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5",
220
+ "/layers/qwen_transformer_decoder_23/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5",
221
+ "/layers/qwen_transformer_decoder_23/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5",
222
+ "/layers/qwen_transformer_decoder_23/_self_attention_layernorm/vars": "model_00002.weights.h5",
223
+ "/layers/qwen_transformer_decoder_24/_feedforward_gate_dense/vars": "model_00002.weights.h5",
224
+ "/layers/qwen_transformer_decoder_24/_feedforward_intermediate_dense/vars": "model_00002.weights.h5",
225
+ "/layers/qwen_transformer_decoder_24/_feedforward_layernorm/vars": "model_00002.weights.h5",
226
+ "/layers/qwen_transformer_decoder_24/_feedforward_output_dense/vars": "model_00002.weights.h5",
227
+ "/layers/qwen_transformer_decoder_24/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5",
228
+ "/layers/qwen_transformer_decoder_24/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5",
229
+ "/layers/qwen_transformer_decoder_24/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5",
230
+ "/layers/qwen_transformer_decoder_24/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5",
231
+ "/layers/qwen_transformer_decoder_24/_self_attention_layernorm/vars": "model_00002.weights.h5",
232
+ "/layers/qwen_transformer_decoder_25/_feedforward_gate_dense/vars": "model_00002.weights.h5",
233
+ "/layers/qwen_transformer_decoder_25/_feedforward_intermediate_dense/vars": "model_00002.weights.h5",
234
+ "/layers/qwen_transformer_decoder_25/_feedforward_layernorm/vars": "model_00002.weights.h5",
235
+ "/layers/qwen_transformer_decoder_25/_feedforward_output_dense/vars": "model_00002.weights.h5",
236
+ "/layers/qwen_transformer_decoder_25/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5",
237
+ "/layers/qwen_transformer_decoder_25/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5",
238
+ "/layers/qwen_transformer_decoder_25/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5",
239
+ "/layers/qwen_transformer_decoder_25/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5",
240
+ "/layers/qwen_transformer_decoder_25/_self_attention_layernorm/vars": "model_00002.weights.h5",
241
+ "/layers/qwen_transformer_decoder_26/_feedforward_gate_dense/vars": "model_00002.weights.h5",
242
+ "/layers/qwen_transformer_decoder_26/_feedforward_intermediate_dense/vars": "model_00002.weights.h5",
243
+ "/layers/qwen_transformer_decoder_26/_feedforward_layernorm/vars": "model_00002.weights.h5",
244
+ "/layers/qwen_transformer_decoder_26/_feedforward_output_dense/vars": "model_00002.weights.h5",
245
+ "/layers/qwen_transformer_decoder_26/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5",
246
+ "/layers/qwen_transformer_decoder_26/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5",
247
+ "/layers/qwen_transformer_decoder_26/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5",
248
+ "/layers/qwen_transformer_decoder_26/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5",
249
+ "/layers/qwen_transformer_decoder_26/_self_attention_layernorm/vars": "model_00002.weights.h5",
250
+ "/layers/qwen_transformer_decoder_27/_feedforward_gate_dense/vars": "model_00002.weights.h5",
251
+ "/layers/qwen_transformer_decoder_27/_feedforward_intermediate_dense/vars": "model_00002.weights.h5",
