RuntimeError: Given groups=1, weight of size [5120, 36, 1, 2, 2], expected input[1, 68, 25, 40, 30] to have 36 channels, but got 68 channels instead

Open
  • Is something wrong?
#2,003 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
48/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Quiet
Tech stack
python, pytorch
Domain
machine-learning

Research direction

Start with the WanVideoImageToVideoEncode node and trace how its inputs are assembled before the model receives the tensor; inspect the linked workflow2.json for hidden or serialized connections, including UE links. Compare the tensor's 68 channels with the model's expected 36 and identify where the additional 32 enter. Done means explaining the source and giving a reproducible way to remove it; the issue does not name tests to run.

Written by the indexing model from the issue text.

Description

RuntimeError: 68 channels instead of 36 - WanVideoImageToVideoEncode

Error:

RuntimeError: Given groups=1, weight of size [5120, 36, 1, 2, 2], expected input[1, 68, 25, 40, 30] to have 36 channels, but got 68 channels instead

Setup:

  • ComfyUI version: 0.19.3
  • ComfyUI-WanVideoWrapper: latest (updated)
  • Model: Wan2.2-I2V-A14B (fp8_e5m2_scaled_KJ)
  • GPU: NVIDIA RTX 3070 Ti Laptop (8GB VRAM)
  • OS: Windows 11

Workflow:

  • Two WanVideoModelLoader nodes (HIGH + LOW)
  • Two WanVideoSampler nodes in cascade
  • One WanVideoImageToVideoEncode node
  • WanVideoLoraSelectMulti with Lightning LoRA
  • fun_or_fi2v_model = disabled
  • end_image = not connected (no visible link)
  • extra_latents = not connected (no visible link)

Problem:
The error persists even with fun_or_fi2v_model disabled and no visible connections on end_image or extra_latents. The 68 channels suggest an end_image or extra latent is being injected somewhere, but I cannot find where.

I suspect it may be coming from a cg-use-everywhere (UE) invisible link, but Reject UE Links on the node doesn't fix it.

Question:
What causes exactly 68 channels (36 + 32) in this context, and how can I identify and remove the source of the extra 32 channels?

Thank you!

workflow2.json

Dominant language
Python
Stars
6.8k
Forks
712
No merged PRs in 30d

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from kijai/ComfyUI-WanVideoWrapper

All issues in kijai/ComfyUI-WanVideoWrapper

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.