Synchronizing local compiler cache.

.gitattributes

@@ -6576,3 +6576,5 @@ neuronxcc-2.15.143.0+e39249ad/MODULE_95f0ce72279e469fa4dc+39f12043/model.neff fi
 neuronxcc-2.15.143.0+e39249ad/MODULE_a42027513ff1c4e1a28b+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.143.0+e39249ad/MODULE_f37ba14e8cecd0289618+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.143.0+e39249ad/MODULE_f6695fce8439b8324614+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_0fff2decd536cc6fae0b+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_b60a34f4644d1f39a41c+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.15.143.0+e39249ad/0_REGISTRY/0.0.28.dev0/inference/granite/ibm-granite/granite-3.1-8b-instruct/a77e916953a3fe0abad7.json

@@ -0,0 +1 @@
+{"architectures": ["GraniteForCausalLM"], "attention_bias": false, "attention_dropout": 0.1, "attention_multiplier": 0.0078125, "bos_token_id": 0, "embedding_multiplier": 12.0, "eos_token_id": 0, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 12800, "logits_scaling": 16.0, "max_position_embeddings": 131072, "mlp_bias": false, "model_type": "granite", "neuron": {"auto_cast_type": "bf16", "batch_size": 4, "checkpoint_id": "ibm-granite/granite-3.1-8b-instruct", "checkpoint_revision": "7f4bb8b56ad1c53b49b07a8f9f0ec43204e07ef2", "compiler_type": "neuronx-cc", "compiler_version": "2.15.143.0+e39249ad", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 40, "num_key_value_heads": 8, "pad_token_id": 0, "residual_multiplier": 0.22, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000000.0, "tie_word_embeddings": true, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 49155}

neuronxcc-2.15.143.0+e39249ad/MODULE_0fff2decd536cc6fae0b+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_0fff2decd536cc6fae0b+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd1fd11eaf38deb423ff4f463a72f2f089e65252945fc35562df4d1f9982f2fc
+size 569348

neuronxcc-2.15.143.0+e39249ad/MODULE_0fff2decd536cc6fae0b+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29bf7c1dee57b0c2b7c30b4c8936d048f65a3e06be296f69fb6c448d028d9241
+size 5991424

neuronxcc-2.15.143.0+e39249ad/MODULE_b60a34f4644d1f39a41c+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_b60a34f4644d1f39a41c+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0bcfdcbad033c64c7d44c35d07dd6af92b6b242b591cc008929e6d285d34421f
+size 467919

neuronxcc-2.15.143.0+e39249ad/MODULE_b60a34f4644d1f39a41c+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:943f575da50c68f37d4e0a1d8d7e796b40cce4de9ae2460e6435ed8be1c98553
+size 11060224