Synchronizing local compiler cache.

.gitattributes

@@ -7196,3 +7196,5 @@ neuronxcc-2.16.372.0+4a9b2326/MODULE_db869219fd71aafdb657+613edded/model.neff fi
 neuronxcc-2.16.372.0+4a9b2326/MODULE_e67fee574b54cbf7316c+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.16.372.0+4a9b2326/MODULE_052ac89b53a3ef1c39e5+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.16.372.0+4a9b2326/MODULE_658a5c1d57118303b833+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.16.372.0+4a9b2326/MODULE_76e8ff114c4f26c8023e+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.16.372.0+4a9b2326/MODULE_c4eacc972a112955fd4c+613edded/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.16.372.0+4a9b2326/0_REGISTRY/0.1.0.dev1/inference/llama/meta-llama/Llama-3.2-3B/60cfa39a4231a8e3e393.json

@@ -0,0 +1 @@
+{"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 128000, "eos_token_id": 128001, "head_dim": 128, "hidden_act": "silu", "hidden_size": 3072, "initializer_range": 0.02, "intermediate_size": 8192, "max_position_embeddings": 131072, "mlp_bias": false, "model_type": "llama", "neuron": {"auto_cast_type": "bf16", "batch_size": 1, "checkpoint_id": "meta-llama/Llama-3.2-3B", "checkpoint_revision": "13afe5124825b4f3751f836b40dafda64c1ed062", "compiler_type": "neuronx-cc", "compiler_version": "2.16.372.0+4a9b2326", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 24, "num_hidden_layers": 28, "num_key_value_heads": 8, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": {"factor": 32.0, "high_freq_factor": 4.0, "low_freq_factor": 1.0, "original_max_position_embeddings": 8192, "rope_type": "llama3"}, "rope_theta": 500000.0, "tie_word_embeddings": true, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 128256}

neuronxcc-2.16.372.0+4a9b2326/MODULE_76e8ff114c4f26c8023e+613edded/compile_flags.json

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

neuronxcc-2.16.372.0+4a9b2326/MODULE_76e8ff114c4f26c8023e+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d06d8e51ff5bbf35070ccf7892a9b0a188171d00302223b9030dce4014601ce
+size 351037

neuronxcc-2.16.372.0+4a9b2326/MODULE_76e8ff114c4f26c8023e+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8a6f619841c52f2da2b599b7500230f43279f5b3bd8f76a3819786d5fd29906
+size 3933184

neuronxcc-2.16.372.0+4a9b2326/MODULE_c4eacc972a112955fd4c+613edded/compile_flags.json

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

neuronxcc-2.16.372.0+4a9b2326/MODULE_c4eacc972a112955fd4c+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e417cc254d8577fb0ee56f758fbd39e47ff69a9746af9eab9636a83c04ee1320
+size 290468

neuronxcc-2.16.372.0+4a9b2326/MODULE_c4eacc972a112955fd4c+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b21c4fcd1693208c9d98febeb8998acc1a316697e8b2fc91cbfddf8ccf84cf50
+size 3421184