| { | |
| "vocab_size": 118, | |
| "d_model": 128, | |
| "n_layers": 2, | |
| "n_heads": 4, | |
| "ff": 256, | |
| "max_len": 256, | |
| "dropout": 0.1, | |
| "pad_id": 0, | |
| "cls_id": 1, | |
| "sep_id": 2, | |
| "unk_id": 3, | |
| "model_type": "tiny-lm", | |
| "architecture": "TinyLM" | |
| } |
| { | |
| "vocab_size": 118, | |
| "d_model": 128, | |
| "n_layers": 2, | |
| "n_heads": 4, | |
| "ff": 256, | |
| "max_len": 256, | |
| "dropout": 0.1, | |
| "pad_id": 0, | |
| "cls_id": 1, | |
| "sep_id": 2, | |
| "unk_id": 3, | |
| "model_type": "tiny-lm", | |
| "architecture": "TinyLM" | |
| } |