Synchronizing local compiler cache.

.gitattributes

@@ -4484,3 +4484,5 @@ neuronxcc-2.15.128.0+56dc5a86/MODULE_93afc660e7e7f555747d+39f12043/model.neff fi
 neuronxcc-2.15.128.0+56dc5a86/MODULE_a91e1e61c179ae79142a+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_e2c7bb3483b611fefe7a+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_e9040de0b7ab62e09639+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_10417c15e9c50b8ff95c+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_dbaefb7a2d2fe2794521+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-2-7b-chat-hf/b1071f4efd70944473d3.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": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 4096, "mlp_bias": false, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 8, "checkpoint_id": "meta-llama/Llama-2-7b-chat-hf", "checkpoint_revision": "f5db02db724555f92da89c216ac04704f23d4590", "compiler_type": "neuronx-cc", "compiler_version": "2.15.128.0+56dc5a86", "num_cores": 8, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 32, "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_10417c15e9c50b8ff95c+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_10417c15e9c50b8ff95c+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87b074d2b2f1b0d3051a2b7f75eceb310b78f196b5b1df9425ede5d3fa8a056c
+size 349797

neuronxcc-2.15.128.0+56dc5a86/MODULE_10417c15e9c50b8ff95c+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:add6ae69c634bd267243a04f6b6422931f8dad9fdd4d3f1026dce396e6657826
+size 5039104

neuronxcc-2.15.128.0+56dc5a86/MODULE_dbaefb7a2d2fe2794521+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_dbaefb7a2d2fe2794521+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b028a61d07d7ca5abf15dacabc2f475c529c2f087c5c3a3bb4c2ae8cd4adfa32
+size 448317

neuronxcc-2.15.128.0+56dc5a86/MODULE_dbaefb7a2d2fe2794521+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8ebea5aa0edadd172f7bfe5697bc1e87d33232b4b1cb8091851ff72c9064402
+size 30198784