Graduate student in Taiwan. Leader of KohakuLab. Researcher/Dev in ComfyOrg
Taiwan
Joined May 2021
- Tweets3.7K
- Following1K
- Followers4.1K
- Likes29.1K
Pinned Tweet
TT-VidT is accepted at #NeurIPS2026 🎉
A temporal-decoupled methodology in video pretraining to provide real "motion" info extracted from the video and only requires small scale training to reach high quality results.
Huge thanks to all collaborators, @NVIDIAAI and @spellbrush
琥珀青葉@KohakuLab retweeted
I made FP16 fixed version of MOSS-image
It support any FP16 compute card (such as V100/2080ti) without any issue. and if your lora or PEFT model only trained QKVO/MLP layers, it can be used directly without issue!
huggingface.co/KBlueLeaf/MOS…
KohakuMossTrainer is released:
github.com/KohakuBlueleaf/Ko…
**Features**
- LyCORIS all algorithms
- SM70 ~ SM120
- Intel Arc GPU
- Windows and LInux
- multi GPU training
- Model offload
***Try to test it now!***
I made FP16 fixed version of MOSS-image
It support any FP16 compute card (such as V100/2080ti) without any issue. and if your lora or PEFT model only trained QKVO/MLP layers, it can be used directly without issue!
huggingface.co/KBlueLeaf/MOS…
KohakuMossTrainer is released:
github.com/KohakuBlueleaf/Ko…
**Features**
- LyCORIS all algorithms
- SM70 ~ SM120
- Intel Arc GPU
- Windows and LInux
- multi GPU training
- Model offload
***Try to test it now!***
琥珀青葉@KohakuLab retweeted
The style lora test on Moss-image-base
with 100step only
KohakuMossTrainer is released:
github.com/KohakuBlueleaf/Ko…
**Features**
- LyCORIS all algorithms
- SM70 ~ SM120
- Intel Arc GPU
- Windows and LInux
- multi GPU training
- Model offload
***Try to test it now!***
The style lora test on Moss-image-base
with 100step only
KohakuMossTrainer is released:
github.com/KohakuBlueleaf/Ko…
**Features**
- LyCORIS all algorithms
- SM70 ~ SM120
- Intel Arc GPU
- Windows and LInux
- multi GPU training
- Model offload
***Try to test it now!***
琥珀青葉@KohakuLab retweeted
KohakuMossTrainer is released:
github.com/KohakuBlueleaf/Ko…
**Features**
- LyCORIS all algorithms
- SM70 ~ SM120
- Intel Arc GPU
- Windows and LInux
- multi GPU training
- Model offload
***Try to test it now!***
KohakuMossTrainer is released:
github.com/KohakuBlueleaf/Ko…
**Features**
- LyCORIS all algorithms
- SM70 ~ SM120
- Intel Arc GPU
- Windows and LInux
- multi GPU training
- Model offload
***Try to test it now!***
KohakuTPU's new compute unit successfully make the attention MFU improved to around 40% (from 28%) with less LUT, less BRAM and less DSP, but also slightly less flexibility (as it is row wise operation unit)
but as most nn never need ability beyond the sru I guess it is fine...
The design have tinygrad support, have its own DSL and IR/Copmiler.
If you have interests, plz check: github.com/KohakuBlueleaf/Ko…
Finally I will target 24 matrix core + 8vector core
The MFU number above is 16 + 8 so 24+8 may have lower MFU.
And as the RV64IMA cpu I have support OS, I will implement KohakuAccel OS as well!
Basically KohakuTPU is now FULLY WORKING!
And I will try to setup my xdma test rig (as windows hate test driver haha) and see if I can run some model inference on it!
Current target is SDXL or Anima at first, then try to run Moss-image (no TE)
KohakuTPU progress:
New framework arch: xache <> sysnode <> mesh fully works, RV64I firmware also works well
software stack redo allow me to reach 25~30% MFU in flash attention, 60~80% MFU in MLP/SwiGLU
FPGA bring-up fully success, it reach 650Mhz on mat and 350Mhz for vec
琥珀青葉@KohakuLab retweeted
いまだに入手性の良さからか、ブラウザアプリ等で水匠5が使われているシーンを目にすることがあります。
水匠5の公開からは5年以上が経過しているので、さすがに水匠5はもういいだろうと思い、水匠5よりパラメーター数が桁違いに少なく、ファイルサイズが1/5以下で、そしてR150以上強い評価関数パラメーターファイルをスマホアプリなどに組み込みやすいようにMIT Licenseで公開することにしました。
ブログにその紹介記事を書きましたので、詳しくはそちらをご覧ください。
yaneuraou.yaneu.com/2026/10/…