Hugging Face's logo

aws-neuron
/
optimum-neuron-cache

Model card Files
xet
Community
629
dacorvo HF Staff commited on
Commit
63d95b9
·
verified ·
1 Parent(s): 5abd0f8

Delete neuronxcc-2.13.66.0+6dfecc895

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.20/training/bert/bert-base-uncased/1598246794078531d79e.json +0 -1
  2. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.20/training/bert/bert-base-uncased/8d09821f6f2cdea95b24.json +0 -1
  3. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.20/training/llama/meta-llama/Llama-2-7b-chat-hf/d403aed4c729665ab5f5.json +0 -1
  4. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/gpt2/gpt2/36e1657c052cd92f031f.json +0 -1
  5. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/gpt2/openai-community/gpt2/3ec20209d69016be9fbb.json +0 -1
  6. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/gpt2/openai-community/gpt2/7b39be0d0672ee1e02f4.json +0 -1
  7. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/gpt2/openai-community/gpt2/864bc63eecc707804124.json +0 -1
  8. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/gpt2/openai-community/gpt2/cc7cda3941881c035cb2.json +0 -1
  9. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/gpt2/openai-community/gpt2/e31ae96d70b5049c0036.json +0 -1
  10. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/01-ai/Yi-34B-200K/5b66966876b95b5064ff.json +0 -1
  11. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/01-ai/Yi-34B-200K/a5537a99bd14f7123d71.json +0 -1
  12. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/HuggingFaceTB/cosmo-1b/b394abcd6529110edc95.json +0 -1
  13. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/HuggingFaceTB/cosmo-1b/e57ca9bc6f160c36bb39.json +0 -1
  14. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/LargeWorldModel/LWM-Text-Chat-1M/3c4003d9da8254b2069d.json +0 -1
  15. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/LargeWorldModel/LWM-Text-Chat-1M/aaaef72dfbb3c4f226c1.json +0 -1
  16. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/NousResearch/Llama-2-13b-hf/0ff25bcc1d48309d50a5.json +0 -1
  17. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/NousResearch/Llama-2-13b-hf/13086a3eb20e43924921.json +0 -1
  18. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/NousResearch/Llama-2-13b-hf/144f7ce7049532e83c52.json +0 -1
  19. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/NousResearch/Llama-2-13b-hf/26b2777e2fc012769dec.json +0 -1
  20. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/NousResearch/Llama-2-13b-hf/d4009ea91ad0897960a3.json +0 -1
  21. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/abacusai/Smaug-72B-v0.1/31a27187f4e1a238da7a.json +0 -1
  22. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/abacusai/Smaug-72B-v0.1/d02cdab26f3dd900edbe.json +0 -1
  23. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/defog/sqlcoder-7b-2/08b79e703c61b2f851cc.json +0 -1
  24. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/defog/sqlcoder-7b-2/a433356ad4b14d82a632.json +0 -1
  25. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/gorilla-llm/gorilla-openfunctions-v2/17974dd7f23264e06557.json +0 -1
  26. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/gorilla-llm/gorilla-openfunctions-v2/29c007fc008352029be8.json +0 -1
  27. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/ibm/labradorite-13b/28cb90229ea727be3410.json +0 -1
  28. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/ibm/labradorite-13b/688e08bfe1af22e247b7.json +0 -1
  29. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/ibm/labradorite-13b/79d7ed3c93cbdb601301.json +0 -1
  30. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/m-a-p/ChatMusician/7fb626f961a26ca04b1d.json +0 -1
  31. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/m-a-p/ChatMusician/ca58f8583b95a120b566.json +0 -1
  32. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/m-a-p/OpenCodeInterpreter-DS-6.7B/0b49b0d495e75121beb7.json +0 -1
  33. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/m-a-p/OpenCodeInterpreter-DS-6.7B/331556dfa1f920d3e73b.json +0 -1
  34. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/03b6c7125f076563bd68.json +0 -1
  35. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/0b9e96edfb00f0f41781.json +0 -1
  36. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/2025960152bb1331bcc8.json +0 -1
  37. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/2311cc196a0f70044562.json +0 -1
  38. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/28dc6684e8ffd1957fcc.json +0 -1
  39. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/2e33fb05af9eda2154f6.json +0 -1
  40. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/4e99ef3d07dc56a0f446.json +0 -1
  41. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/66c4d16781d41c4da7f7.json +0 -1
  42. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/8bf98c2e52fcfe14c5d7.json +0 -1
  43. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/9a378c29a33a95a92f36.json +0 -1
  44. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/baaf1893070cf034975e.json +0 -1
  45. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/da44f10585a29a3d996c.json +0 -1
  46. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/e9e9e0fa42e555b34152.json +0 -1
  47. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-70b-chat-hf/1c0ffc384d07e27fbe8a.json +0 -1
