Synchronizing local compiler cache.

.gitattributes

@@ -1498,3 +1498,5 @@ neuronxcc-2.15.143.0+e39249ad/MODULE_c0b96836ba4c3a46f1d1+39f12043/model.neff fi
 neuronxcc-2.15.143.0+e39249ad/MODULE_c7fee5e6e0f12889c294+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.143.0+e39249ad/MODULE_2eeecff6f903dd1aaf30+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.143.0+e39249ad/MODULE_f46c84f4be3f40eafec4+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.16.345.0+69131dd3/MODULE_ae142d763c7fe49cf917+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.16.345.0+69131dd3/MODULE_eeea370b5fb22f353681+613edded/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.16.345.0+69131dd3/0_REGISTRY/0.0.28.dev0/inference/llama/NousResearch/Hermes-2-Theta-Llama-3-8B/6747313734a381250e06.json

@@ -0,0 +1 @@
+{"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 128000, "eos_token_id": 128003, "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": "2.16.345.0+69131dd3", "num_cores": 2, "sequence_length": 2048, "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-2.16.345.0+69131dd3/MODULE_ae142d763c7fe49cf917+613edded/compile_flags.json

@@ -0,0 +1 @@
+["--target=trn1", "--model-type=transformer", "--auto-cast=none", "--execute-repetition=1"]

neuronxcc-2.16.345.0+69131dd3/MODULE_ae142d763c7fe49cf917+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:603185b382959e404710a4ecc30622d9a9debb0f88faa347822c5259af1d29a7
+size 302136

neuronxcc-2.16.345.0+69131dd3/MODULE_ae142d763c7fe49cf917+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28e77af4d976d6cd257e61b1b53452894bcd1c57f775fe7dacdf938fe6c8f59a
+size 2571264

neuronxcc-2.16.345.0+69131dd3/MODULE_eeea370b5fb22f353681+613edded/compile_flags.json

@@ -0,0 +1 @@
+["--target=trn1", "--model-type=transformer", "--auto-cast=none", "--execute-repetition=1"]

neuronxcc-2.16.345.0+69131dd3/MODULE_eeea370b5fb22f353681+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03a4257e79223681830effe26e22fc5ef1b0bad633edd181880dc4203e7470c7
+size 363743

neuronxcc-2.16.345.0+69131dd3/MODULE_eeea370b5fb22f353681+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d822c72f0e15f53d84cba9d8c8fc9af246f9dc62114e295caab2ad9c1195d963
+size 9790464