Synchronizing local compiler cache.

.gitattributes

@@ -6064,3 +6064,5 @@ neuronxcc-2.15.143.0+e39249ad/MODULE_923bc8ee2b722ecd5bd3+39f12043/model.neff fi
 neuronxcc-2.15.143.0+e39249ad/MODULE_a26bcbd6aeae1e6c24a8+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.143.0+e39249ad/MODULE_b26e890076ad95278803+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.143.0+e39249ad/MODULE_dee2f0c1b80f5d3c7edb+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_a4549af46c0cb7ca1cfd+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_c3eb0724b4c268206eb3+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.15.143.0+e39249ad/0_REGISTRY/0.0.27.dev0/inference/llama/princeton-nlp/Sheared-LLaMA-1.3B/b583d8ca856d4fb50088.json

@@ -0,0 +1 @@
+{"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "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": "2.15.143.0+e39249ad", "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-2.15.143.0+e39249ad/MODULE_a4549af46c0cb7ca1cfd+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_a4549af46c0cb7ca1cfd+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:994d91dbc95bfb7b004388c6c71ba019d3b6454a2cfd8e5f0dc1e8b1d01ba7a3
+size 324010

neuronxcc-2.15.143.0+e39249ad/MODULE_a4549af46c0cb7ca1cfd+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aceeae18ae104010c1fa7ab1a7e5982fc1853cf38e733989301c66d60d54d6f6
+size 30342144

neuronxcc-2.15.143.0+e39249ad/MODULE_c3eb0724b4c268206eb3+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_c3eb0724b4c268206eb3+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f326dddefe5b56d004b06c3b97be42fc1a8dff991c0ac1929f9418d53ddfe7d4
+size 254730

neuronxcc-2.15.143.0+e39249ad/MODULE_c3eb0724b4c268206eb3+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74c4416c7f32b6ba49f23400d4cf093e0942cc43b1b4fa15e98ac37de0c7b262
+size 3707904