Synchronizing local compiler cache.

.gitattributes

@@ -4848,3 +4848,5 @@ neuronxcc-2.15.128.0+56dc5a86/MODULE_bfbe3efe685c32920201+39f12043/model.neff fi
 neuronxcc-2.15.128.0+56dc5a86/MODULE_cb6145854d51552ad2ae+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_e7bd7a2134c7496f5ce9+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_f8c98d89b4c3276b429e+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_02e53ffc6085b74dd1c4+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_e820688eff57f3ac8212+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.15.128.0+56dc5a86/0_REGISTRY/0.0.25.dev0/inference/llama/abacusai/Smaug-72B-v0.1/58cf755a61876f0d9cbd.json

@@ -0,0 +1 @@
+{"architectures": ["LlamaForCausalLM"], "attention_bias": true, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 151643, "hidden_act": "silu", "hidden_size": 8192, "initializer_range": 0.02, "intermediate_size": 24576, "max_position_embeddings": 32768, "mlp_bias": false, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "abacusai/Smaug-72B-v0.1", "checkpoint_revision": "a1d657156f82c24b670158406378648233487011", "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": 64, "pad_token_id": 151643, "pretraining_tp": 1, "rms_norm_eps": 1e-06, "rope_scaling": null, "rope_theta": 1000000, "seq_length": 32768, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 152064}

neuronxcc-2.15.128.0+56dc5a86/MODULE_02e53ffc6085b74dd1c4+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_02e53ffc6085b74dd1c4+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:838cf9a1dfcc6f8f3706d073353da3fa6de08e08a507f316ce3861440bc9b3ba
+size 880597

neuronxcc-2.15.128.0+56dc5a86/MODULE_02e53ffc6085b74dd1c4+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb59a9a22d10b32881e2dc744d29c1427b4a99be60d75121683707ac08ad6231
+size 10814464

neuronxcc-2.15.128.0+56dc5a86/MODULE_e820688eff57f3ac8212+39f12043/compile_flags.json

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

neuronxcc-2.15.128.0+56dc5a86/MODULE_e820688eff57f3ac8212+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45ff713a7e13b345abec835a6540adf65fef6fb38da81d1e4c3d78f052af7967
+size 1122182

neuronxcc-2.15.128.0+56dc5a86/MODULE_e820688eff57f3ac8212+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d064838e49d48e6058deda2cd652fdd98bfd54b5cc8d9ed509267b8e5c60dce
+size 2561024