| { | |
| "code_dim": 1024, | |
| "d_model": 1024, | |
| "n_layer": 8, | |
| "n_head": 16, | |
| "d_ff": 4096, | |
| "max_seq_len": 128, | |
| "dropout": 0.1, | |
| "stop_loss_weight": 1.0, | |
| "train_steps": 1000, | |
| "sample_steps": 20, | |
| "denoiser_layers": 4, | |
| "denoiser_hidden": 2048 | |
| } |
| { | |
| "code_dim": 1024, | |
| "d_model": 1024, | |
| "n_layer": 8, | |
| "n_head": 16, | |
| "d_ff": 4096, | |
| "max_seq_len": 128, | |
| "dropout": 0.1, | |
| "stop_loss_weight": 1.0, | |
| "train_steps": 1000, | |
| "sample_steps": 20, | |
| "denoiser_layers": 4, | |
| "denoiser_hidden": 2048 | |
| } |