eps is the SD eps-parameterized vector field with the FM vector field is defined as z - eps First of all convert the x_t from the rectified flow trajectory to the original diffusion trajectory Then calculate the vector field from the eps-parameterized vector field
(self, noise, x_t, t)
| 167 | return vector_field |
| 168 | |
| 169 | def get_vector_field_from_eps(self, noise, x_t, t): |
| 170 | """ |
| 171 | eps is the SD eps-parameterized vector field with |
| 172 | the FM vector field is defined as z - eps |
| 173 | |
| 174 | First of all convert the x_t from the rectified flow trajectory to the original diffusion trajectory |
| 175 | Then calculate the vector field from the eps-parameterized vector field |
| 176 | """ |
| 177 | z_pred = self.predict_start_from_eps(x_t, t, noise) |
| 178 | eps_pred = noise |
| 179 | vector_field = z_pred - eps_pred # z - eps |
| 180 | return vector_field |
| 181 | |
| 182 | def forward(self, x, t, **kwargs): |
| 183 | """ |
no test coverage detected