| { | |
| "architectures": [ | |
| "SykoLLM_CMN" | |
| ], | |
| "model_type": "sykollm", | |
| "vocab_size": 32000, | |
| "d_model": 768, | |
| "n_heads": 6, | |
| "n_layers": 24, | |
| "chunk_size": 128, | |
| "intermediate_size": 3072, | |
| "num_memory_tokens": 16, | |
| "context_size": 1024, | |
| "overlap_size": 16, | |
| "code_overlap_size": 64, | |
| "abstract_head_hidden": 256, | |
| "abstract_head_layers": 2, | |
| "token_loss_weight": 0.7, | |
| "abstract_loss_weight": 0.3, | |
| "bos_token_id": 2, | |
| "eos_token_id": 3, | |
| "pad_token_id": 0 | |
| } |