Synchronizing local compiler cache.

.gitattributes

@@ -4596,3 +4596,5 @@ neuronxcc-2.15.128.0+56dc5a86/MODULE_d1da223a247ee788be66+39f12043/model.neff fi
 neuronxcc-2.15.128.0+56dc5a86/MODULE_e7605daedd5fe0be5d0d+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_f6f1a154381db618da89+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_fbb6ab8551395a88e298+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_3a5c4fca71eb84dd0a9e+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_efb99cce1e155bef3a4b+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.15.128.0+56dc5a86/0_REGISTRY/0.0.25.dev0/inference/llama/meta-llama/Llama-2-13b-chat-hf/d5aa038b99e99a8951fb.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": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "mlp_bias": false, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 8, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "a2cb7a712bb6e5e736ca7f8cd98167f81a0b5bd8", "compiler_type": "neuronx-cc", "compiler_version": "2.15.128.0+56dc5a86", "num_cores": 12, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float16", "use_cache": true, "vocab_size": 32000}

neuronxcc-2.15.128.0+56dc5a86/MODULE_3a5c4fca71eb84dd0a9e+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_3a5c4fca71eb84dd0a9e+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d4011e4e7792fadd6c8c3fda20a2b9a65b248923e442ce658631e457cfe3c97
+size 560463

neuronxcc-2.15.128.0+56dc5a86/MODULE_3a5c4fca71eb84dd0a9e+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85bb4300e9f459dd4aee47243c4a54bdcaabcb8bb6c07f9f057fc6a72bdee0cb
+size 2151424

neuronxcc-2.15.128.0+56dc5a86/MODULE_efb99cce1e155bef3a4b+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_efb99cce1e155bef3a4b+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:acec36500d392257854e3d2b4c059314bf8581cea5dbba295c0e9f7ec1bfce5d
+size 437405

neuronxcc-2.15.128.0+56dc5a86/MODULE_efb99cce1e155bef3a4b+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:741ab3a80c01ddea9ccad66d5e146c66335a28d01d06169a583e867c04999cdc
+size 6503424