Owner

kohya-ss

6 indexed repositories · View on GitHub

sd-scripts

Python · 7240 stars
musubi-tuner

Python · 2054 stars
sd-webui-additional-networks

Python · 1815 stars
ControlNet-LLLite-ComfyUI

Python · 225 stars
ComfyUI-Anima-LLLite

Python · 215 stars
pixel-space-sdxl
Pixel-space SDXL training

Python · 8 stars

Beginner friendly Everything

Open All

MiniMax-H3 training hangs on first step (Windows, RTX 5090, torch 2.14+cu130) — inference works Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

kohya-ss/musubi-tuner#1123 · Opened Sep 18, 2026
MiniMax-H3: apply rope without the batched 2x2 matmul Open

Difficulty 3/5 1-2 days Newbie friendliness 78/100

kohya-ss/musubi-tuner#1122 · 1 reaction · Opened Sep 17, 2026
MInimax h3 ref2av train error Open

Difficulty 4/5 3-5 days Newbie friendliness 30/100

kohya-ss/musubi-tuner#1087 · 1 comment · Opened Sep 3, 2026
Inconsistent retention semantics: --save_last_n_steps is a step window, --save_last_n_epochs is a checkpoint count Open

Difficulty 4/5 3-5 days Newbie friendliness 55/100

kohya-ss/musubi-tuner#1082 · Opened Sep 1, 2026
qwen_image_generate_image: TypeError when guidance_scale > 1 and no --negative_prompt given Open

Difficulty 2/5 1-3 hours Newbie friendliness 82/100

kohya-ss/musubi-tuner#1080 · Opened Aug 31, 2026
--attn_mode sdpa is offered by the CLIs but unimplemented ("Unsupported attention mode: sdpa" / KeyError) Open

Difficulty 2/5 1-3 hours Newbie friendliness 84/100

kohya-ss/musubi-tuner#1077 · Opened Aug 30, 2026
qwen_image_generate_image: --fp8_scaled fails with "self and mat2 must have the same dtype" on the non-scaled_mm path (ROCm) Open

Difficulty 1/5 Under an hour Newbie friendliness 84/100

kohya-ss/musubi-tuner#1076 · Opened Aug 30, 2026
Krea2 training hangs indefinitely (0 steps, GPU 100% util but abnormally low power) when `batch_size >= 2` combined with block swapping, on RTX 5070 Ti (Blackwell) Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

kohya-ss/musubi-tuner#1071 · 5 comments · Opened Aug 25, 2026
Tensor not moved to CPU on numpy 2 Open

Difficulty 3/5 1-2 days Newbie friendliness 64/100

kohya-ss/musubi-tuner#1068 · Opened Aug 24, 2026
ValueError: Audio source is materially short / Audio stream is discontinuous (Includes patchfix) Open

Difficulty 3/5 1-2 days Newbie friendliness 55/100

kohya-ss/musubi-tuner#1066 · 4 comments · Opened Aug 21, 2026
`torch.cuda.amp` in `__main__` self-test blocks and the IPEX hijack — deprecated since torch 2.3/2.4, unguarded on the main install path Open

Difficulty 3/5 1-2 days Newbie friendliness 78/100

kohya-ss/sd-scripts#2420 · Opened Aug 20, 2026
MiniMax H3: shared cache_directory leaks validation samples into training buckets Open

Difficulty 3/5 1-2 days Newbie friendliness 68/100

kohya-ss/musubi-tuner#1060 · 1 comment · Opened Aug 16, 2026
LoRA trained for Krea 2 shows noise patterns in generated output Open

Difficulty 4/5 3-5 days Newbie friendliness 42/100

kohya-ss/musubi-tuner#1053 · 3 comments · 1 reaction · Opened Aug 14, 2026
Feature Request: Add support for SenseNova-U1.5 Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

kohya-ss/musubi-tuner#1052 · Opened Aug 13, 2026
If max_train_steps == sample/save every n, it will sample/save it twice on the final step Open

Difficulty 3/5 1-2 days Newbie friendliness 68/100

kohya-ss/musubi-tuner#1048 · 1 reaction · Opened Aug 12, 2026
[Feature Request] Add option to disable cuDNN/MIOpen backend (workaround for AMD RDNA 4 GPU crashes) Open

Difficulty 3/5 1-2 days Newbie friendliness 52/100

kohya-ss/sd-scripts#2417 · Opened Aug 12, 2026
MiniMax-H3 Support Roadmap Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

kohya-ss/musubi-tuner#1029 · 14 comments · 9 reactions · Opened Aug 9, 2026
WeightsUnpickler error: Unsupported operand 56 Open

Difficulty 2/5 1-3 hours Newbie friendliness 68/100

kohya-ss/sd-scripts#2414 · Opened Aug 8, 2026
Feature Request: Add support for Explorative Modeling (explorative-modeling.github.io) Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

kohya-ss/musubi-tuner#1019 · Opened Aug 5, 2026
The training stops at epoch 1 and does not continue to increase the number of steps Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

kohya-ss/musubi-tuner#1015 · Opened Jul 31, 2026
Documentation: Include AMD ROCm instructions Open

Difficulty 2/5 1-3 hours Newbie friendliness 68/100

kohya-ss/musubi-tuner#1013 · 1 reaction · Opened Jul 29, 2026
GUI MusubiSuite Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

kohya-ss/musubi-tuner#1009 · Opened Jul 22, 2026
[bug] --skip_until_initial_step 使用時、--gradient_accumulation_steps > 1 で再開位置を過剰にスキップして学習が終了する Open

