Synchronizing local compiler cache.

.gitattributes

@@ -4464,3 +4464,5 @@ neuronxcc-2.15.128.0+56dc5a86/MODULE_3abd647589f9579c6c09+39f12043/model.neff fi
 neuronxcc-2.15.128.0+56dc5a86/MODULE_5b25ad1caf95e9a59155+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_5921fe7851e4324b0368+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_a908ab12c136ae19ce8b+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_314799b42bb1f45cec02+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_3e25032356233fc84aee+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.15.128.0+56dc5a86/0_REGISTRY/0.0.25.dev0/inference/llama/princeton-nlp/Sheared-LLaMA-1.3B/10d86b739c51fa605a82.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": "2.15.128.0+56dc5a86", "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-2.15.128.0+56dc5a86/MODULE_314799b42bb1f45cec02+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_314799b42bb1f45cec02+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b379fbffa9e946a6b1b89ca9ef0845f3ad3b656c1beeb4a929991a4282fd2f4
+size 262218

neuronxcc-2.15.128.0+56dc5a86/MODULE_314799b42bb1f45cec02+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d05bb31c08aa357466f17d33ddd3d55a57908354763554a6c417602ab0253ed9
+size 3718144

neuronxcc-2.15.128.0+56dc5a86/MODULE_3e25032356233fc84aee+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_3e25032356233fc84aee+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57fa6f5e2529f92b5e4c6f4968e0a3651f4dacbfbc3d000b7d668bfe3bf06f7a
+size 336202

neuronxcc-2.15.128.0+56dc5a86/MODULE_3e25032356233fc84aee+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b92ba7a8af1c9459af4e9990308e7b7c964671af759af1ccb692311cbe9493fe
+size 29481984