  48. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-70b-chat-hf/8d49601d4e2484beb8d2.json +0 -1
  49. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-7b-chat-hf/09937949b9feb216648b.json +0 -1
  50. neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-7b-chat-hf/13cbb9ec2234e7ca3c26.json +0 -1
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.20/training/bert/bert-base-uncased/1598246794078531d79e.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["BertForMaskedLM"], "pad_token_id": 0, "gradient_checkpointing": false, "model_type": "bert", "vocab_size": 30522, "hidden_size": 768, "num_hidden_layers": 12, "num_attention_heads": 12, "hidden_act": "gelu", "intermediate_size": 3072, "hidden_dropout_prob": 0.1, "attention_probs_dropout_prob": 0.1, "max_position_embeddings": 512, "type_vocab_size": 2, "initializer_range": 0.02, "layer_norm_eps": 1e-12, "position_embedding_type": "absolute", "use_cache": true, "classifier_dropout": null, "neuron": {"model_class": "BertForSequenceClassification", "precision": "bfloat16", "num_neuron_cores_per_node": 2, "compiler_version": "2.13.66.0+6dfecc895", "tensor_parallel_size": 1, "pipeline_parallel_size": 1, "training": true, "input_specs": {"input_ids": [4, 256], "token_type_ids": [4, 256], "attention_mask": [4, 256], "labels": [4]}}}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.20/training/bert/bert-base-uncased/8d09821f6f2cdea95b24.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["BertForMaskedLM"], "pad_token_id": 0, "gradient_checkpointing": false, "model_type": "bert", "vocab_size": 30522, "hidden_size": 768, "num_hidden_layers": 12, "num_attention_heads": 12, "hidden_act": "gelu", "intermediate_size": 3072, "hidden_dropout_prob": 0.1, "attention_probs_dropout_prob": 0.1, "max_position_embeddings": 512, "type_vocab_size": 2, "initializer_range": 0.02, "layer_norm_eps": 1e-12, "position_embedding_type": "absolute", "use_cache": true, "classifier_dropout": null, "neuron": {"model_class": "BertForSequenceClassification", "precision": "bfloat16", "num_neuron_cores_per_node": 2, "compiler_version": "2.13.66.0+6dfecc895", "tensor_parallel_size": 1, "pipeline_parallel_size": 1, "training": false, "input_specs": {"input_ids": [3, 256], "token_type_ids": [3, 256], "attention_mask": [3, 256], "labels": [3]}}}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.20/training/llama/meta-llama/Llama-2-7b-chat-hf/d403aed4c729665ab5f5.json DELETED
@@ -1 +0,0 @@
1
- {"vocab_size": 32000, "max_position_embeddings": 4096, "hidden_size": 4096, "intermediate_size": 11008, "num_hidden_layers": 32, "num_attention_heads": 32, "num_key_value_heads": 32, "hidden_act": "silu", "initializer_range": 0.02, "rms_norm_eps": 1e-05, "pretraining_tp": 1, "use_cache": true, "rope_theta": 10000.0, "rope_scaling": null, "attention_bias": false, "attention_dropout": 0.0, "torch_dtype": "float16", "tie_word_embeddings": false, "architectures": ["LlamaForCausalLM"], "bos_token_id": 1, "eos_token_id": 2, "model_type": "llama", "neuron": {"model_class": "LlamaForCausalLM", "precision": "bfloat16", "num_neuron_cores_per_node": 32, "compiler_version": "2.13.66.0+6dfecc895", "tensor_parallel_size": 8, "pipeline_parallel_size": 1, "training": false, "input_specs": {"input_ids": [2, 512], "attention_mask": [2, 512], "labels": [2, 512]}}}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/gpt2/gpt2/36e1657c052cd92f031f.json DELETED
@@ -1 +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": "fp32", "batch_size": 4, "checkpoint_id": "gpt2", "checkpoint_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/gpt2/openai-community/gpt2/3ec20209d69016be9fbb.json DELETED
@@ -1 +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": 1, "checkpoint_id": "openai-community/gpt2", "checkpoint_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 1, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/gpt2/openai-community/gpt2/7b39be0d0672ee1e02f4.json DELETED
@@ -1 +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": 16, "checkpoint_id": "openai-community/gpt2", "checkpoint_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 1, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/gpt2/openai-community/gpt2/864bc63eecc707804124.json DELETED
@@ -1 +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": "openai-community/gpt2", "checkpoint_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/gpt2/openai-community/gpt2/cc7cda3941881c035cb2.json DELETED
@@ -1 +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": 16, "checkpoint_id": "openai-community/gpt2", "checkpoint_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/gpt2/openai-community/gpt2/e31ae96d70b5049c0036.json DELETED
