PR #14100 Add Nunchaku Lite single-file quantization

Add Nunchaku Lite single-file quantization #14100

sayakpaul merged 33 commits into huggingface:main from rootonchair:feature/nunchaku-lite-single-file

Add Nunchaku Lite single-file quantization

7f4a3a00

github-actions added size/L

github-actions removed size/L

github-actions added quantization

github-actions added tests

github-actions added single-file

rootonchair marked this pull request as draft 85 days ago

sayakpaul commented on 2026-07-02

Support config-backed Nunchaku Lite loading

1a66ac91

github-actions added size/L

sayakpaul commented on 2026-07-02

Remove Nunchaku runtime manifest metadata loading

36b4fc41

Simplify Nunchaku compact config loading

db05e0b0

Add Nunchaku Lite quantization tests

5d4822a3

sayakpaul commented on 2026-07-03

Document Nunchaku Lite checkpoint loading

c630a88b

github-actions removed single-file

github-actions added documentation

Refine Nunchaku Lite quantization docs

ab246af3

sayakpaul requested a review from

asomoza 83 days ago

sayakpaul requested a review from

SunMarc 83 days ago

sergereview requested changes on 2026-07-03

Remove unused Nunchaku smooth factor original weights

57323cd9

Update docs/source/en/quantization/nunchaku.md

5fef94f8

Update src/diffusers/quantizers/nunchaku/utils.py

fb334305

Update src/diffusers/quantizers/nunchaku/nunchaku_quantizer.py

8e14847b

Address Nunchaku Lite review feedback

f5c0179f

github-actions added utils

rootonchair marked this pull request as ready for review 81 days ago

Update Nunchaku Lite quantizer

8ba15fff

SunMarc commented on 2026-07-07

Rename Nunchaku Lite fp4 precision to nvfp4

661d6f7a

Allow torch_dtype to override Nunchaku compute dtype

1c58a6f9

Keep Nunchaku replacement modules on meta during loading

4c77a517

Require kernels when importing Nunchaku runtime utils

16a9cd09

Use kernels requirement decorators for Nunchaku tests

7b186f09

Validate Nunchaku CUDA capability requirements

452de403

sayakpaul commented on 2026-07-09

Merge branch 'main' into feature/nunchaku-lite-single-file

b849b18a

Apply suggestions from code review

9284dd09

rootonchair force pushed from 03399a0a to 55229544 77 days ago

docs: update nunchaku quantization guide

48065859

test: remove nunchaku quantization tests

d687751f

fix: reject hopper for nunchaku lite

86cd7c86

rootonchair force pushed from 55229544 to 86cd7c86 77 days ago

rootonchair commented on 2026-07-09

Apply suggestion from @rootonchair

fd588930

docs: update nunchaku lite compile example

f6897aff

Merge branch 'feature/nunchaku-lite-single-file' of github.com:rooton…

3df700a0

Merge branch 'main' into feature/nunchaku-lite-single-file

0d772788

sayakpaul requested a review from

stevhliu 73 days ago

sayakpaul approved these changes on 2026-07-13

Apply style fixes

b23fd3cf

stevhliu approved these changes on 2026-07-13

Merge branch 'main' into feature/nunchaku-lite-single-file

10236fe3

Apply suggestions from code review

5fdc0266

update doc to be more reasonable

0d63eea3

fix import that requires kernels

d03f80a4

sayakpaul approved these changes on 2026-07-14

sayakpaul merged 60ec6f72 into main 73 days ago

Reviewers

sayakpaul

stevhliu

sergereview

SunMarc

asomoza

Assignees

No one assigned

Labels

documentation quantization tests size/L utils

Milestone

No milestone

diffusers Add Nunchaku Lite single-file quantization #14100 Merged

Add Nunchaku Lite single-file quantization #14100

diffusers
Add Nunchaku Lite single-file quantization
#14100

Merged