modelscope / modelscope/DiffSynth-Studio

[FETAURE] Add WAN Vace LoRA train with full dataset

Open
#1,326 1 comment 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

I want to train LoRA for Vace version of WAN2.2.
My main task is make stable video outpainting.
So, I have dataset data
cropped.mp4 - cropped video with grey border.
mask.mp4 - video, where every frame is mask
orig.mp4 - original video to compare output
description.txt - text file with positive prompt

Dataset arch:

- <root_dir>
-- <video_guid>
---- cropped.mp4
---- mask.mp4
---- orig.mp4
---- description.txt
-- <video_guid>
---- cropped.mp4
---- mask.mp4
---- orig.mp4
---- description.txt

How can I train LoRA for Vace layers with full dataset?

Model for ex.: https://huggingface.co/linoyts/Wan2.2-VACE-Fun-14B-diffusers

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 by reviewing the requested Wan2.2-VACE-Fun-14B-diffusers model and the dataset files cropped.mp4, mask.mp4, orig.mp4, and description.txt. Determine how the existing training workflow handles the full dataset layout and VACE layers; done means a defined, supported way to train the requested LoRA for stable video outpainting.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
ai, machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.