@@ -1 +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": 1, "checkpoint_id": "openai-community/gpt2", "checkpoint_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/01-ai/Yi-34B-200K/5b66966876b95b5064ff.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 7168, "initializer_range": 0.02, "intermediate_size": 20480, "max_position_embeddings": 200000, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "01-ai/Yi-34B-200K", "checkpoint_revision": "4277a28017f0e631526d3a59d50dfcd3f583fb30", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 24, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 56, "num_hidden_layers": 60, "num_key_value_heads": 8, "pad_token_id": 0, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 64000}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/01-ai/Yi-34B-200K/a5537a99bd14f7123d71.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 7168, "initializer_range": 0.02, "intermediate_size": 20480, "max_position_embeddings": 200000, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "01-ai/Yi-34B-200K", "checkpoint_revision": "4277a28017f0e631526d3a59d50dfcd3f583fb30", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 24, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 56, "num_hidden_layers": 60, "num_key_value_heads": 8, "pad_token_id": 0, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 64000}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/HuggingFaceTB/cosmo-1b/b394abcd6529110edc95.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 8192, "max_position_embeddings": 2048, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "HuggingFaceTB/cosmo-1b", "checkpoint_revision": "0d5e341cfe835dffc81b6186f9715c094889f8ce", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 16, "num_hidden_layers": 24, "num_key_value_heads": 16, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float32", "use_cache": true, "vocab_size": 32000}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/HuggingFaceTB/cosmo-1b/e57ca9bc6f160c36bb39.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 8192, "max_position_embeddings": 2048, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "HuggingFaceTB/cosmo-1b", "checkpoint_revision": "0d5e341cfe835dffc81b6186f9715c094889f8ce", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 16, "num_hidden_layers": 24, "num_key_value_heads": 16, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float32", "use_cache": true, "vocab_size": 32000}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/LargeWorldModel/LWM-Text-Chat-1M/3c4003d9da8254b2069d.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 1048576, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "LargeWorldModel/LWM-Text-Chat-1M", "checkpoint_revision": "0598c443b02aeb1a1f9f6788e9af85ea762a452d", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 32, "pad_token_id": 0, "pretraining_tp": 1, "rms_norm_eps": 1e-06, "rope_scaling": null, "rope_theta": 50000000.0, "tie_word_embeddings": false, "torch_dtype": "float16", "use_cache": true, "vocab_size": 32000}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/LargeWorldModel/LWM-Text-Chat-1M/aaaef72dfbb3c4f226c1.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 1048576, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "LargeWorldModel/LWM-Text-Chat-1M", "checkpoint_revision": "0598c443b02aeb1a1f9f6788e9af85ea762a452d", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 32, "pad_token_id": 0, "pretraining_tp": 1, "rms_norm_eps": 1e-06, "rope_scaling": null, "rope_theta": 50000000.0, "tie_word_embeddings": false, "torch_dtype": "float16", "use_cache": true, "vocab_size": 32000}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/NousResearch/Llama-2-13b-hf/0ff25bcc1d48309d50a5.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "NousResearch/Llama-2-13b-hf", "checkpoint_revision": "b0491461253755d8c60bf22f0d696b9e337c6375", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 24, "sequence_length": 2048, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "pad_token_id": 0, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/NousResearch/Llama-2-13b-hf/13086a3eb20e43924921.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "NousResearch/Llama-2-13b-hf", "checkpoint_revision": "b0491461253755d8c60bf22f0d696b9e337c6375", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 12, "sequence_length": 2048, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "pad_token_id": 0, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/NousResearch/Llama-2-13b-hf/144f7ce7049532e83c52.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "NousResearch/Llama-2-13b-hf", "checkpoint_revision": "b0491461253755d8c60bf22f0d696b9e337c6375", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 8, "sequence_length": 1024, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "pad_token_id": 0, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/NousResearch/Llama-2-13b-hf/26b2777e2fc012769dec.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "NousResearch/Llama-2-13b-hf", "checkpoint_revision": "b0491461253755d8c60bf22f0d696b9e337c6375", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 16, "sequence_length": 1024, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "pad_token_id": 0, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/NousResearch/Llama-2-13b-hf/d4009ea91ad0897960a3.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "NousResearch/Llama-2-13b-hf", "checkpoint_revision": "b0491461253755d8c60bf22f0d696b9e337c6375", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 4, "sequence_length": 1024, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "pad_token_id": 0, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/abacusai/Smaug-72B-v0.1/31a27187f4e1a238da7a.json DELETED
