Synchronizing local compiler cache.

.gitattributes

@@ -5675,3 +5675,5 @@ neuronxcc-2.15.128.0+56dc5a86/MODULE_d2af06511ee8aec7657a+39f12043/model.neff fi
 neuronxcc-2.15.128.0+56dc5a86/MODULE_e3246d94d543670da65b+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_ea00da9543e66ea4e87a+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_ed6f3686a322093122c3+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_56847724d7506138eff8+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_ab87c4008ae885ec0008+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-3.1-70B-Instruct/416eeffa8bea7170eafc.json

@@ -0,0 +1 @@
+{"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 128000, "eos_token_id": [128001, 128008, 128009], "hidden_act": "silu", "hidden_size": 8192, "initializer_range": 0.02, "intermediate_size": 28672, "max_position_embeddings": 131072, "mlp_bias": false, "model_type": "llama", "neuron": {"auto_cast_type": "bf16", "batch_size": 4, "checkpoint_id": "meta-llama/Llama-3.1-70B-Instruct", "checkpoint_revision": "945c8663693130f8be2ee66210e062158b2a9693", "compiler_type": "neuronx-cc", "compiler_version": "2.15.128.0+56dc5a86", "num_cores": 24, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 64, "num_hidden_layers": 80, "num_key_value_heads": 8, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": {"factor": 8.0, "high_freq_factor": 4.0, "low_freq_factor": 1.0, "original_max_position_embeddings": 8192, "rope_type": "llama3"}, "rope_theta": 500000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 128256}

neuronxcc-2.15.128.0+56dc5a86/MODULE_56847724d7506138eff8+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_56847724d7506138eff8+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bed65e23839648afafceb82960879c1608e13463410b0a32f48562d52e695fea
+size 915713

neuronxcc-2.15.128.0+56dc5a86/MODULE_56847724d7506138eff8+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23277b04d594cab5bcb7bec59872292ecdfdceb0df55b7237f2f40aaef37bf87
+size 10681344

neuronxcc-2.15.128.0+56dc5a86/MODULE_ab87c4008ae885ec0008+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_ab87c4008ae885ec0008+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3094067afc6de1115fe8b735a462bd35a6c7c2310dbe515b41b7f918085ba809
+size 1128340

neuronxcc-2.15.128.0+56dc5a86/MODULE_ab87c4008ae885ec0008+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd3e7d5593cabf8f2a7ebc5a8657b43e0c6a122b7eeff6f3f6108a5ae4043389
+size 2612224