| { | |
| "vocab_size": 8192, | |
| "context_len": 12000, | |
| "n_embd": 768, | |
| "n_head": 12, | |
| "n_layer": 12, | |
| "attn_dropout": 0.1, | |
| "resid_dropout": 0.1, | |
| "embd_dropout": 0.1, | |
| "mlp_dropout": 0.1 | |
| } |
| { | |
| "vocab_size": 8192, | |
| "context_len": 12000, | |
| "n_embd": 768, | |
| "n_head": 12, | |
| "n_layer": 12, | |
| "attn_dropout": 0.1, | |
| "resid_dropout": 0.1, | |
| "embd_dropout": 0.1, | |
| "mlp_dropout": 0.1 | |
| } |