Synchronizing local compiler cache.

.gitattributes

@@ -6544,3 +6544,19 @@ neuronxcc-2.15.128.0+56dc5a86/MODULE_1670378662283822540+0662544f/model.neff fil
 neuronxcc-2.15.128.0+56dc5a86/MODULE_17680146082376704246+0662544f/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_7546651762822850137+0662544f/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_8908022223708495108+0662544f/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_1f6cff7b8c7ee3876236+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_2fafc1ff132d5f941aad+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_3af7e7ef3791e98d3c6d+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_46e098601535b4170185+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_4875ee13adf7ccce2fd8+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_4b5c886cd13ba7d10be9+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_4f8465ac364975d34bcb+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_640ad25f4ab9e0ce4ac7+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_722d3c4ebc79e2bbd531+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_b420f7ff165f4f0517c3+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_bb86f1f35879e84a7798+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_c0b96836ba4c3a46f1d1+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_c7fee5e6e0f12889c294+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_f63ff654073837167064+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_fcdf5b6858bafa453910+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.143.0+e39249ad/MODULE_ff275b5606db64ceaea0+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.15.143.0+e39249ad/0_REGISTRY/0.0.27.dev0/inference/granite/ibm-granite/granite-3.0-8b-instruct/89417199e26022010c7a.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": 4096, "mlp_bias": false, "model_type": "granite", "neuron": {"auto_cast_type": "bf16", "batch_size": 4, "checkpoint_id": "ibm-granite/granite-3.0-8b-instruct", "checkpoint_revision": "8fe1e202a17f7763bd0af471253e00cc846d1c05", "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": 10000, "tie_word_embeddings": true, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 49155}

neuronxcc-2.15.143.0+e39249ad/0_REGISTRY/0.0.28.dev0/inference/gpt2/gpt2/aadac442fccfdd730bb5.json

@@ -0,0 +1 @@
+{"activation_function": "gelu_new", "architectures": ["GPT2LMHeadModel"], "attn_pdrop": 0.1, "bos_token_id": 50256, "embd_pdrop": 0.1, "eos_token_id": 50256, "initializer_range": 0.02, "layer_norm_epsilon": 1e-05, "model_type": "gpt2", "n_ctx": 1024, "n_embd": 768, "n_head": 12, "n_inner": null, "n_layer": 12, "n_positions": 1024, "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "gpt2", "checkpoint_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "compiler_type": "neuronx-cc", "compiler_version": "2.15.143.0+e39249ad", "num_cores": 2, "sequence_length": 1024, "task": "text-generation"}, "reorder_and_upcast_attn": false, "resid_pdrop": 0.1, "scale_attn_by_inverse_layer_idx": false, "scale_attn_weights": true, "summary_activation": null, "summary_first_dropout": 0.1, "summary_proj_to_labels": true, "summary_type": "cls_index", "summary_use_proj": true, "task_specific_params": {"text-generation": {"do_sample": true, "max_length": 50}}, "use_cache": true, "vocab_size": 50257}

neuronxcc-2.15.143.0+e39249ad/0_REGISTRY/0.0.28.dev0/inference/granite/ibm-granite/granite-3.1-2b-instruct/5861522c9be356156fd7.json

@@ -0,0 +1 @@
+{"architectures": ["GraniteForCausalLM"], "attention_bias": false, "attention_dropout": 0.1, "attention_multiplier": 0.015625, "bos_token_id": 0, "embedding_multiplier": 12.0, "eos_token_id": 0, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 8192, "logits_scaling": 8.0, "max_position_embeddings": 131072, "mlp_bias": false, "model_type": "granite", "neuron": {"auto_cast_type": "bf16", "batch_size": 4, "checkpoint_id": "ibm-granite/granite-3.1-2b-instruct", "checkpoint_revision": "a06c9a0fef05d8111ca1f77b60f477443c526043", "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": 5000000.0, "tie_word_embeddings": true, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 49155}

neuronxcc-2.15.143.0+e39249ad/0_REGISTRY/0.0.28.dev0/inference/granite/ibm-granite/granite-3.1-2b-instruct/a6c9d0b4d634131c86b8.json

@@ -0,0 +1 @@
+{"architectures": ["GraniteForCausalLM"], "attention_bias": false, "attention_dropout": 0.1, "attention_multiplier": 0.015625, "bos_token_id": 0, "embedding_multiplier": 12.0, "eos_token_id": 0, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 8192, "logits_scaling": 8.0, "max_position_embeddings": 131072, "mlp_bias": false, "model_type": "granite", "neuron": {"auto_cast_type": "bf16", "batch_size": 1, "checkpoint_id": "ibm-granite/granite-3.1-2b-instruct", "checkpoint_revision": "a06c9a0fef05d8111ca1f77b60f477443c526043", "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": 5000000.0, "tie_word_embeddings": true, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 49155}

neuronxcc-2.15.143.0+e39249ad/0_REGISTRY/0.0.28.dev0/inference/llama/NousResearch/Hermes-2-Theta-Llama-3-8B/c1effc43754275f46f73.json

@@ -0,0 +1 @@
+{"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 128000, "eos_token_id": 128003, "head_dim": 128, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 14336, "max_position_embeddings": 8192, "mlp_bias": false, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "NousResearch/Hermes-2-Theta-Llama-3-8B", "checkpoint_revision": "57a73110702e7b05ba3f39fef36297454c680725", "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": 32, "num_key_value_heads": 8, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 500000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 128256}

neuronxcc-2.15.143.0+e39249ad/0_REGISTRY/0.0.28.dev0/inference/mistral/optimum/mistral-1.1b-testing/37aed4e6734aa8542ee8.json

