magic-animate icon indicating copy to clipboard operation
magic-animate copied to clipboard

Failed to generate animation results with errors

Open snakeninny opened this issue 1 year ago • 5 comments

I have Ubuntu 22.04.3 server with Tesla V100 SXM2 32GB.

I have followed:

  1. 🏃‍♂️ Getting Started
  2. ⚒️ Installation
  3. 💃 Inference
  4. 🎨 Gradio Demo

The WebUI was launched successfully.

But, with the monalisa image and the running motion sequence, the default random seed (1), sampling steps (25) and guidance scale (7.5), an error occurred on the webpage: Screenshot 2023-12-07 at 18 55 55

while the loading indicator in terminal was around 50%:

/home/zyl/magic-animate/magicanimate/pipelines/pipeline_animation.py:624: FutureWarning: Accessing config attribute `in_channels` directly via 'UNet3DConditionModel' object attribute is deprecated. Please access 'in_channels' over 'UNet3DConditionModel's config object instead, e.g. 'unet.config.in_channels'.
  num_channels_latents = self.unet.in_channels
 50%|██████████████████████████████████████████████████████████████████▌                             

When the loading indicator reached 100% I got this in the terminal:

/home/zyl/magic-animate/magicanimate/pipelines/pipeline_animation.py:624: FutureWarning: Accessing config attribute `in_channels` directly via 'UNet3DConditionModel' object attribute is deprecated. Please access 'in_channels' over 'UNet3DConditionModel's config object instead, e.g. 'unet.config.in_channels'.
  num_channels_latents = self.unet.in_channels
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████| 25/25 [02:00<00:00,  4.81s/it]
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████| 48/48 [00:02<00:00, 16.84it/s]
IMAGEIO FFMPEG_WRITER WARNING: input image is not divisible by macro_block_size=16, resizing from (1544, 516) to (1552, 528) to ensure video compatibility with most codecs and players. To prevent resizing, make your input image divisible by the macro_block_size or set the macro_block_size to 1 (risking incompatibility).
[swscaler @ 0x69835c0] Warning: data is not aligned! This can lead to a speed loss

and the webpage showed error on the top center: Screenshot 2023-12-07 at 18 57 43

Any ideas? Thank you!

snakeninny avatar Dec 07 '23 11:12 snakeninny

I have Ubuntu 22.04.3 server with Tesla V100 SXM2 32GB.

I have followed:

  1. 🏃‍♂️ Getting Started
  2. ⚒️ Installation
  3. 💃 Inference
  4. 🎨 Gradio Demo

The WebUI was launched successfully.

But, with the monalisa image and the running motion sequence, the default random seed (1), sampling steps (25) and guidance scale (7.5), an error occurred on the webpage: Screenshot 2023-12-07 at 18 55 55

while the loading indicator in terminal was around 50%:

/home/zyl/magic-animate/magicanimate/pipelines/pipeline_animation.py:624: FutureWarning: Accessing config attribute `in_channels` directly via 'UNet3DConditionModel' object attribute is deprecated. Please access 'in_channels' over 'UNet3DConditionModel's config object instead, e.g. 'unet.config.in_channels'.
  num_channels_latents = self.unet.in_channels
 50%|██████████████████████████████████████████████████████████████████▌                             

When the loading indicator reached 100% I got this in the terminal:

/home/zyl/magic-animate/magicanimate/pipelines/pipeline_animation.py:624: FutureWarning: Accessing config attribute `in_channels` directly via 'UNet3DConditionModel' object attribute is deprecated. Please access 'in_channels' over 'UNet3DConditionModel's config object instead, e.g. 'unet.config.in_channels'.
  num_channels_latents = self.unet.in_channels
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████| 25/25 [02:00<00:00,  4.81s/it]
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████| 48/48 [00:02<00:00, 16.84it/s]
IMAGEIO FFMPEG_WRITER WARNING: input image is not divisible by macro_block_size=16, resizing from (1544, 516) to (1552, 528) to ensure video compatibility with most codecs and players. To prevent resizing, make your input image divisible by the macro_block_size or set the macro_block_size to 1 (risking incompatibility).
[swscaler @ 0x69835c0] Warning: data is not aligned! This can lead to a speed loss

and the webpage showed error on the top center: Screenshot 2023-12-07 at 18 57 43

Any ideas? Thank you!

Sorry, I cannot reproduce this bug on my side.

zcxu-eric avatar Dec 07 '23 15:12 zcxu-eric

I had the same issue. The video is still processed and you can find the results in the demo/ folder. Your terminal should still continue showing progress. Seems to be some gradio issue. Or use pose.rip

hassantsyed avatar Dec 07 '23 19:12 hassantsyed

As @hassantsyed said, it seems to be a gradio issue; after 100% progress I got results in magic-animate/demo/outputs

snakeninny avatar Dec 08 '23 01:12 snakeninny

The problem is on Gradio timeout setting, when the loading is too long.

I created a pull request : https://github.com/magic-research/magic-animate/pull/94/commits/3c53a883483126c53e9f67be9e4029f1adfdbe97

Or you can simply change line 96 in demo/gradio_animate.py demo.queue().launch(share=True)

catyung avatar Dec 11 '23 10:12 catyung

The problem is on Gradio timeout setting, when the loading is too long.

I created a pull request : 3c53a88

Or you can simply change line 96 in demo/gradio_animate.py demo.queue().launch(share=True)

That solves my problem. Thank you!

codeayzy avatar Dec 12 '23 09:12 codeayzy