modelscope / modelscope/DiffSynth-Studio

Inquiry regarding Full Training speed and First-frame color shift on Wan2.2 TI2V 5B

Open
#1,195 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
13.1k
Forks
1.3k
Avg merge
13h 12m
Merged PRs (30d)
45

Description

Hi authors and community,

Huge thanks for open-sourcing the training scripts! It's truly great contribution to the field.

I have two questions regarding the training process and would appreciate any insights from those who have experimented with it:

  1. I am currently attempting a full train of the Wan 2.2 TI2V 5B model. I was wondering if anyone could share their training speed?
    (like, what is the approximate seconds per step (s/step) I should expect? )

  2. I noticed a specific issue during generation/training: there seems to be a slight but noticeable sudden color shift immediately after the first frame (the conditioning image). The subsequent frames have a slightly different color tone compared to the initial frame.
    Has anyone else encountered this issue? Is this related to VAE encoding/decoding or a specific training configuration?

Any advice or reference data would be greatly appreciated. Thanks again!

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

No files, tests, or entry points are named. Start by reproducing the Wan 2.2 TI2V 5B run, recording seconds per step and comparing the conditioning frame with subsequent frames while checking the VAE encoding/decoding and training configuration; done means providing reference speed data and confirming or isolating the color-shift cause.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.