@@ -0,0 +1 @@
+{"architectures": ["MistralForCausalLM"], "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 64, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 5632, "max_position_embeddings": 32768, "model_type": "mistral", "neuron": {"auto_cast_type": "bf16", "batch_size": 4, "checkpoint_id": "optimum/mistral-1.1b-testing", "checkpoint_revision": "ce03bc8d47dbd2c173ff65f3a8de1325ba724195", "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": 22, "num_key_value_heads": 4, "rms_norm_eps": 1e-05, "rope_theta": 1000000.0, "sliding_window": null, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 32000}

neuronxcc-2.15.143.0+e39249ad/0_REGISTRY/0.0.28.dev0/inference/qwen2/Qwen/Qwen2.5-0.5B/a71f9f96f5754288c758.json

@@ -0,0 +1 @@
+{"architectures": ["Qwen2ForCausalLM"], "attention_dropout": 0.0, "bos_token_id": 151643, "eos_token_id": 151643, "hidden_act": "silu", "hidden_size": 896, "initializer_range": 0.02, "intermediate_size": 4864, "max_position_embeddings": 32768, "max_window_layers": 24, "model_type": "qwen2", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "Qwen/Qwen2.5-0.5B", "checkpoint_revision": "060db6499f32faf8b98477b0a26969ef7d8b9987", "compiler_type": "neuronx-cc", "compiler_version": "2.15.143.0+e39249ad", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 14, "num_hidden_layers": 24, "num_key_value_heads": 2, "rms_norm_eps": 1e-06, "rope_scaling": null, "rope_theta": 1000000.0, "sliding_window": null, "tie_word_embeddings": true, "torch_dtype": "bfloat16", "use_cache": true, "use_mrope": false, "use_sliding_window": false, "vocab_size": 151936}

neuronxcc-2.15.143.0+e39249ad/MODULE_0321b5b02683246a4daf+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_0321b5b02683246a4daf+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cb5b751bac3098f37be0e0913868ba23f9012c99c7b8645b115b0b96a79f956
+size 45952

neuronxcc-2.15.143.0+e39249ad/MODULE_0728e5a2562898fb6304+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_0728e5a2562898fb6304+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:947cf33aaeb5d03ce1e8ae9975578e3cfb1c54807ba3fb44e84a9be3f02eece4
+size 66329

neuronxcc-2.15.143.0+e39249ad/MODULE_08029859a80afc7e62d3+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_08029859a80afc7e62d3+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:addd1e24b6a8cb2a88bdab82060ebda6b3648021b7b9667bfdcad841306fcb17
+size 41445

neuronxcc-2.15.143.0+e39249ad/MODULE_0af38ef912cf5f22e544+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_0af38ef912cf5f22e544+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c6ea47f92cefbcb20665c27d11df6da0b45afb22514a04211d829a6b0877c68
+size 69161

neuronxcc-2.15.143.0+e39249ad/MODULE_12a0d260bd97c0c503b9+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_12a0d260bd97c0c503b9+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b46c9c744d8f649bd2fd1e82572fdb16910ead866713ac326600bc9da7f47a6e
+size 44866

neuronxcc-2.15.143.0+e39249ad/MODULE_12d9000464abb773502a+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_12d9000464abb773502a+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba1a30ce3421122641a40517673aff5b38b757e500824cc20c9b35832a37a39b
+size 55044

neuronxcc-2.15.143.0+e39249ad/MODULE_12f2d7e7596ea62b3de1+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_12f2d7e7596ea62b3de1+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b409d8245e57378632c1f66a90e3c4b98b0bfc8aeb90843123a87a69ce4af817
+size 20632

neuronxcc-2.15.143.0+e39249ad/MODULE_1f6cff7b8c7ee3876236+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_1f6cff7b8c7ee3876236+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea2767f6783bc33ee98b300d892e7df1fc9eb03c9a353bef26fd2d5c9a0b297a
+size 426108

neuronxcc-2.15.143.0+e39249ad/MODULE_1f6cff7b8c7ee3876236+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:762b9be3f390373aa44c767c26c294ccc69e31cece925752f79fc0a0b311847a
+size 3124224

neuronxcc-2.15.143.0+e39249ad/MODULE_220ccbe10e6e8be33dbf+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_220ccbe10e6e8be33dbf+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2771652ffc854d34036171b43e8ec3c7c08c54d9b51fb66e026c3e7451b27879
+size 54675

neuronxcc-2.15.143.0+e39249ad/MODULE_2fafc1ff132d5f941aad+39f12043/compile_flags.json

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

neuronxcc-2.15.143.0+e39249ad/MODULE_2fafc1ff132d5f941aad+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d504df9721933b6e4428c1410d737f8d46e48cfb3fa62d489e57324907a9fe6b
+size 332573

neuronxcc-2.15.143.0+e39249ad/MODULE_2fafc1ff132d5f941aad+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af2f9e2ad514b97b416bb6aef9e34afe3d9274d4699802b2151053d474f7499a
+size 10845184

neuronxcc-2.15.143.0+e39249ad/MODULE_3af7e7ef3791e98d3c6d+39f12043/compile_flags.json

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