Synchronizing local compiler cache.

.gitattributes

@@ -7739,3 +7739,5 @@ neuronxcc-2.17.194.0+d312836f/MODULE_8afc89b9c18489780471+613edded/model.neff fi
 neuronxcc-2.17.194.0+d312836f/MODULE_c27b585b0715d8fd83b8+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.17.194.0+d312836f/MODULE_95dae70c69a000bb4eb6+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.17.194.0+d312836f/MODULE_9f4e39af4e0b35889393+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.17.194.0+d312836f/MODULE_984c41b800d85c96193b+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.17.194.0+d312836f/MODULE_cebb33a758d78c3cffe9+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/Llama-3.1-70B-Instruct/0893aa250f27c3bca5d9.json

@@ -0,0 +1,45 @@
+{
+  "_entry_class": "SingleModelCacheEntry",
+  "_model_id": "meta-llama/Llama-3.1-70B-Instruct",
+  "_task": "text-generation",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "head_dim": 128,
+  "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": "1605565b47bb9346c5515c34102e054115b4f98b",
+    "compiler_type": "neuronx-cc",
+    "compiler_version": "2.17.194.0+d312836f",
+    "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,
+  "use_cache": true,
+  "vocab_size": 128256
+}

neuronxcc-2.17.194.0+d312836f/MODULE_984c41b800d85c96193b+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_984c41b800d85c96193b+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2572d3fef65a73a90017c14abada895860339ee7d3b99ac9208513d853ae7e0
+size 1168672

neuronxcc-2.17.194.0+d312836f/MODULE_984c41b800d85c96193b+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ede71d556400ab2e5d27b428bc4cd7274b046973bb0353549450bf57797d66f8
+size 2683904

neuronxcc-2.17.194.0+d312836f/MODULE_cebb33a758d78c3cffe9+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_cebb33a758d78c3cffe9+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b837fc2378fffecd859ba44115b41a9762865941d70d3065d0e414c758eaa73
+size 915725

neuronxcc-2.17.194.0+d312836f/MODULE_cebb33a758d78c3cffe9+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:948907db5be2d545502e15836f73101fd8dbc8a667afd4c7c692dd41963644d0
+size 11633664