repo stringclasses 1
value | github_id int64 1.27B 4.76B | github_node_id stringlengths 18 24 | number int64 8 14.1k | html_url stringlengths 49 53 | api_url stringlengths 59 63 | title stringlengths 1 402 | body stringlengths 1 62.9k ⌀ | state stringclasses 2
values | state_reason stringclasses 4
values | locked bool 2
classes | comments_count int64 0 99 | labels listlengths 0 5 | assignees listlengths 0 5 | created_at stringdate 2022-06-09 16:28:35 2026-06-26 21:21:14 | updated_at stringdate 2022-06-12 22:18:01 2026-06-26 23:42:20 | closed_at stringdate 2022-06-12 22:18:01 2026-06-26 23:42:20 ⌀ | author_association stringclasses 3
values | milestone_title stringclasses 0
values | snapshot_id stringclasses 107
values | extracted_at stringdate 2026-04-07 13:34:13 2026-06-27 00:35:12 | author_login stringlengths 3 28 | author_id int64 1.54k 296M | author_node_id stringlengths 12 20 | author_type stringclasses 3
values | author_site_admin bool 1
class |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
huggingface/diffusers | 3,311,998,871 | I_kwDOHa8MBc7FaReX | 12,125 | https://github.com/huggingface/diffusers/issues/12125 | https://api.github.com/repos/huggingface/diffusers/issues/12125 | CompVis/stable-diffusion-v1-4 is missing fp16 files | Please note that authentication is recommended but still optional to access public models or datasets.
warnings.warn(
model_index.json: 100%
543/543 [00:00<00:00, 30.5kB/s]
/usr/local/lib/python3.11/dist-packages/diffusers/pipelines/pipeline_loading_utils.py:325: FutureWarning: You are loading the variant fp16 from ... | closed | completed | false | 2 | [
"stale"
] | [] | 2025-08-12T00:00:45Z | 2026-01-09T23:07:10Z | 2026-01-09T23:07:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | thepreetam | 137,159,354 | U_kgDOCCziug | User | false |
huggingface/diffusers | 3,312,429,648 | I_kwDOHa8MBc7Fb6pQ | 12,126 | https://github.com/huggingface/diffusers/issues/12126 | https://api.github.com/repos/huggingface/diffusers/issues/12126 | wan2.2 inference error Bug Report: `NoneType` object has no attribute `cache_context` when switching from `transformer` to `transformer2` | ## 🐛 Bug Report: `NoneType` object has no attribute `cache_context` when switching from `transformer` to `transformer2`
### Description
I'm using the `WanPipeline` from the `diffusers` library to generate video frames. When running the pipeline, I encounter an error during inference, specifically when the model swit... | closed | completed | false | 2 | [
"bug"
] | [] | 2025-08-12T03:58:00Z | 2025-08-25T09:28:01Z | 2025-08-25T09:28:01Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Passenger12138 | 47,623,700 | MDQ6VXNlcjQ3NjIzNzAw | User | false |
huggingface/diffusers | 1,441,940,763 | I_kwDOHa8MBc5V8kUb | 1,213 | https://github.com/huggingface/diffusers/issues/1213 | https://api.github.com/repos/huggingface/diffusers/issues/1213 | Does RePaintPipeline work with Stable diffusion? | Hello,
First of all, thank you so much for adding RePaintPipeline. This pipeline works much better than stable diffusion inpainting when I use DDPMs (such as ddpm-ema-bedroom-256 and ddpm-bedroom-256) . However, when I use CompVis/stable-diffusion-v1-4, some bugs appear. For example, I get the following error:
`T... | closed | completed | false | 6 | [] | [
"anton-l"
] | 2022-11-09T11:41:24Z | 2022-12-08T09:48:03Z | 2022-11-17T17:14:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FBehrad | 41,340,554 | MDQ6VXNlcjQxMzQwNTU0 | User | false |
huggingface/diffusers | 3,314,364,371 | I_kwDOHa8MBc7FjS_T | 12,131 | https://github.com/huggingface/diffusers/issues/12131 | https://api.github.com/repos/huggingface/diffusers/issues/12131 | inference with lora suddenly significantly slower | inference with lora suddenly significantly slower than with diffusers==0.33.1 and peft==0.14.0
Used to be 3.9it/s now 2.9it/s
_Originally posted by @burgalon in https://github.com/huggingface/diffusers/pull/11999#issuecomment-3179413365_ | closed | completed | false | 19 | [
"bug",
"stale",
"performance"
] | [] | 2025-08-12T13:48:54Z | 2026-01-09T22:47:56Z | 2026-01-09T22:47:56Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | a-r-r-o-w | 72,266,394 | MDQ6VXNlcjcyMjY2Mzk0 | User | false |
huggingface/diffusers | 3,317,089,788 | I_kwDOHa8MBc7FtsX8 | 12,136 | https://github.com/huggingface/diffusers/issues/12136 | https://api.github.com/repos/huggingface/diffusers/issues/12136 | How to use Diffusers to Convert Safetensors SDXL 1.0 to Onnx? | Hello,
I'm trying to convert a safetensors checkpoint for SDXL to onnx format.
