Synchronizing local compiler cache.

.gitattributes

@@ -7163,3 +7163,5 @@ neuronxcc-2.16.372.0+4a9b2326/MODULE_324968d40a0692640f69/text_encoder/model.neu
 neuronxcc-2.16.372.0+4a9b2326/MODULE_324968d40a0692640f69/unet/model.neuron filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.16.372.0+4a9b2326/MODULE_324968d40a0692640f69/vae_decoder/model.neuron filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.16.372.0+4a9b2326/MODULE_324968d40a0692640f69/vae_encoder/model.neuron filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.16.372.0+4a9b2326/MODULE_1b6d7576802d84bea017+613edded/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.16.372.0+4a9b2326/MODULE_c578025927b993537ac9+613edded/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.16.372.0+4a9b2326/0_REGISTRY/0.1.0.dev1/inference/llama/meta-llama/Llama-2-7b-hf/678b8395e743df567a53.json

@@ -0,0 +1 @@
+{"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "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": "2.16.372.0+4a9b2326", "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, "torch_dtype": "float16", "use_cache": true, "vocab_size": 32000}

neuronxcc-2.16.372.0+4a9b2326/MODULE_1b6d7576802d84bea017+613edded/compile_flags.json

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

neuronxcc-2.16.372.0+4a9b2326/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-2.16.372.0+4a9b2326/MODULE_1b6d7576802d84bea017+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:306c29d117aba7afc1e1bd03397b743ae9eb6f2174f5834bc584eb3d38aeae5e
+size 3257344

neuronxcc-2.16.372.0+4a9b2326/MODULE_c578025927b993537ac9+613edded/compile_flags.json

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

neuronxcc-2.16.372.0+4a9b2326/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-2.16.372.0+4a9b2326/MODULE_c578025927b993537ac9+613edded/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1220c9245b0f397d12bd3cddaa79992d3e749278981e0f01034883ff105b25d9
+size 7015424