@@ -1 +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, "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": "2.13.66.0+6dfecc895", "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/abacusai/Smaug-72B-v0.1/d02cdab26f3dd900edbe.json DELETED
@@ -1 +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, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "abacusai/Smaug-72B-v0.1", "checkpoint_revision": "a1d657156f82c24b670158406378648233487011", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/defog/sqlcoder-7b-2/08b79e703c61b2f851cc.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 16384, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "defog/sqlcoder-7b-2", "checkpoint_revision": "7e5b6f7981c0aa7d143f6bec6fa26625bdfcbe66", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "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": 1000000, "tie_word_embeddings": false, "torch_dtype": "float16", "use_cache": true, "vocab_size": 32016}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/defog/sqlcoder-7b-2/a433356ad4b14d82a632.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 16384, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "defog/sqlcoder-7b-2", "checkpoint_revision": "7e5b6f7981c0aa7d143f6bec6fa26625bdfcbe66", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "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": 1000000, "tie_word_embeddings": false, "torch_dtype": "float16", "use_cache": true, "vocab_size": 32016}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/gorilla-llm/gorilla-openfunctions-v2/17974dd7f23264e06557.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 100000, "eos_token_id": 100015, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "gorilla-llm/gorilla-openfunctions-v2", "checkpoint_revision": "0f91d705e64b77fb55e35a7eab5d03bf965c9b5c", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 30, "num_key_value_heads": 32, "pretraining_tp": 1, "rms_norm_eps": 1e-06, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": false, "vocab_size": 102400}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/gorilla-llm/gorilla-openfunctions-v2/29c007fc008352029be8.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 100000, "eos_token_id": 100015, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "gorilla-llm/gorilla-openfunctions-v2", "checkpoint_revision": "0f91d705e64b77fb55e35a7eab5d03bf965c9b5c", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 30, "num_key_value_heads": 32, "pretraining_tp": 1, "rms_norm_eps": 1e-06, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": false, "vocab_size": 102400}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/ibm/labradorite-13b/28cb90229ea727be3410.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "end_token_id": 32000, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "ibm/labradorite-13b", "checkpoint_revision": "26921cf1d63c9c657e8260d08eaacc14fca43da0", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 8, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "pad_token_id": 32001, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 32008}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/ibm/labradorite-13b/688e08bfe1af22e247b7.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "end_token_id": 32000, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "ibm/labradorite-13b", "checkpoint_revision": "26921cf1d63c9c657e8260d08eaacc14fca43da0", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 8, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "pad_token_id": 32001, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 32008}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/ibm/labradorite-13b/79d7ed3c93cbdb601301.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "end_token_id": 32000, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 8, "checkpoint_id": "ibm/labradorite-13b", "checkpoint_revision": "26921cf1d63c9c657e8260d08eaacc14fca43da0", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 8, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "pad_token_id": 32001, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 32008}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/m-a-p/ChatMusician/7fb626f961a26ca04b1d.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 2048, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "m-a-p/ChatMusician", "checkpoint_revision": "cd8064911f26cd6701561eb733319adfa88383df", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 32, "pad_token_id": 0, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/m-a-p/ChatMusician/ca58f8583b95a120b566.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 2048, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "m-a-p/ChatMusician", "checkpoint_revision": "cd8064911f26cd6701561eb733319adfa88383df", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 32, "pad_token_id": 0, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/m-a-p/OpenCodeInterpreter-DS-6.7B/0b49b0d495e75121beb7.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 32013, "eos_token_id": 32021, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 16384, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "m-a-p/OpenCodeInterpreter-DS-6.7B", "checkpoint_revision": "60b89884df814590abd76757a6db4a527cbdfc91", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 32, "pretraining_tp": 1, "rms_norm_eps": 1e-06, "rope_scaling": {"factor": 4.0, "type": "linear"}, "rope_theta": 100000, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": false, "vocab_size": 32256}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/m-a-p/OpenCodeInterpreter-DS-6.7B/331556dfa1f920d3e73b.