I've tried Optimum already but it fails everytime.
Please help. | closed | completed | false | 0 | [] | [] | 2025-08-13T06:33:22Z | 2025-10-31T03:13:28Z | 2025-10-31T03:13:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CypherpunkSamurai | 66,906,402 | MDQ6VXNlcjY2OTA2NDAy | User | false |
huggingface/diffusers | 3,317,594,525 | I_kwDOHa8MBc7Fvnmd | 12,138 | https://github.com/huggingface/diffusers/issues/12138 | https://api.github.com/repos/huggingface/diffusers/issues/12138 | requirest for the model use | ### Did you like the remote VAE solution?
I like it cause I can use it to generate many nice images.
### What can be improved about the current solution?
/
### What other VAEs you would like to see if the pilot goes well?
faster and high-quality images
### Notify the members of the team
_No response_ | open | null | false | 1 | [
"stale",
"remote-vae"
] | [] | 2025-08-13T09:16:13Z | 2026-01-09T15:13:57Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kjy09 | 22,428,544 | MDQ6VXNlcjIyNDI4NTQ0 | User | false |
huggingface/diffusers | 3,317,662,627 | I_kwDOHa8MBc7Fv4Oj | 12,139 | https://github.com/huggingface/diffusers/issues/12139 | https://api.github.com/repos/huggingface/diffusers/issues/12139 | The IP Adapter provides different conditional guidance for each frame generated by AnimateDiff. | The IP Adapter seems to only provide the same conditional guidance for all frames generated by AnimateDiff. I am unsure whether the IP Adapter can provide different conditional guidance for each frame. | closed | completed | false | 2 | [
"stale",
"consider-for-modular-diffusers"
] | [] | 2025-08-13T09:34:30Z | 2026-01-09T22:48:24Z | 2026-01-09T22:48:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tlemangen | 53,872,583 | MDQ6VXNlcjUzODcyNTgz | User | false |
huggingface/diffusers | 1,441,964,260 | I_kwDOHa8MBc5V8qDk | 1,214 | https://github.com/huggingface/diffusers/issues/1214 | https://api.github.com/repos/huggingface/diffusers/issues/1214 | CLIPTextModel does not work at half precision | ### Describe the bug
RuntimeError exception thrown when using CLIPTextModel.How should I use fp16?
```
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/xxx/anaconda3/envs/yyy/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forwa... | closed | completed | false | 5 | [
"bug"
] | [] | 2022-11-09T11:58:57Z | 2022-11-18T15:52:03Z | 2022-11-18T15:52:03Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chinoll | 22,575,232 | MDQ6VXNlcjIyNTc1MjMy | User | false |
huggingface/diffusers | 3,318,093,157 | I_kwDOHa8MBc7FxhVl | 12,141 | https://github.com/huggingface/diffusers/issues/12141 | https://api.github.com/repos/huggingface/diffusers/issues/12141 | WanVACEPipeline doesn't work with bfloat16/foat16 vae | ### Describe the bug
When I use bfloat16 or float16 VAE computing type - I have error of different input and model types.
