kohya-ss / kohya-ss/sd-scripts
Adaptive LR for each block (SDXL)
Open
- Dominant language
- Python
- Stars
- 7.2k
- Forks
- 1.2k
- Avg merge
- 11m
- Merged PRs (30d)
- 2
Description
Could using an adaptive LR for each trainable Unet block be a good idea?
Contributor guide
No contributing guide indexed for this repository
Research direction
No file, test, or entry point is identified. First clarify the adaptive-learning-rate algorithm and which trainable SDXL U-Net blocks it should cover, then locate the relevant training and optimizer entry points. Done means an agreed design and an implementation validated by appropriate training tests or evidence.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100