lllyasviel / lllyasviel/ControlNet

ControlNet Training in pixel space rather than latent space is hard to produce good results

Open
#752 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
34.1k
Forks
3k
PR merge metrics
No merged PRs in 30d

Description

I have a problem when training ControlNet.
I find the model desn't generate images according to the input condition, regardless of more than 50k iterations
Maybe the reason is that direct learning in pixel space is hard compared with latent space.
Thanks!

Contributor guide

No contributing guide indexed for this repository

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.

Research direction

The issue names no file, test, or entry point, and provides no reproducible training configuration or expected-versus-actual example. Start by documenting the ControlNet training setup and reproducing the reported pixel-space behavior; completion would require identifying whether the training approach or configuration causes the condition to be ignored and defining a verified improvement.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.