| { | |
| "vocab_size": 16384, | |
| "context_length": 256, | |
| "embedding_dim": 640, | |
| "num_layers": 6, | |
| "num_heads": 10, | |
| "ffn_multiplier": 4, | |
| "dropout": 0.1, | |
| "weight_tying": false | |
| } |
| { | |
| "vocab_size": 16384, | |
| "context_length": 256, | |
| "embedding_dim": 640, | |
| "num_layers": 6, | |
| "num_heads": 10, | |
| "ffn_multiplier": 4, | |
| "dropout": 0.1, | |
| "weight_tying": false | |
| } |