| { | |
| "vocab_size": 30522, | |
| "max_position_embeddings": 1024, | |
| "hidden_size": 512, | |
| "num_attention_heads": 8, | |
| "intermediate_size": 2048, | |
| "num_hidden_layers": 8, | |
| "dropout_rate": 0.1, | |
| "weight_decay": 0.01, | |
| "learning_rate": 0.0001, | |
| "restart_period": 500, | |
| "total_steps": 2500, | |
| "warmup_steps": 500 | |
| } |