| { | |
| "activation": "relu", | |
| "compile": false, | |
| "d_cond": 128, | |
| "d_model": 768, | |
| "dim_feedforward": 3072, | |
| "dropout": 0.1, | |
| "force_flash_attn": false, | |
| "inner_autocast": true, | |
| "layer_norm_eps": 1e-05, | |
| "mask_idx": 4, | |
| "max_length": 260, | |
| "max_rate": 100.0, | |
| "min_rate": 0.0, | |
| "nhead": 12, | |
| "num_embeddings": 1882, | |
| "num_layers": 12, | |
| "num_rate_bins": 1000, | |
| "padding_idx": 3, | |
| "rate_head_type": "scalar", | |
| "rate_head_use_mlp": true, | |
| "rotary_emb_dim": 64, | |
| "scalar_fn": "exp", | |
| "schedule_type": "simplified-kuma" | |
| } |