| { | |
| "architectures": [ | |
| "ArcticLSTMSpeculatorPreTrainedModel" | |
| ], | |
| "base_model_name_or_path": "meta-llama/Llama-3.1-70B-Instruct", | |
| "base_model_archs": [ | |
| "LlamaForCausalLM", | |
| "LlamaSwiftKVForCausalLM" | |
| ], | |
| "input_hidden_dim": 8192, | |
| "inner_dim": "4096", | |
| "proj_dim": "4096", | |
| "emb_dim": "4096", | |
| "model_type": "mlp_speculator", | |
| "n_candidates": 3, | |
| "n_predict": 3, | |
| "scale_input": true, | |
| "tie_weights": true, | |
| "tie_lstm_embs": true, | |
| "top_k_tokens_per_head": [ | |
| 1, | |
| 1, | |
| 1 | |
| ], | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.47.0", | |
| "vocab_size": 128256, | |
| "method": "sum_lstm" | |
| } |