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 32013, "eos_token_id": 32021, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 16384, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "m-a-p/OpenCodeInterpreter-DS-6.7B", "checkpoint_revision": "60b89884df814590abd76757a6db4a527cbdfc91", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 32, "pretraining_tp": 1, "rms_norm_eps": 1e-06, "rope_scaling": {"factor": 4.0, "type": "linear"}, "rope_theta": 100000, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": false, "vocab_size": 32256}
 
 
neuronxcc-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/03b6c7125f076563bd68.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 32, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 12, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/0b9e96edfb00f0f41781.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 8, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/2025960152bb1331bcc8.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 8, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 12, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/2311cc196a0f70044562.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 8, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 8, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/28dc6684e8ffd1957fcc.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 16, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 12, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/2e33fb05af9eda2154f6.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 8, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 24, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/4e99ef3d07dc56a0f446.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 32, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 24, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/66c4d16781d41c4da7f7.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 24, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/8bf98c2e52fcfe14c5d7.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 12, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/9a378c29a33a95a92f36.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 12, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/baaf1893070cf034975e.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 8, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/da44f10585a29a3d996c.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 24, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-13b-chat-hf/e9e9e0fa42e555b34152.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 5120, "initializer_range": 0.02, "intermediate_size": 13824, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 16, "checkpoint_id": "meta-llama/Llama-2-13b-chat-hf", "checkpoint_revision": "29655417e51232f4f2b9b5d3e1418e5a9b04e80e", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 24, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 40, "num_hidden_layers": 40, "num_key_value_heads": 40, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-70b-chat-hf/1c0ffc384d07e27fbe8a.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 8192, "initializer_range": 0.02, "intermediate_size": 28672, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 4, "checkpoint_id": "meta-llama/Llama-2-70b-chat-hf", "checkpoint_revision": "8b17e6f4e86be78cf54afd49ddb517d4e274c13f", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 24, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 64, "num_hidden_layers": 80, "num_key_value_heads": 8, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-70b-chat-hf/8d49601d4e2484beb8d2.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 8192, "initializer_range": 0.02, "intermediate_size": 28672, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "meta-llama/Llama-2-70b-chat-hf", "checkpoint_revision": "8b17e6f4e86be78cf54afd49ddb517d4e274c13f", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 24, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 64, "num_hidden_layers": 80, "num_key_value_heads": 8, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-7b-chat-hf/09937949b9feb216648b.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 1, "checkpoint_id": "meta-llama/Llama-2-7b-chat-hf", "checkpoint_revision": "92011f62d7604e261f748ec0cfe6329f31193e33", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 2, "sequence_length": 4096, "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-2.13.66.0+6dfecc895/0_REGISTRY/0.0.21/inference/llama/meta-llama/Llama-2-7b-chat-hf/13cbb9ec2234e7ca3c26.json DELETED
@@ -1 +0,0 @@
1
- {"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, "max_position_embeddings": 4096, "model_type": "llama", "neuron": {"auto_cast_type": "fp16", "batch_size": 8, "checkpoint_id": "meta-llama/Llama-2-7b-chat-hf", "checkpoint_revision": "92011f62d7604e261f748ec0cfe6329f31193e33", "compiler_type": "neuronx-cc", "compiler_version": "2.13.66.0+6dfecc895", "num_cores": 24, "sequence_length": 4096, "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}