Difficulty 3/5 1-2 days Newbie friendliness 70/100

kohya-ss/sd-scripts#2407 · 1 comment · Opened Jul 20, 2026
--tag_replacement doesn't actually replace anything (variable shadowing bug) Open

Difficulty 1/5 Under an hour Newbie friendliness 88/100

kohya-ss/sd-scripts#2405 · Opened Jul 19, 2026
Int8 training ? Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

kohya-ss/musubi-tuner#1007 · Opened Jul 17, 2026
Race in block-swap offloading: intermittent "Expected all tensors to be on the same device, cuda:0 and cpu" during multi-prompt inference Open

Difficulty 2/5 1-3 hours Newbie friendliness 76/100

kohya-ss/musubi-tuner#1006 · Opened Jul 16, 2026
[Feature Request] SCAIL 2 LoRa Training Open

Difficulty 5/5 Over a week Newbie friendliness 30/100

kohya-ss/musubi-tuner#1005 · Opened Jul 16, 2026
Silent `adaln` drop while anima lora training Open

Difficulty 3/5 1-2 days Newbie friendliness 50/100

kohya-ss/sd-scripts#2403 · Opened Jul 16, 2026
Krea 2 text encoder silently truncates captions longer than 512 tokens Open

Difficulty 4/5 3-5 days Newbie friendliness 55/100

kohya-ss/musubi-tuner#1002 · Opened Jul 13, 2026
Feat: I have updated the Musubi Tuner GUI Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

kohya-ss/musubi-tuner#996 · 3 reactions · Opened Jul 10, 2026
LECO: verify_training_args() crashes on missing sample_every_n_epochs/sample_every_n_steps Open

Difficulty 2/5 1-3 hours Newbie friendliness 72/100

kohya-ss/sd-scripts#2402 · Opened Jul 8, 2026
[Feature Request] Support Krea 2 Edit LoRA training Open

Difficulty 5/5 Over a week Newbie friendliness 42/100

kohya-ss/musubi-tuner#993 · 2 comments · 1 reaction · Opened Jul 7, 2026
Issue to convert full finetuned model ( ZIB ) generated with musubi tuner to int8 Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

kohya-ss/musubi-tuner#991 · Opened Jul 6, 2026
How to train lora with trigger word? Open

Difficulty 3/5 1-2 days Newbie friendliness 35/100

kohya-ss/musubi-tuner#990 · 1 comment · Opened Jul 2, 2026
Incompatible with int8 models Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

kohya-ss/ComfyUI-Anima-LLLite#7 · 1 comment · Opened Jun 27, 2026
Krea 2 Lora train with 16gb vram Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

kohya-ss/musubi-tuner#985 · 17 comments · Opened Jun 26, 2026
Krea 2 Error on Cache Text Encoder Outputs Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

kohya-ss/musubi-tuner#983 · 6 comments · Opened Jun 24, 2026
Training ideogram 4 lora issues. Open

Difficulty 4/5 3-5 days Newbie friendliness 30/100

kohya-ss/musubi-tuner#978 · 6 comments · Opened Jun 21, 2026
Adding Anima ControlNet support Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

kohya-ss/sd-scripts#2394 · 4 reactions · Opened Jun 20, 2026
"Missing keys in state dict" solved when training lora on some anima ft checkpoints Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

kohya-ss/sd-scripts#2392 · 4 reactions · Opened Jun 18, 2026
Feature Request: Support for Boogu-Image-0.1-Edit Open

Difficulty 4/5 3-5 days Newbie friendliness 42/100

kohya-ss/musubi-tuner#974 · 3 comments · Opened Jun 17, 2026
So where's weights? Open

Difficulty 2/5 1-3 hours Newbie friendliness 35/100

kohya-ss/ComfyUI-Anima-LLLite#6 · 1 comment · Opened Jun 17, 2026
Successful HiDreamO1 training? Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

kohya-ss/musubi-tuner#973 · 3 comments · Opened Jun 13, 2026
This is a ads。add online/offline RL method to musubi-tuner Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

kohya-ss/musubi-tuner#965 · Opened Jun 11, 2026
Can we get FLUX 2 Klein full fine tuning if possible? Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

kohya-ss/musubi-tuner#961 · 3 comments · 1 reaction · Opened Jun 8, 2026
Ideogram v4 Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

kohya-ss/musubi-tuner#958 · 1 comment · 1 reaction · Opened Jun 7, 2026
bf16/float RuntimeError - dtype incompatibility Open

Difficulty 3/5 1-2 days Newbie friendliness 38/100

kohya-ss/sd-scripts#2364 · Opened May 29, 2026
Tag info saved in model metadata ignores keep_tokens_separator Open

Difficulty 3/5 1-2 days Newbie friendliness 55/100

kohya-ss/sd-scripts#2323 · Opened May 6, 2026
Support for ernie and nucleus image Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

kohya-ss/musubi-tuner#942 · 1 comment · Opened May 4, 2026
lora save error when output name not set Open

Difficulty 2/5 1-3 hours Newbie friendliness 68/100

kohya-ss/musubi-tuner#940 · Opened Apr 27, 2026

Showing the newest 100

This page lists what was indexed most recently. Advanced filter has the whole inventory, narrowed by language, difficulty and how long a task takes.

Open advanced filter

kohya-ss

Showing the newest 100

Get new issues in your inbox