Synchronizing local compiler cache.

.gitattributes

@@ -4786,3 +4786,5 @@ neuronxcc-2.15.128.0+56dc5a86/MODULE_841428df9a7f7fcd1384/text_encoder_2/model.n
 neuronxcc-2.15.128.0+56dc5a86/MODULE_841428df9a7f7fcd1384/unet/model.neuron filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_841428df9a7f7fcd1384/vae_decoder/model.neuron filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_841428df9a7f7fcd1384/vae_encoder/model.neuron filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_1155707dee9639252808+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_2492d0aac41996a9baa4+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/1f8e0c4bbb36017fae31.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": 16, "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_1155707dee9639252808+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_1155707dee9639252808+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e3d03b67f7c9319f546367158489936766f7d4839ec90b12115710e96ba90ea
+size 437405

neuronxcc-2.15.128.0+56dc5a86/MODULE_1155707dee9639252808+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce132d579983c44a7c0cceb6d74ca236232182107c686feb8591c5cfa17bda90
+size 9688064

neuronxcc-2.15.128.0+56dc5a86/MODULE_2492d0aac41996a9baa4+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_2492d0aac41996a9baa4+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0dc7ed8b9deb0c9d8b89e3bcccd8ce921bc165ed652fca65def46f6c2cd6376
+size 560463

neuronxcc-2.15.128.0+56dc5a86/MODULE_2492d0aac41996a9baa4+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac70b8dd59c64f6b66a20a7ca4c679ad769bf882e133b71206b93d548d2d41a0
+size 2151424