Synchronizing local compiler cache.

.gitattributes

@@ -1630,3 +1630,5 @@ neuronxcc-2.16.372.0+4a9b2326/MODULE_b8f93d36af83ab17ab85+39f12043/model.neff fi
 neuronxcc-2.16.372.0+4a9b2326/MODULE_f567467231139f46c50e+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.143.0+e39249ad/MODULE_2e229618015e416964c5+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.143.0+e39249ad/MODULE_fa839a3354ecb514534b+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_22728fdf8942185937a3+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_f6b2b8267d631f2f0fad+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.15.143.0+e39249ad/0_REGISTRY/0.0.28.dev2/inference/llama/NousResearch/Hermes-2-Theta-Llama-3-8B/c1effc43754275f46f73.json

@@ -0,0 +1 @@
+{"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 128000, "eos_token_id": 128003, "head_dim": 128, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 14336, "max_position_embeddings": 8192, "mlp_bias": false, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "NousResearch/Hermes-2-Theta-Llama-3-8B", "checkpoint_revision": "57a73110702e7b05ba3f39fef36297454c680725", "compiler_type": "neuronx-cc", "compiler_version": "2.15.143.0+e39249ad", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 8, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 500000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 128256}

neuronxcc-2.15.143.0+e39249ad/MODULE_22728fdf8942185937a3+39f12043/compile_flags.json

@@ -0,0 +1 @@
+["--model-type=transformer", "--auto-cast=none", "--execute-repetition=1"]

neuronxcc-2.15.143.0+e39249ad/MODULE_22728fdf8942185937a3+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6262be51029c3eea47a4a1dbedec75caf2c13c8af4ec4fe2b335e94514525004
+size 444306

neuronxcc-2.15.143.0+e39249ad/MODULE_22728fdf8942185937a3+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c78cc265f1153e0896e4b17646ece64a861022f8d1cabdea967a5c97eedbcd6
+size 6042624

neuronxcc-2.15.143.0+e39249ad/MODULE_f6b2b8267d631f2f0fad+39f12043/compile_flags.json

@@ -0,0 +1 @@
+["--model-type=transformer", "--auto-cast=none", "--execute-repetition=1"]

neuronxcc-2.15.143.0+e39249ad/MODULE_f6b2b8267d631f2f0fad+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc1a14daf775573428682f87075bb804d80212e0b04b1157bc26d991a572d6ec
+size 363704

neuronxcc-2.15.143.0+e39249ad/MODULE_f6b2b8267d631f2f0fad+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d33d9a9268df35fb36dccf5821be6c6ab5ec23a8512c096f31098a7b57755b53
+size 10466304