Synchronizing local compiler cache.

.gitattributes

@@ -7892,3 +7892,5 @@ neuronxcc-2.17.194.0+d312836f/MODULE_c4162c03a5bca6d7620e/text_encoder_2/model.n
 neuronxcc-2.17.194.0+d312836f/MODULE_c4162c03a5bca6d7620e/unet/model.neuron filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.17.194.0+d312836f/MODULE_c4162c03a5bca6d7620e/vae_decoder/model.neuron filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.17.194.0+d312836f/MODULE_c4162c03a5bca6d7620e/vae_encoder/model.neuron filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.17.194.0+d312836f/MODULE_1b6d7576802d84bea017+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.17.194.0+d312836f/MODULE_c578025927b993537ac9+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-2-7b-hf/d75e02defdb0d74b4773.json

@@ -0,0 +1,39 @@
+{
+  "_entry_class": "SingleModelCacheEntry",
+  "_model_id": "meta-llama/Llama-2-7b-hf",
+  "_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": 11008,
+  "max_position_embeddings": 4096,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "neuron": {
+    "auto_cast_type": "bf16",
+    "batch_size": 1,
+    "checkpoint_id": "meta-llama/Llama-2-7b-hf",
+    "checkpoint_revision": "01c7f73d771dfac7d292323805ebc428287df4f9",
+    "compiler_type": "neuronx-cc",
+    "compiler_version": "2.17.194.0+d312836f",
+    "num_cores": 2,
+    "sequence_length": 2048,
+    "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,
+  "use_cache": true,
+  "vocab_size": 32000
+}

neuronxcc-2.17.194.0+d312836f/MODULE_1b6d7576802d84bea017+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_1b6d7576802d84bea017+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f88cce4270594ff7789282fa8116c1bc85743bc9a5c375660f852053fd43d5f1
+size 244397

neuronxcc-2.17.194.0+d312836f/MODULE_1b6d7576802d84bea017+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af14da9d25db2bd55dc36d9a7b7f8f2391cb0098665ae34cfb7841d2e3c94f0e
+size 3329024

neuronxcc-2.17.194.0+d312836f/MODULE_c578025927b993537ac9+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_c578025927b993537ac9+613edded/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fcb186a7e5c50350864f29d2abb1ac641565637b2e807baf979ba16e7439ee1
+size 308060

neuronxcc-2.17.194.0+d312836f/MODULE_c578025927b993537ac9+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9e8216d4525804f0c33096f9d01a45d456c14668284dfc66cdea4cf4b9c6f6e
+size 7650304