252
+ "/layers/qwen_transformer_decoder_27/_feedforward_layernorm/vars": "model_00002.weights.h5",
253
+ "/layers/qwen_transformer_decoder_27/_feedforward_output_dense/vars": "model_00002.weights.h5",
254
+ "/layers/qwen_transformer_decoder_27/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5",
255
+ "/layers/qwen_transformer_decoder_27/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5",
256
+ "/layers/qwen_transformer_decoder_27/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5",
257
+ "/layers/qwen_transformer_decoder_27/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5",
258
+ "/layers/qwen_transformer_decoder_27/_self_attention_layernorm/vars": "model_00002.weights.h5",
259
+ "/layers/qwen_layer_norm/vars": "model_00002.weights.h5"
260
+ }
261
+ }
model_00000.weights.h5 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4784f9dc08e787ef9f984ba473d6b3f1fffcf62df773b05704e67b4c96357eec
3
+ size 10496772680
model_00001.weights.h5 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d38d74febdddb60c132abfc89bcdb39b2a40c9c4c594b8971b825149b246d3d
3
+ size 10644013712
model_00002.weights.h5 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e1213db4fd7d43addda5ef61c1ca1e15814781d1afe429c274b44af660a4aed
3
+ size 9322680800
preprocessor.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.qwen.qwen_causal_lm_preprocessor",
3
+ "class_name": "QwenCausalLMPreprocessor",
4
+ "config": {
5
+ "name": "qwen_causal_lm_preprocessor_2",
6
+ "trainable": true,
7
+ "dtype": {
8
+ "module": "keras",
9
+ "class_name": "DTypePolicy",
10
+ "config": {
11
+ "name": "float32"
12
+ },
13
+ "registered_name": null
14
+ },
15
+ "tokenizer": {
16
+ "module": "keras_hub.src.models.qwen.qwen_tokenizer",
17
+ "class_name": "QwenTokenizer",
18
+ "config": {
19
+ "name": "qwen_tokenizer",
20
+ "trainable": true,
21
+ "dtype": {
22
+ "module": "keras",
23
+ "class_name": "DTypePolicy",
24
+ "config": {
25
+ "name": "int32"
26
+ },
27
+ "registered_name": null
28
+ },
29
+ "config_file": "tokenizer.json",
30
+ "sequence_length": null,
31
+ "add_prefix_space": false,
32
+ "unsplittable_tokens": [
33
+ "<|object_ref_end|>",
34
+ "<|repo_name|>",
35
+ "<|fim_suffix|>",
36
+ "<|fim_prefix|>",
37
+ "<|im_start|>",
38
+ "<|vision_end|>",
39
+ "</tool_call>",
40
+ "<tool_call>",
41
+ "<|im_end|>",
42
+ "<|box_start|>",
43
+ "<|image_pad|>",
44
+ "<|file_sep|>",
45
+ "<|quad_end|>",
46
+ "<|video_pad|>",
47
+ "<|vision_start|>",
48
+ "<|fim_middle|>",
49
+ "<|object_ref_start|>",
50
+ "<|endoftext|>",
51
+ "<|box_end|>",
52
+ "<|fim_pad|>",
53
+ "<|vision_pad|>",
54
+ "<|quad_start|>"
55
+ ]
56
+ },
57
+ "registered_name": "keras_hub>QwenTokenizer"
58
+ },
59
+ "config_file": "preprocessor.json",
60
+ "sequence_length": 1024,
61
+ "add_start_token": true,
62
+ "add_end_token": true
63
+ },
64
+ "registered_name": "keras_hub>QwenCausalLMPreprocessor"
65
+ }
task.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.qwen.qwen_causal_lm",
3
+ "class_name": "QwenCausalLM",
4
+ "config": {
5
+ "backbone": {
6
+ "module": "keras_hub.src.models.qwen.qwen_backbone",
7
+ "class_name": "QwenBackbone",
8
+ "config": {
9
+ "name": "qwen_backbone",
10
+ "trainable": true,
