如何控制生成视频的尺寸大小

[Zeldalina]

我现在想要生成横版视频，输入视频为竖版视频，尺寸为2160 × 2880，输入参考图像为横版，尺寸为2500 × 1875，调整参数 resolution为720。运行程序报错： warnings.warn( DWPose: 100%|█████████████████████████████████████| 171/171 [03:46<00:00, 1.32s/it] 0%| | 0/75 [00:02<?, ?it/s] Traceback (most recent call last): File "inference.py", line 136, in main(args) File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context return func(*args, kwargs) File "inference.py", line 100, in main _video_frames = run_pipeline( File "inference.py", line 68, in run_pipeline frames = pipeline( File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context return func(*args, *kwargs) File "/root/MimicMotion/mimicmotion/pipelines/pipeline_mimicmotion.py", line 570, in call _noise_pred = self.unet( File "/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(args, kwargs) File "/root/MimicMotion/mimicmotion/modules/unet.py", line 481, in forward sample = upsample_block( File "/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(*args, **kwargs) File "/root/miniconda3/lib/python3.8/site-packages/diffusers/models/unets/unet_3d_blocks.py", line 2361, in forward hidden_states = torch.cat([hidden_states, res_hidden_states], dim=1) RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 24 but got size 23 for tensor number 1 in the list.

[Zeldalina]

同样的输入数据，调整参数 resolution为默认的576，则运行成功。目前不理解问题所在。

[Icedcocon]

Have the same problem as you.

[SAWRJJ]

请问 这个问题 有解决吗