Synchronizing local compiler cache.

.gitattributes

@@ -4494,3 +4494,5 @@ neuronxcc-2.15.128.0+56dc5a86/MODULE_7f99a181290326a67234+39f12043/model.neff fi
 neuronxcc-2.15.128.0+56dc5a86/MODULE_d35838e9962cd0fda4dd+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_34053e7320475bab291f+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_e2b4adc50d217e9a8485+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_dd559faf80bd6656a54c+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_e9d088c006b4ad88cd17+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/099c77f14ea9d810a1d7.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": 32, "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_dd559faf80bd6656a54c+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_dd559faf80bd6656a54c+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f8cc7d33f4dc85362cc018ba4fe2014be11ae105a0bd2e4200fb0266efd8f94
+size 349797

neuronxcc-2.15.128.0+56dc5a86/MODULE_dd559faf80bd6656a54c+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc03641482fee2c3a404e6e7071e98226838b54201f63959678c30fbf119404f
+size 12964864

neuronxcc-2.15.128.0+56dc5a86/MODULE_e9d088c006b4ad88cd17+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_e9d088c006b4ad88cd17+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11a0da334580a4a29c906efb5326d7c27dd778d384ed80da1c8e6de2d0fd0c51
+size 448317

neuronxcc-2.15.128.0+56dc5a86/MODULE_e9d088c006b4ad88cd17+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fc152e01289f8779d581baef924039fc67bed938856edbe36f84d3f8a9725f2
+size 30198784