Synchronizing local compiler cache.

.gitattributes

@@ -6942,3 +6942,5 @@ neuronxcc-2.15.143.0+e39249ad/MODULE_1b6d7576802d84bea017+39f12043/model.neff fi
 neuronxcc-2.15.143.0+e39249ad/MODULE_c578025927b993537ac9+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.143.0+e39249ad/MODULE_3f1e5540fd9596fb060c+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.143.0+e39249ad/MODULE_d1581937ee5c6df81b5c+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_190feb29500cacd7e00a+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_6db75f07c04ef97a526d+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.15.143.0+e39249ad/0_REGISTRY/0.0.28.dev2/inference/llama/princeton-nlp/Sheared-LLaMA-1.3B/6463b97c9f6a564faf0c.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": "2.15.143.0+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_190feb29500cacd7e00a+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_190feb29500cacd7e00a+39f12043/model.hlo_module.pb

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

neuronxcc-2.15.143.0+e39249ad/MODULE_190feb29500cacd7e00a+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c76f2861d7c656d51ae83de34db3110e9b1ed570f2c917ee024cf042ea9e613
+size 3707904

neuronxcc-2.15.143.0+e39249ad/MODULE_6db75f07c04ef97a526d+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_6db75f07c04ef97a526d+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52e0587e7186fbf2b98c3745a509f849dca3f599b1e1a1ec46008aac58e23c0f
+size 324022

neuronxcc-2.15.143.0+e39249ad/MODULE_6db75f07c04ef97a526d+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:343dc609eea22c2350d617dad79f8b3573613fcbf94e0879d826db965c270362
+size 30342144