| depth: 24 | |
| diffusion_type: flow | |
| heads: 16 | |
| hidden: 2816 | |
| learn_sigma: false | |
| lr: 0.0001 | |
| p2_gamma: 1.2 | |
| p2_k: 1.0 | |
| p_mean: -1.2 | |
| p_std: 1.2 | |
| seq_len: 128 | |
| sigma_data: 1.0 | |
| sigma_max: 15.0 | |
| sigma_min: 0.002 | |
| warmup_ratio: 0.05 | |
| warmup_steps: null | |
| depth: 24 | |
| diffusion_type: flow | |
| heads: 16 | |
| hidden: 2816 | |
| learn_sigma: false | |
| lr: 0.0001 | |
| p2_gamma: 1.2 | |
| p2_k: 1.0 | |
| p_mean: -1.2 | |
| p_std: 1.2 | |
| seq_len: 128 | |
| sigma_data: 1.0 | |
| sigma_max: 15.0 | |
| sigma_min: 0.002 | |
| warmup_ratio: 0.05 | |
| warmup_steps: null | |