| { | |
| "architectures": [ | |
| "ElasticGPT" | |
| ], | |
| "model_type": "elastic", | |
| "vocab_size": 16000, | |
| "max_seq_len": 1024, | |
| "hidden_size": 512, | |
| "num_layers": 10, | |
| "num_heads": 4, | |
| "head_dim": 128, | |
| "tie_embeddings": true, | |
| "use_rope": true, | |
| "qk_norm": true, | |
| "use_positional_encoding": true, | |
| "pad_token_id": 0, | |
| "bos_token_id": 1, | |
| "eos_token_id": 1 | |
| } |