Synchronizing local compiler cache.

.gitattributes

@@ -7639,3 +7639,5 @@ neuronxcc-2.17.194.0+d312836f/MODULE_349901478ff7d1ca3ce4+613edded/model.neff fi
 neuronxcc-2.17.194.0+d312836f/MODULE_4a56093d8b6239d969e7+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.17.194.0+d312836f/MODULE_a6cfe103ba38d7ced324+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.17.194.0+d312836f/MODULE_d02bc244b8a319b871c1+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.17.194.0+d312836f/MODULE_34d371c764b2edea8dcd+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.17.194.0+d312836f/MODULE_3925d01728915c012cb3+613edded/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.17.194.0+d312836f/0_REGISTRY/0.2.0.dev2/inference/llama/meta-llama/Meta-Llama-3.1-8B/9580f944931f95eff7e8.json

@@ -0,0 +1,45 @@
+{
+  "_entry_class": "SingleModelCacheEntry",
+  "_model_id": "meta-llama/Meta-Llama-3.1-8B",
+  "_task": "text-generation",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 131072,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "neuron": {
+    "auto_cast_type": "bf16",
+    "batch_size": 1,
+    "checkpoint_id": "meta-llama/Meta-Llama-3.1-8B",
+    "checkpoint_revision": "d04e592bb4f6aa9cfee91e2e20afa771667e1d4b",
+    "compiler_type": "neuronx-cc",
+    "compiler_version": "2.17.194.0+d312836f",
+    "num_cores": 2,
+    "sequence_length": 4096,
+    "task": "text-generation"
+  },
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "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,
+  "use_cache": true,
+  "vocab_size": 128256
+}

neuronxcc-2.17.194.0+d312836f/MODULE_34d371c764b2edea8dcd+613edded/compile_flags.json

@@ -0,0 +1 @@
+["--target=trn1", "--model-type=transformer", "--auto-cast=none", "--execute-repetition=1"]

neuronxcc-2.17.194.0+d312836f/MODULE_34d371c764b2edea8dcd+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b3e44cf86dd1d0e68e104812550ae07601f6173588a9533c6638fc6b8119354
+size 331813

neuronxcc-2.17.194.0+d312836f/MODULE_34d371c764b2edea8dcd+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f2d4856bbcad34d17c68ef684ff0d11c9acfacd31834dab80dcaedfa9ffaf1c
+size 8346624

neuronxcc-2.17.194.0+d312836f/MODULE_3925d01728915c012cb3+613edded/compile_flags.json

@@ -0,0 +1 @@
+["--target=trn1", "--model-type=transformer", "--auto-cast=none", "--execute-repetition=1"]

neuronxcc-2.17.194.0+d312836f/MODULE_3925d01728915c012cb3+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e75fb9291e8c024cec13a06162453c87d518e5319f621027126b1fc60a730bf
+size 421622

neuronxcc-2.17.194.0+d312836f/MODULE_3925d01728915c012cb3+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89a10a6c9feb38268c507a44e79ea6677d82cea1288747322742ed4b18558e0b
+size 6339584