On encode step.
### Reproduction
```
# --- VAE / VAE ---
vae = AutoencoderKLWan.from_pretrained(
"Wan-AI/Wan2.1-VACE-14B-diffusers",
subfolder="vae",
torch_dtype=torch... | closed | completed | false | 1 | [
"bug"
] | [] | 2025-08-13T11:35:47Z | 2025-08-13T21:50:51Z | 2025-08-13T21:50:51Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | SlimRG | 39,348,033 | MDQ6VXNlcjM5MzQ4MDMz | User | false |
huggingface/diffusers | 3,322,188,783 | I_kwDOHa8MBc7GBJPv | 12,145 | https://github.com/huggingface/diffusers/issues/12145 | https://api.github.com/repos/huggingface/diffusers/issues/12145 | resize_token_embeddings (?) (different result from new and old version) | ### System Info
model, and embending https://huggingface.co/fofr/sdxl-emoji
detail image: [emoji_20250813_002621_2c661c36-1825-496f-84b4-60ec59a961fe.json](https://github.com/user-attachments/files/21756027/emoji_20250813_002621_2c661c36-1825-496f-84b4-60ec59a961fe.json)
from [requirements.txt](https://github.com/u... | open | null | false | 5 | [
"bug",
"stale"
] | [] | 2025-08-13T13:54:55Z | 2026-01-09T15:13:51Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Red-Owl | 36,891,854 | MDQ6VXNlcjM2ODkxODU0 | User | false |
huggingface/diffusers | 3,322,625,745 | I_kwDOHa8MBc7GCz7R | 12,146 | https://github.com/huggingface/diffusers/issues/12146 | https://api.github.com/repos/huggingface/diffusers/issues/12146 | Wan 2.1 / Wan2.2 Lightning & Sparse Distillation & Tiny VAE Acceleration Methods | <img width="5538" height="2705" alt="Image" src="https://github.com/user-attachments/assets/7bf435e2-999f-40b7-8c26-0381a68c0923" />
**Is your feature request related to a problem? Please describe.**
Some incredible acceleration methods have been released for both Wan2.1 and Wan2.2. These are currently appear to be u... | open | null | false | 18 | [
"stale"
] | [] | 2025-08-14T15:22:59Z | 2026-02-03T15:19:19Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghunkins | 12,562,057 | MDQ6VXNlcjEyNTYyMDU3 | User | false |
huggingface/diffusers | 3,323,258,088 | I_kwDOHa8MBc7GFOTo | 12,147 | https://github.com/huggingface/diffusers/issues/12147 | https://api.github.com/repos/huggingface/diffusers/issues/12147 | Wan2.2 Lightning LoRA Loading Fails | ### Describe the bug
Loading Wan2.2 LightX2V Lightning LoRAs fails.
### Reproduction
```python
import torch
from diffusers import WanImageToVideoPipeline, UniPCMultistepScheduler
from huggingface_hub import hf_hub_download
from diffusers.loaders.lora_conversion_utils import _convert_non_diffusers_wan_lora_to_diffuse... | closed | completed | false | 5 | [
"bug"
] | [] | 2025-08-14T18:54:17Z | 2025-09-12T03:30:26Z | 2025-08-19T16:11:15Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghunkins | 12,562,057 | MDQ6VXNlcjEyNTYyMDU3 | User | false |
huggingface/diffusers | 3,324,142,482 | I_kwDOHa8MBc7GImOS | 12,149 | https://github.com/huggingface/diffusers/issues/12149 | https://api.github.com/repos/huggingface/diffusers/issues/12149 | The diffusers0.34.0 installed by pip is not aligned with the source code | ### Describe the bug
The following files cannot be found in the diffusers0.34.0 installed by pip “diffusers/models/attention_dispatch.py”. But i can see it in the github repository.
