GitLab

* add v2 flux examples

* add the docs

* add docs

* update

* finished ops

* add ops

* update

* update

* update

* update

* update

* update

* update

* update docstrings

* update

* update

* update

* update

* update

* update

* update

* finished the api docs

* update

* update

* tmp

* update

* update

* finished the offloading impl

* the offloading is buggy

* update utils

* the offloading is still buggy

* update

* correctness and speedup done; need to check the vram overhead

* done

* final debugging

* update

* update

* correct now

* fix

* update

* use per-layer offloading

* fix the offloading on 5090

* support setting the num_blocks_on_gpu

* change the import name

* feat: add support for teacache in flux kontext

* merge main and make linter happy

* caching_v2

* rename fb cache and write docstring

* lint

* rename utils to fbcache

* no need maintain sana for caching

* update

* update

* update README

* update dos

* update docs

* improve the lightning script

* update the example script

* change the repo name

* start refract the codebase

* update

* update

* start to implement ops

* add gemm

* write the docstrings

* define the w4a4 svdq linear

* update

* make the linter happy

* finished the SVDQW4A4Linear

* finished the SVDQW4A4Linear

* update

* update

* add a patcher to the model

* update

* add adanormsinglezero

* update

* update

* finished the naive implementation of nunchaku flux

* add ff

* finished the naive forward

* update

* svdq linear

* start debugging

* fix some issues

* successfully built the model

* update

* successfully load the model

* update

* update

* update

* try to making it runnable

* debugging

* debugging

* debugging

* add bias to awq linear

* run through

* fix the normalization

* update

* update

* update

* fix the attention

* fix the no fuse nvfp models

* update

* finished the fused ff

* make linter happy

* make linter happy

* make linter happy

* debugging the fp16 attn

* nunchaku fp16 is buggy

* finish the fp16 attn

* fp4 done

* fix the lora scales

* add a default value for alpha; need to debug int4

* fix input4

* update

* update

* ff does not work

* specialize the processors

* qwen transformer done. start debugging

* make linter happy

* add schnell v2 for metrics eval

* chore: schnellv2 eval

* update

* ff and attention correct

* need to check what happened to module

* fp4 done

* make linter happy

* update an example script

* reformat

* add an example script

* add the annoucement

* remove a misleading info

* ready to release

* fix teacache_batch

* lint

* Fix FLUX.1-Kontext LoRA support and dimension mismatch issues

- Added convert_keys_to_diffusers() for ComfyUI/PEFT format conversion
- Fixed dimension mismatch in LoRA weight concatenation
- Added preprocessing for single_blocks LoRA structure
- Added comprehensive test suite for Kontext LoRA
- Added example script for FLUX.1-Kontext with LoRA

Fixes #354

* lint

* FAL.AI and relight-kontext-lora patch

* feat: support IP-adapter

* FBCache and comfyUI

* fixing conflicts

* update

* update example

* update example

* style: make linter happy

* update

* update ipa test

* add docs and rename IP to ip

* docs: add docs for ipa

* docs: add docs for ipa

* add an example for pulid

* update

* save gpu memory

* change the threshold to 0.8

---------
Co-authored-by: Muyang Li <lmxyy1999@foxmail.com>

* add colossus script

* finalize colossus

* update kontext examples

* update tests

* add tests for kontext

* remove the warning of txt_ids and img_ids

* chore: add kontext to be synced from hf to ms

* add kontext demo

* make linter happy

* style: make linter happy

* update docs

* fix:cache issue if offload is set to True

* fix: lint

* add a script to merge models

* finished

* try to merge t5

* merge the config into meta files

* rewrite the t5 model loading

* consider the case of subfolder

* merged the qencoder files

* make the linter happy and fix the tests

* pass tests

* add deprecation messages

* add a script to merge models

* schnell script runnable

* update sana

* modify the model paths

* fix the model paths

* style: make the linter happy

* remove the debugging assertion

* chore: fix the qencoder lpips

* fix the lpips

* Updating quantized t5 encoder

* Fix formatting based on pre-commit hook

* Update test cases

* Fixing linter issue

* Fix linter reformatting

* support fp4

* style: make linter happy

* update the fp4 lpips

* Prevent downloading original t5 model

* Make sure model in eval mode

---------
Co-authored-by: muyangli <lmxyy1999@foxmail.com>

* style: reformated codes

* style: reformated codes

* add pulid

* Add the feature that allows the mixed use of pulid and non-pulid after loading pulid to generate the pipeline.

* Added the feature to load LoRA at any time.

* Organized the directory structure.

* Organized the code.

* Removed unused related code from eva-clip.

* style: apply Ruff formatting

* Refactored code and verified pulid works.

* add pulid tests

* auto detect precision in test

* Updated requirements.txt

* update requirements

* style: reformat the example

* style: reformat the example

* style: rename cb to call_back

* style: format the codes

* style: format the codes

* reformated the codes

* fix the repo forward

* clean some dead codes

* wrap up for pulid

---------
Co-authored-by: kkkxue <kkkxue@tencent.com>
Co-authored-by: muyangli <lmxyy1999@foxmail.com>

* feat: expose norm1 layer to support TeaCache

* feat: add TeaCache example

* feat: add idx as optional parameter

* chore: rename function

* refactor: move TeaCache decorator into example script

* test: add a test for the combination of Nunchaku with TeaCache

* feat: expose norm1 layer to support TeaCache

* feat: add TeaCache example

* feat: add idx as optional parameter

* chore: rename function

* refactor: move TeaCache decorator into example script

* test: add a test for the combination of Nunchaku with TeaCache

* fix: make tests run on low memory hardware

* fix: ensure that memory is correctly released between tests

* fix: avoid moving pipeline to device prematurely

* gpu memory does not release

* need to figure out a way to get compatible with offloading

* wrap up the teacache

---------
Co-authored-by: muyangli <lmxyy1999@foxmail.com>

* DoubleFBCache

* rename > DoubleFBCache to use_double_fb_cache

Ready to release v0.2.0

Co-authored-by: Zhekai Zhang <sxtyzhangzk@gmail.com>
Co-authored-by: Muyang Li <lmxyy1999@foxmail.com>
Co-authored-by: Yujun Lin <16437040+synxlin@users.noreply.github.com>

[major] update the lora conversion instructions; add customized lora comfyui nodes; upgrade the version