| { | |
| "decoder_dropout": 0.1, | |
| "embedding_dropout": 0.1, | |
| "max_sequence_length": 1331, | |
| "model_dim": 768, | |
| "num_classes": 20000, | |
| "num_heads": 6, | |
| "num_layers": 6 | |
| } |
| { | |
| "decoder_dropout": 0.1, | |
| "embedding_dropout": 0.1, | |
| "max_sequence_length": 1331, | |
| "model_dim": 768, | |
| "num_classes": 20000, | |
| "num_heads": 6, | |
| "num_layers": 6 | |
| } |