<img width="394" height="636" alt="Image" src="https://github.com/user-attachments/assets/2dff3d9f-07ff-447b-a383-a228ad52fcea" />
### ... | closed | completed | false | 2 | [
"bug"
] | [] | 2025-08-15T02:38:11Z | 2025-08-26T05:28:11Z | 2025-08-26T05:28:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LTayfaker | 85,294,970 | MDQ6VXNlcjg1Mjk0OTcw | User | false |
huggingface/diffusers | 3,324,864,556 | I_kwDOHa8MBc7GLWgs | 12,154 | https://github.com/huggingface/diffusers/issues/12154 | https://api.github.com/repos/huggingface/diffusers/issues/12154 | dje | null | closed | not_planned | false | 0 | [
"spam"
] | [] | 2025-08-15T09:50:22Z | 2025-08-17T22:07:27Z | 2025-08-17T22:07:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jcagentco-sudo | 226,686,628 | U_kgDODYL2pA | User | false |
huggingface/diffusers | 3,324,865,578 | I_kwDOHa8MBc7GLWwq | 12,155 | https://github.com/huggingface/diffusers/issues/12155 | https://api.github.com/repos/huggingface/diffusers/issues/12155 | monsieur | null | closed | completed | false | 0 | [] | [] | 2025-08-15T09:50:57Z | 2025-08-15T09:51:17Z | 2025-08-15T09:51:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jcagentco-sudo | 226,686,628 | U_kgDODYL2pA | User | false |
huggingface/diffusers | 3,324,869,059 | I_kwDOHa8MBc7GLXnD | 12,156 | https://github.com/huggingface/diffusers/issues/12156 | https://api.github.com/repos/huggingface/diffusers/issues/12156 | cool man | i am a cool man | closed | completed | false | 1 | [] | [] | 2025-08-15T09:52:44Z | 2025-08-15T09:53:00Z | 2025-08-15T09:53:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jcagentco-sudo | 226,686,628 | U_kgDODYL2pA | User | false |
huggingface/diffusers | 3,325,084,126 | I_kwDOHa8MBc7GMMHe | 12,157 | https://github.com/huggingface/diffusers/issues/12157 | https://api.github.com/repos/huggingface/diffusers/issues/12157 | When training the Qwen-Image model using --bnb_quantization_config_path, the following error occurred: 'torch.device' object is not iterable. | ### Describe the bug
When I perform post-training on Qwen-Image using diffusers/examples/dreambooth/train_dreambooth_lora_qwen_image.py, I encounter the error: 'torch.device' object is not iterable. The difference from the official example script is that I added the **_--bnb_quantization_config_path_** parameter.
```... | closed | completed | false | 1 | [
"bug"
] | [] | 2025-08-15T11:39:19Z | 2025-08-18T08:26:18Z | 2025-08-18T08:26:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DarwinYang | 733,525 | MDQ6VXNlcjczMzUyNQ== | User | false |
huggingface/diffusers | 3,326,075,546 | I_kwDOHa8MBc7GP-Ka | 12,159 | https://github.com/huggingface/diffusers/issues/12159 | https://api.github.com/repos/huggingface/diffusers/issues/12159 | LoRA loading compatibility with some Qwen Image LoRAs | ### Describe the bug
Some Qwen Image LoRAs are not compatible with diffusers.
