Comfy-Org / Comfy-Org/ComfyUI

Dynamic vram ltx 2.3 lora audio above 0 strength issue - slow first step

Open
#14,168 6 comments 0 reactions 1 assignee Claimed by @rattus128 View on GitHub
Potential Bug
Dominant language
Python
Stars
133k
Forks
15.7k
Avg merge
1d 7h
Merged PRs (30d)
158

Description

### Custom Node Testing

- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Expected Behavior

When using dynamic vram with ltx2.3 and loras with audio strength above 0 and running another job without changing anything, first step should start immediately.

### Actual Behavior

You run the next job without changing anything and first step seems to reload some weights anyways causing first step to be slow. (You can see GPU usage fluctuations)

### Steps to Reproduce

Load loras with audio strength above 0. Related to all lora loaders.

Top run without audio, bottom with.
Image

### Debug Logs

```powershell
with audio

[INFO] got prompt
[WARNING] Warning: TAESD previews enabled, but could not find models/vae_approx/None
[INFO] 0 models unloaded.
[INFO] Model LTXAV prepared for dynamic VRAM loading. 16775MB Staged. 1440 patches attached. Force pre-loaded 608 weights:
3303 KB.
100%|██████████████████████████████████████████████████████████████████████████████████████| 13/13 [00:51<00:00, 3.93s/it]
[INFO] Requested to load AudioVAE
[INFO] loaded completely; 693.46 MB loaded, full load: True
[INFO] 0 models unloaded.
[INFO] Model VideoVAE prepared for dynamic VRAM loading. 1384MB Staged. 0 patches attached.
[INFO] Prompt executed in 67.45 seconds

without audio

[INFO] got prompt
[WARNING] Warning: TAESD previews enabled, but could not find models/vae_approx/None
[INFO] 0 models unloaded.
[INFO] Model LTXAV prepared for dynamic VRAM loading. 16775MB Staged. 480 patches attached. Force pre-loaded 608 weights: 3
303 KB.
100%|██████████████████████████████████████████████████████████████████████████████████████| 13/13 [00:38<00:00, 2.96s/it]
[INFO] Requested to load AudioVAE
[INFO] loaded completely; 693.46 MB loaded, full load: True
[INFO] 0 models unloaded.
[INFO] Model VideoVAE prepared for dynamic VRAM loading. 1384MB Staged. 0 patches attached.
[INFO] Prompt executed in 50.80 seconds
```

### Other

When only the distill lora with audio strength is loaded this bug does not occur for some reason. (Certain others seem unaffected as well.)

When loras are loaded with 0 audio strength the job begins immediately.

Most loras seem to cause this issue, and multiple loras can cause the first step to become EVEN slower.

When dynamic vram is disabled, the issue vanishes completely.

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.