Comfy-Org / Comfy-Org/ComfyUI

--reserve-vram doesn't seem to be respected on SamplerCustom step

Open
#5,958 0 comments 0 reactions 0 assignees View on GitHub
AMD Potential Bug
Dominant language
Python
Stars
133k
Forks
15.7k
Avg merge
1d 7h
Merged PRs (30d)
158

Description

### Expected Behavior

I have set --reserve-vram 6 because my AMD 7800 XT keeps crashing if I use all my memory on latest kernels.

### Actual Behavior

All memory is used.

### Steps to Reproduce

Use this workflow and observe that all 16GB vram is filled.
[LTX-video-fail.json](https://github.com/user-attachments/files/18050478/LTX-video-fail.json)

### Debug Logs

```powershell
stable-diffusion-comfyui-1 | got prompt
stable-diffusion-comfyui-1 | model weight dtype torch.bfloat16, manual cast: None
stable-diffusion-comfyui-1 | model_type FLUX
stable-diffusion-comfyui-1 | no CLIP/text encoder weights in checkpoint, the text encoder model will not be loaded.
stable-diffusion-comfyui-1 | Requested to load MochiTEModel_
stable-diffusion-comfyui-1 | loaded completely 9.5367431640625e+25 9083.38671875 True
stable-diffusion-comfyui-1 | 0 models unloaded.
stable-diffusion-comfyui-1 | Requested to load LTXV
stable-diffusion-comfyui-1 | loaded completely 9.5367431640625e+25 3667.902587890625 True
100%|██████████| 30/30 [01:28<00:00, 2.96s/it]
stable-diffusion-comfyui-1 | Requested to load VideoVAE
stable-diffusion-comfyui-1 | loaded completely 9.5367431640625e+25 1599.096866607666 True
```

### Other

I'm running from docker container using this guide.
https://github.com/hartmark/sd-rocm

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.