modelscope / modelscope/DiffSynth-Studio

support for wan2.1 mini-batch finetune

Open
#682 2 comments 1 reaction 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,
When finetune wan2.1 on my own dataset, I noted that the current finetuning script doesn't support batch training. It seems that the code only support 1sample/GPU. Like the flow matching schedule, noise generation, etc.

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

Start with the current Wan2.1 finetuning script and trace how batch size, the flow-matching schedule, and noise generation are handled. Done means the script can finetune with more than one sample per GPU, including correct batch behavior for these steps.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.