Skip to content

[Bug] view size is not compatible with input tensor's size and stride #598

Description

@greyaz

Checklist

Describe the Bug

The KSampler throw an error when executing an nunchaku-qwen-image workflow:

view size is not compatible with input tensor's size and stride (at least one dimension spans across two contiguous subspaces). Use .reshape(...) instead.

The error report:
report.md

Environment

System Information

  • ComfyUI Version: 0.3.59
  • Arguments: main.py --listen 0.0.0.0
  • OS: Ubuntu
  • Python Version: 3.12.11 | packaged by Anaconda, Inc. | (main, Jun 5 2025, 13:09:17) [GCC 11.2.0]
  • Embedded Python: false
  • PyTorch Version: 2.8.0+cu128
  • Nunchaku: 1.0.0
  • ComfyUI-nunchaku: 1.0.1

Devices

  • Name: cuda:0 NVIDIA GeForce RTX 4090 : cudaMallocAsync
  • Type: cuda
  • VRAM Total: 25386352640
  • VRAM Free: 24975835136
  • Torch VRAM Total: 0
  • Torch VRAM Free: 0

Reproduction Steps

qwen-imge-basic.json

Image

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions