modelscope/DiffSynth-Studio

View on GitHub

Enjoy the magic of Diffusion models!

Stars: 13.1k
Forks: 1.3k
Open beginner issues: 1
Indexed issues: 542
Avg merge: 13h 12m
Merged PRs (30d): 45

Dominant language: Python
License: Apache-2.0
Last GitHub push: Sep 20, 2026
Latest indexed: Sep 20, 2026
Contributing guide: No contributing guide
Code of conduct: No code of conduct
Beginner labels: No beginner labels indexed

Beginner friendly Everything

Open All

542 open issues indexed

微调过程的显存消耗 Open

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

modelscope/DiffSynth-Studio#1295 · 1 comment · Opened Feb 11, 2026
我想知道全参数微调qwen-image-edit那个meta-json的数据格式是什么样的？以及图片目录里面的数据格式是什么样的？ Open

Difficulty 3/5 1-2 days Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1294 · 1 comment · Opened Feb 11, 2026
通过配置显存管理和loadhot为True情况下，跑十几张图，每张图来回切不同lora方式，显存会不断往上涨，有什么办法解决吗，让显存能稳定下来？ Open

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

modelscope/DiffSynth-Studio#1292 · 3 comments · Opened Feb 11, 2026
预计什么时候会适配qwen-image2.0 Open

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

modelscope/DiffSynth-Studio#1291 · 1 comment · 2 reactions · Opened Feb 11, 2026
训练的wan模型lora key不匹配 Open

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

modelscope/DiffSynth-Studio#1287 · 3 comments · Opened Feb 10, 2026
Reduce the VAP denosing steps Open

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

modelscope/DiffSynth-Studio#1286 · 1 comment · Opened Feb 10, 2026
基于Z-Image Base模型训练的lora可以用于Turbo模型的推理吗？ Open

Difficulty 4/5 3-5 days Newbie friendliness 20/100

modelscope/DiffSynth-Studio#1285 · 2 comments · Opened Feb 10, 2026
本地加载生图模型，为啥这么慢，怎么多卡推理 Open

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

modelscope/DiffSynth-Studio#1284 · 1 comment · Opened Feb 10, 2026
lora_base_model疑问 Open

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

modelscope/DiffSynth-Studio#1279 · 1 comment · Opened Feb 10, 2026
how to enable 'vram_management_enabled' variable when loading model and lora? Open

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

modelscope/DiffSynth-Studio#1277 · 3 comments · Opened Feb 9, 2026
训练qwen-image-edit的lora或full微调后，推理时候完全不会对图像做修改，为什么？ Open

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

modelscope/DiffSynth-Studio#1276 · 3 comments · Opened Feb 9, 2026
The inference in DiffSynth-Studio is clearer than in ComfyUI. Is there a problem with the model conversion script? Open

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

modelscope/DiffSynth-Studio#1273 · 1 comment · Opened Feb 6, 2026
flux2-klein-base-4b训练出错 RuntimeError: FlashAttention only support fp16 and bf16 data type Open

Difficulty 4/5 3-5 days Newbie friendliness 20/100

modelscope/DiffSynth-Studio#1271 · 2 comments · Opened Feb 6, 2026
black color picture in nunchaku Open

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

modelscope/DiffSynth-Studio#1269 · Opened Feb 6, 2026
The Klein LoRA training results lack fine details and are blurry, with performance inferior to both Kontext and Qwen-Image-Edit-2511. Open

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

modelscope/DiffSynth-Studio#1268 · 1 comment · Opened Feb 6, 2026
klein lora: ERROR lora diffusion_model.double_blocks.4.txt_attn.qkv.weight shape '[9216, 3072]' is invalid for input of size 9437184 Open

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

modelscope/DiffSynth-Studio#1267 · Opened Feb 6, 2026
support Skywork UniPic 3.0 Open

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

modelscope/DiffSynth-Studio#1266 · 1 comment · Opened Feb 5, 2026
DiffSynth-Studio 训练的klein lora模型无法在diffuser 框架下加载！ Open

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

modelscope/DiffSynth-Studio#1264 · 2 comments · Opened Feb 5, 2026
问下qwen-image-edit 能用openai方式加载吗 Open

Difficulty 4/5 3-5 days Newbie friendliness 20/100

modelscope/DiffSynth-Studio#1263 · Opened Feb 5, 2026
QwenImage多卡推理，单卡推理特别慢正常吗 Open

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

modelscope/DiffSynth-Studio#1262 · 4 comments · Opened Feb 5, 2026
显存占用明显大于diffusers框架 Open

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

modelscope/DiffSynth-Studio#1257 · 9 comments · Opened Feb 4, 2026
Wan2.2-Fun-A14B-Control-Camera 训练 Open

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

modelscope/DiffSynth-Studio#1254 · 1 comment · Opened Feb 3, 2026
Flux.2训练lora时无法加载transformer模型 Open

