Lipeng Wang

``` RuntimeError: function ProjectGaussiansBackward returned an incorrect number of gradients (expected 14, got 13) ``` it seems that the ``backward`` function in ``project_gaussians.py`` should return a ``None`` more corresponding to...

``batch['text']`` adds ``[self.mapping_task[task_name]]`` twice

1

function [``log_images``](https://github.com/lighten001/UniControl/blob/472c6cd347df2d2bccba3875d66d27946e847290/cldm/cldm.py#L366) will call ``get_input`` (which is the second time be called) on train batch end ```python z, c = self.get_input(batch, self.first_stage_key, bs=N) ``` so ``batch['text']`` will add ``[self.mapping_task[task_name]]`` twice...