Different Conditioning for each Transformer block
- Dominant language
- Python
- Stars
- 133k
- Forks
- 15.7k
- Avg merge
- 1d 7h
- Merged PRs (30d)
- 158
Description
### Feature Idea
From my research, some transformer blocks is very sensitive and fragile to emphasized conditioning.
If we can feed different conditioning to such transformer blocks, we can use much stronger emphasis.
However, to use above technique, current Comfyui implementation has two problems:
1. I have to encode model conds manually. I don't want to do it as it would be compatible issue.
2. Comfyui backend may split batches which we cannot deal with.
Please make your implementation easier to deal with above problems.
### Existing Solutions
_No response_
### Other
Examples:
Feeding same conditioning to every transformer blocks
Feeding emphasized conditioning only to /model.diffusion_model.output_blocks\.0\.1\.transformer_blocks\.[0-9]/ .
Contributor guide
Assessment
This issue has not been assessed yet.