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

modelscope/DiffSynth-Studio#1252 · 2 comments · Opened Feb 3, 2026
Qwen-Image controlnet inpaint + controlnet canny Open

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

modelscope/DiffSynth-Studio#1250 · 2 comments · Opened Feb 2, 2026
MOVA: Unified Audio-Video model support? Open

Difficulty 5/5 Over a week Newbie friendliness 15/100

modelscope/DiffSynth-Studio#1249 · 1 comment · Opened Feb 2, 2026
mu训练和推理对齐问题 Open

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

modelscope/DiffSynth-Studio#1248 · 1 comment · Opened Feb 2, 2026
vram_limit is not taking effect Open

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

modelscope/DiffSynth-Studio#1247 · 1 comment · Opened Feb 2, 2026
如何实现文生图和编辑混合训练，我看edit_images设置为空[]，目前框架会报错 Open

Difficulty 3/5 1-2 days Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1239 · 3 comments · Opened Jan 29, 2026
H20 96G显存 qwen-image-edit lora微调为什么还是报错 Open

Difficulty 4/5 3-5 days Newbie friendliness 15/100

modelscope/DiffSynth-Studio#1238 · 1 comment · Opened Jan 29, 2026
怎么进行多卡推理呢 Open

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

modelscope/DiffSynth-Studio#1237 · 4 comments · Opened Jan 29, 2026
pyptoject.tomp issue Open

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

modelscope/DiffSynth-Studio#1236 · Opened Jan 28, 2026
qwen_edit 的metadata格式 Open

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

modelscope/DiffSynth-Studio#1235 · 2 comments · Opened Jan 28, 2026
Z-Image微调 Open

Difficulty 5/5 Over a week Newbie friendliness 15/100

modelscope/DiffSynth-Studio#1234 · 10 comments · Opened Jan 28, 2026
Traceback (most recent call last): from diffsynth import ModelManager, save_video, VideoData ImportError: cannot import name 'ModelManager' from 'diffsynth' Open

Difficulty 2/5 1-3 hours Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1230 · 2 comments · Opened Jan 24, 2026
各位大佬好，是不是目前基于flux-klein edit 微调的lora 推理是不支持的呀？ Open

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

modelscope/DiffSynth-Studio#1225 · 8 comments · Opened Jan 21, 2026
FLUX.2-klein-base-9B微调的lora在comfy中无法读取 Open

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

modelscope/DiffSynth-Studio#1224 · 3 comments · Opened Jan 21, 2026
Wan 2.2 T2V 14B，微调后的模型不支持多卡推理吗？推理出来全是噪声 Open

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

modelscope/DiffSynth-Studio#1216 · 2 comments · Opened Jan 19, 2026
为什么qwen-image-edit进入vae的图像的长和宽要限制为32的倍数呢？ Open

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

modelscope/DiffSynth-Studio#1213 · Opened Jan 19, 2026
lora微调z-image报错q,k和v的dtype不同 Open

Difficulty 4/5 3-5 days Newbie friendliness 20/100

modelscope/DiffSynth-Studio#1212 · Opened Jan 18, 2026
配置要求 Open

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

modelscope/DiffSynth-Studio#1211 · 2 comments · Opened Jan 16, 2026
训练数据集问题 Open

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

modelscope/DiffSynth-Studio#1210 · 1 comment · Opened Jan 16, 2026
请问支持FLUX-2-Klein吗 Open

Difficulty 5/5 Over a week Newbie friendliness 15/100

modelscope/DiffSynth-Studio#1208 · 2 comments · Opened Jan 16, 2026
请问准备支持GLM-Image训练吗 Open

Difficulty 5/5 Over a week Newbie friendliness 15/100

modelscope/DiffSynth-Studio#1206 · 1 comment · Opened Jan 15, 2026
Choice of `lora_target_modules` Open

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

modelscope/DiffSynth-Studio#1203 · 1 reaction · Opened Jan 15, 2026
how to reduce video memory when running lora Open

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

modelscope/DiffSynth-Studio#1201 · 3 comments · Opened Jan 14, 2026
wan2.2 animte 首帧色差 Open

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

modelscope/DiffSynth-Studio#1197 · 4 comments · 2 reactions · Opened Jan 13, 2026
how to use local model weight?? Open

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

modelscope/DiffSynth-Studio#1196 · 1 comment · Opened Jan 13, 2026
Inquiry regarding Full Training speed and First-frame color shift on Wan2.2 TI2V 5B Open

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

modelscope/DiffSynth-Studio#1195 · 3 comments · Opened Jan 12, 2026
Implementation of the 'trajectory_imitation' Open

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

modelscope/DiffSynth-Studio#1192 · Opened Jan 12, 2026
hello，现在已经开源了longcat-video的训练脚本？能否开源longcat video-avatar的训练脚本？ Open

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

modelscope/DiffSynth-Studio#1186 · Opened Jan 9, 2026

modelscope/DiffSynth-Studio

Get new issues in your inbox