[Bug]: Gets stuck after a few generations

[F14V10]

Is there an existing issue for this?

• [X] I have searched the existing issues and checked the recent builds/commits of both this extension and the webui

Have you read FAQ on README?

• [X] I have updated WebUI and this extension to the latest version

What happened?

Every few generations (sometimes even the first) the progress bar in the terminal gets stuck at 100% of the generation progress and in the web UI at 95%. The terminal then crashes without an error and the current generation aborts. I can then only restart the process. But it usually works a few times and then stops for no apparent reason.

Steps to reproduce the problem

1. generate a few GIFs or videos with AnimateDiff enabled
2. the progress bar gets stuck and the terminal crashes

What should have happened?

It shouldn't get stuck. :point_right::point_left:

Commit where the problem happens

webui: v1.6.0 extension: v1.9.4 (https://github.com/continue-revolution/sd-webui-animatediff/commit/2f0e68d08028b0440a899c57dcfaf4972a5f0185)

What browsers do you use to access the UI ?

Mozilla Firefox

Command Line Arguments

--xformers --precision full --no-half --no-half-vae --api

Console logs

2023-10-22 18:43:20,922 - AnimateDiff - INFO - AnimateDiff process start.
2023-10-22 18:43:20,923 - AnimateDiff - INFO - You are using tested mm mm_sd_v15_v2.safetensors.
2023-10-22 18:43:20,932 - AnimateDiff - INFO - Injecting motion module mm_sd_v15_v2.safetensors into SD1.5 UNet middle block.
2023-10-22 18:43:20,933 - AnimateDiff - INFO - Injecting motion module mm_sd_v15_v2.safetensors into SD1.5 UNet input blocks.
2023-10-22 18:43:20,933 - AnimateDiff - INFO - Injecting motion module mm_sd_v15_v2.safetensors into SD1.5 UNet output blocks.
2023-10-22 18:43:20,934 - AnimateDiff - INFO - Setting DDIM alpha.
2023-10-22 18:43:20,938 - AnimateDiff - INFO - Injection finished.
2023-10-22 18:43:20,938 - AnimateDiff - INFO - Hacking lora to support motion lora
2023-10-22 18:43:20,938 - AnimateDiff - INFO - Hacking CFGDenoiser forward function.
2023-10-22 18:43:20,939 - AnimateDiff - INFO - Hacking ControlNet.
100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [01:10<00:00,  3.51s/it]
Total progress: 100%|██████████████████████████████████████████████████████████████████| 20/20 [01:07<00:00,  3.57s/it]

Additional information

Thank you for your effort and this great extension! :heart:

[continue-revolution]

Please observe your output dir to see if there is any new output images, especially when this happens again. It can help me figure out where the problem is.

[F14V10]

There are no videos, images or grids of failed generations in the output directory. Only those of the successful ones. I have activated the preview during the generation. The preview grid that is shown last before it gets stuck already looks a lot like the final result. It is just not saved or converted to a video or gif.

[LIQUIDMIND111]

Please observe your output dir to see if there is any new output images, especially when this happens again. It can help me figure out where the problem is.

hello mate, one question, for VIDEO to VIDEO, when we enable control net, i get this error: RuntimeError: You have not selected any ControlNet Model. - DOES THIS MEANS I NEED TO CHOOSE A MODEL? Which one you suggest? Or i ignore that?

[Tyranomaster]

@continue-revolution I've had the same issue as the original post, and did some troubleshooting and gotten some extra details for you.

Issue only occurs when PNG output type is selected (for me at least). When PNG is turned off, proceeds as intended and produces output.

Further PNG won't fail if you interrupt early enough in the process (for a 40 step generation, interrupting by step 20 tends to succeed in finishing, but it isn't consistent).

[alieverlove]

same Issue.. using a1111 web ui .. :')