e.g.: https://huggingface.co/itspoidaman/qwenglitch
### Reproduction
with `load_lora_weights` from https://huggingface.co/itspoidaman/qwenglitch we get:
```
ValueError: Target modules {'lora_unet_transformer_blocks_59_attn_to_add_out', '... | closed | completed | false | 6 | [
"bug"
] | [] | 2025-08-15T19:03:37Z | 2025-08-18T15:26:29Z | 2025-08-18T15:26:29Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | apolinario | 788,417 | MDQ6VXNlcjc4ODQxNw== | User | false |
huggingface/diffusers | 3,326,912,899 | I_kwDOHa8MBc7GTKmD | 12,162 | https://github.com/huggingface/diffusers/issues/12162 | https://api.github.com/repos/huggingface/diffusers/issues/12162 | Not compatible with Flash ATTN 2.8.2 | Module 'diffusers' load failed. If you don't have it installed, do it:
pip install diffusers
After I installed flash attn 2.8.2, both diffusers 0.34 and dev failed to start | closed | completed | false | 8 | [
"stale"
] | [] | 2025-08-16T05:40:25Z | 2026-01-09T22:53:06Z | 2026-01-09T22:53:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hargret | 161,938,407 | U_kgDOCab75w | User | false |
huggingface/diffusers | 3,328,057,282 | I_kwDOHa8MBc7GXh_C | 12,165 | https://github.com/huggingface/diffusers/issues/12165 | https://api.github.com/repos/huggingface/diffusers/issues/12165 | Failed to finetune the pre-trained model of 'stable-diffusion-v1-4' on image inpainting task | I finetuned the pre-trained model of 'stable-diffusion-inpainting' on image inpainting task, and all work well as the model is trained on image inpainting. But when I finetuned with the pre-trained model of 'stable-diffusion-v1-4' which is trained on text-to-image, the loss is NaN and the result is pure black.
As the... | closed | completed | false | 7 | [] | [] | 2025-08-17T07:15:36Z | 2025-09-07T09:35:38Z | 2025-09-04T11:22:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | micklexqg | 13,776,012 | MDQ6VXNlcjEzNzc2MDEy | User | false |
huggingface/diffusers | 3,329,258,885 | I_kwDOHa8MBc7GcHWF | 12,173 | https://github.com/huggingface/diffusers/issues/12173 | https://api.github.com/repos/huggingface/diffusers/issues/12173 | Wan i2i with latent upscale | ### Model/Pipeline/Scheduler description
I made a wan i2i pipeline that can also somewhat handle upscaling (not really a true upscaler since it's very creative). It only uses the low noise transformer and can generate good results in 20 steps. It's useful to transform low quality pictures and to change the overall vib... | open | null | false | 1 | [
"stale"
] | [] | 2025-08-18T05:49:47Z | 2026-01-09T15:13:27Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | luke14free | 166,602 | MDQ6VXNlcjE2NjYwMg== | User | false |
huggingface/diffusers | 3,329,298,047 | I_kwDOHa8MBc7GcQ5_ | 12,174 | https://github.com/huggingface/diffusers/issues/12174 | https://api.github.com/repos/huggingface/diffusers/issues/12174 | 'super' object has no attribute '__getattr__' | ### Describe the bug
#### Installed diffusers version：
pip install -e . -i https://mirrors.aliyun.com/pypi/simple/ --no-build-isolation --timeout=500
Looking in indexes: https://mirrors.aliyun.com/pypi/simple/
Obtaining file:///app/diffusers
Checking if build backend supports build_editable ... done
Preparing ed... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2025-08-18T06:07:41Z | 2026-01-09T22:55:44Z | 2026-01-09T22:55:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | White-Friday | 37,582,362 | MDQ6VXNlcjM3NTgyMzYy | User | false |
huggingface/diffusers | 3,329,338,882 | I_kwDOHa8MBc7Gca4C | 12,175 | https://github.com/huggingface/diffusers/issues/12175 | https://api.github.com/repos/huggingface/diffusers/issues/12175 | Bad Image Quality generated by Qwen Image | ### Describe the bug
I installed the latest version of the diffusers library from the main branch, and used the inference code example in the application for inference, but the effect was very poor, as shown in the following figure:
