(sigma_from, sigma_to, eta=1.0)
| 133 | |
| 134 | |
| 135 | def get_ancestral_step(sigma_from, sigma_to, eta=1.0): |
| 136 | if not eta: |
| 137 | return sigma_to, 0.0 |
| 138 | sigma_up = torch.minimum( |
| 139 | sigma_to, |
| 140 | eta * (sigma_to**2 * (sigma_from**2 - sigma_to**2) / sigma_from**2) ** 0.5, |
| 141 | ) |
| 142 | sigma_down = (sigma_to**2 - sigma_up**2) ** 0.5 |
| 143 | return sigma_down, sigma_up |
| 144 | |
| 145 | |
| 146 | def to_d(x, sigma, denoised): |
no outgoing calls
no test coverage detected