Synchronizing local compiler cache.

.gitattributes

@@ -7174,3 +7174,5 @@ neuronxcc-2.16.372.0+4a9b2326/MODULE_c27b585b0715d8fd83b8+613edded/model.neff fi
 neuronxcc-2.16.372.0+4a9b2326/MODULE_3b9cc972c9e78ef8a619+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.16.372.0+4a9b2326/MODULE_5f194d47737258040d3c+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.16.372.0+4a9b2326/MODULE_6c97395a3fefb42ecdbc+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.16.372.0+4a9b2326/MODULE_190feb29500cacd7e00a+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.16.372.0+4a9b2326/MODULE_a639a850df3eb7f79ea2+613edded/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.16.372.0+4a9b2326/0_REGISTRY/0.1.0.dev1/inference/llama/princeton-nlp/Sheared-LLaMA-1.3B/5dc9f8288bce68812153.json

@@ -0,0 +1 @@
+{"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 128, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 5504, "max_position_embeddings": 4096, "mlp_bias": false, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "princeton-nlp/Sheared-LLaMA-1.3B", "checkpoint_revision": "a4b76938edbf571ea7d7d9904861cbdca08809b4", "compiler_type": "neuronx-cc", "compiler_version": "2.16.372.0+4a9b2326", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 16, "num_hidden_layers": 24, "num_key_value_heads": 16, "pad_token_id": 0, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float32", "use_cache": true, "vocab_size": 32000}

neuronxcc-2.16.372.0+4a9b2326/MODULE_190feb29500cacd7e00a+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_190feb29500cacd7e00a+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16171d3648242f807b8d3cd37ed129977f71d43699d8ffebc1788583aacd537e
+size 254742

neuronxcc-2.16.372.0+4a9b2326/MODULE_190feb29500cacd7e00a+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e22dd3fe518570069cc416297c11d5a1b02dc32ad5b4a279f7a0b196ee1fdb94
+size 3687424

neuronxcc-2.16.372.0+4a9b2326/MODULE_a639a850df3eb7f79ea2+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_a639a850df3eb7f79ea2+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22165059b6f2fd7e0acf2f3abae98d233cceab2bf85fb8c488a890077effae58
+size 325462

neuronxcc-2.16.372.0+4a9b2326/MODULE_a639a850df3eb7f79ea2+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08425a98a5cb4b003116003e3d6632c678862b5f0e499ad1e2cab4cf6889c6f0
+size 28417024