11
+ "vocabulary_size": 152064,
12
+ "num_layers": 28,
13
+ "num_query_heads": 28,
14
+ "hidden_dim": 3584,
15
+ "intermediate_dim": 18944,
16
+ "rope_max_wavelength": 1000000.0,
17
+ "rope_scaling_factor": 1.0,
18
+ "num_key_value_heads": 4,
19
+ "layer_norm_epsilon": 1e-06,
20
+ "dropout": 0,
21
+ "tie_word_embeddings": false,
22
+ "use_sliding_window_attention": false,
23
+ "sliding_window_size": 131072
24
+ },
25
+ "registered_name": "keras_hub>QwenBackbone"
26
+ },
27
+ "preprocessor": {
28
+ "module": "keras_hub.src.models.qwen.qwen_causal_lm_preprocessor",
29
+ "class_name": "QwenCausalLMPreprocessor",
30
+ "config": {
31
+ "name": "qwen_causal_lm_preprocessor_2",
32
+ "trainable": true,
33
+ "dtype": {
34
+ "module": "keras",
35
+ "class_name": "DTypePolicy",
36
+ "config": {
37
+ "name": "float32"
38
+ },
39
+ "registered_name": null
40
+ },
41
+ "tokenizer": {
42
+ "module": "keras_hub.src.models.qwen.qwen_tokenizer",
43
+ "class_name": "QwenTokenizer",
44
+ "config": {
45
+ "name": "qwen_tokenizer",
46
+ "trainable": true,
47
+ "dtype": {
48
+ "module": "keras",
49
+ "class_name": "DTypePolicy",
50
+ "config": {
51
+ "name": "int32"
52
+ },
53
+ "registered_name": null
54
+ },
55
+ "config_file": "tokenizer.json",
56
+ "sequence_length": null,
57
+ "add_prefix_space": false,
58
+ "unsplittable_tokens": [
59
+ "<|object_ref_end|>",
60
+ "<|repo_name|>",
61
+ "<|fim_suffix|>",
62
+ "<|fim_prefix|>",
63
+ "<|im_start|>",
64
+ "<|vision_end|>",
65
+ "</tool_call>",
66
+ "<tool_call>",
67
+ "<|im_end|>",
68
+ "<|box_start|>",
69
+ "<|image_pad|>",
70
+ "<|file_sep|>",
71
+ "<|quad_end|>",
72
+ "<|video_pad|>",
73
+ "<|vision_start|>",
74
+ "<|fim_middle|>",
75
+ "<|object_ref_start|>",
76
+ "<|endoftext|>",
77
+ "<|box_end|>",
78
+ "<|fim_pad|>",
79
+ "<|vision_pad|>",
80
+ "<|quad_start|>"
81
+ ]
82
+ },
83
+ "registered_name": "keras_hub>QwenTokenizer"
84
+ },
85
+ "config_file": "preprocessor.json",
86
+ "sequence_length": 1024,
87
+ "add_start_token": true,
88
+ "add_end_token": true
89
+ },
90
+ "registered_name": "keras_hub>QwenCausalLMPreprocessor"
91
+ },
92
+ "name": "qwen_causal_lm"
93
+ },
94
+ "registered_name": "keras_hub>QwenCausalLM"
95
+ }
tokenizer.json ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.qwen.qwen_tokenizer",
3
+ "class_name": "QwenTokenizer",
4
+ "config": {
5
+ "name": "qwen_tokenizer",
6
+ "trainable": true,
7
+ "dtype": {
8
+ "module": "keras",
9
+ "class_name": "DTypePolicy",
10
+ "config": {
11
+ "name": "int32"
12
+ },
13
+ "registered_name": null
14
+ },
15
+ "config_file": "tokenizer.json",
16
+ "sequence_length": null,
17
+ "add_prefix_space": false,
18
+ "unsplittable_tokens": [
19
+ "<|object_ref_end|>",
20
+ "<|repo_name|>",
21
+ "<|fim_suffix|>",
22
+ "<|fim_prefix|>",
23
+ "<|im_start|>",
24
+ "<|vision_end|>",
25
+ "</tool_call>",
26
+ "<tool_call>",
27
+ "<|im_end|>",
28
+ "<|box_start|>",
29
+ "<|image_pad|>",
30
+ "<|file_sep|>",
31
+ "<|quad_end|>",
32
+ "<|video_pad|>",
33
+ "<|vision_start|>",
34
+ "<|fim_middle|>",
35
+ "<|object_ref_start|>",
36
+ "<|endoftext|>",
37
+ "<|box_end|>",
38
+ "<|fim_pad|>",
39
+ "<|vision_pad|>",
40
+ "<|quad_start|>"
41
+ ]
42
+ },
43
+ "registered_name": "keras_hub>QwenTokenizer"
44
+ }