kohya-ss / kohya-ss/sd-scripts

save_state does not work with accelerate==0.19.0

Open
#715 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
7.2k
Forks
1.2k
Avg merge
11m
Merged PRs (30d)
2

Description

I'm training a SDXL lora using a A4000 16GB gpu.

With batch size = 1, the vram usage is around 7xxxMB, but OOM occur during save_state.

I tried to upgrade accelerate to the latest version (0.21.0), and save_state is working.

Contributor guide

No contributing guide indexed for this repository

Research direction

Start by reproducing save_state with accelerate 0.19.0 using the reported SDXL LoRA setup and compare its behavior with accelerate 0.21.0. Done means the out-of-memory failure is fixed or the supported accelerate version and workaround are clearly documented.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.