<img width="1024" height="1024" alt="Image" src="https://github.com/user-attachments/... | closed | completed | false | 15 | [
"bug"
] | [] | 2025-08-18T06:22:39Z | 2025-08-27T11:15:52Z | 2025-08-21T16:21:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vagitablebirdcode | 42,081,162 | MDQ6VXNlcjQyMDgxMTYy | User | false |
huggingface/diffusers | 3,332,762,002 | I_kwDOHa8MBc7GpemS | 12,185 | https://github.com/huggingface/diffusers/issues/12185 | https://api.github.com/repos/huggingface/diffusers/issues/12185 | What's the difference between DreamBooth LoRa and traditional LoRa? | I see a lot of examples using DreamBooth LoRa training code. What's the difference between this and traditional LoRa training? Can this DreamBooth LoRa training code be adapted to standard SFT LoRa code? Does disabling with_prior_preservation return normal LoRa training? | closed | completed | true | 4 | [
"stale"
] | [] | 2025-08-19T03:32:30Z | 2026-01-09T22:56:05Z | 2026-01-09T22:56:05Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MetaInsight7 | 30,329,784 | MDQ6VXNlcjMwMzI5Nzg0 | User | false |
huggingface/diffusers | 3,332,978,858 | I_kwDOHa8MBc7GqTiq | 12,186 | https://github.com/huggingface/diffusers/issues/12186 | https://api.github.com/repos/huggingface/diffusers/issues/12186 | `WanVACETransformer3DModel` does not inherit `AttentionMixin` while `WanTransformer3DModel` does | Hello,
While installing the latest version of diffusers from source and implementing a custom Wan VACE pipeline, I noticed a potential inconsistency that I hope may be helpful to point out.
`WanTransformer3DModel` inherits the following:
`ModelMixin`, `ConfigMixin`, `PeftAdapterMixin`, `FromOriginalModelMixin`, `Cach... | closed | completed | false | 4 | [] | [] | 2025-08-19T05:49:58Z | 2025-09-11T03:35:37Z | 2025-09-09T12:14:07Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | J4BEZ | 43,560,917 | MDQ6VXNlcjQzNTYwOTE3 | User | false |
huggingface/diffusers | 3,333,507,455 | I_kwDOHa8MBc7GsUl_ | 12,189 | https://github.com/huggingface/diffusers/issues/12189 | https://api.github.com/repos/huggingface/diffusers/issues/12189 | When using the qwen-image model to generate images, diffusers takes twice as long as diffsynth-studio | <img width="2056" height="258" alt="Image" src="https://github.com/user-attachments/assets/2b7b927a-ae3f-4cf1-a40e-3cdbd8b7bc55" />
diffsynth-studio took 9 minutes and 58 seconds
<img width="2058" height="280" alt="Image" src="https://github.com/user-attachments/assets/53689ec2-2de6-4639-b502-9ac7c83b2f3a" />
diffusrs... | closed | completed | true | 3 | [] | [] | 2025-08-19T08:51:49Z | 2026-01-09T22:57:05Z | 2026-01-09T22:57:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | gelinhua | 10,258,602 | MDQ6VXNlcjEwMjU4NjAy | User | false |
huggingface/diffusers | 3,336,519,981 | I_kwDOHa8MBc7G30Et | 12,195 | https://github.com/huggingface/diffusers/issues/12195 | https://api.github.com/repos/huggingface/diffusers/issues/12195 | Version 0.35.0 AutoencoderKL failed with PyTorch 2.3.1: AutoAttributeError: module 'torch.library' has no attribute 'custom_op' | ### Describe the bug
This should be an old issue, but it appears again in version 0.35.0, maybe also in 0.35.1.
Maybe need to protect the import again if PyTorch 2.3.1 is still in the supporting list?
### Reproduction
from diffusers import AutoencoderKL
import AutoencoderKL will lead to : AutoAttributeError: module... | closed | completed | false | 5 | [
"bug"
] | [] | 2025-08-20T04:27:21Z | 2025-09-04T11:51:07Z | 2025-08-23T07:22:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | skedin | 5,986,503 | MDQ6VXNlcjU5ODY1MDM= | User | false |
huggingface/diffusers | 3,337,836,773 | I_kwDOHa8MBc7G81jl | 12,198 | https://github.com/huggingface/diffusers/issues/12198 | https://api.github.com/repos/huggingface/diffusers/issues/12198 | Add support for guidance-interval during sampling | ### Issue
**Describe the solution you'd like.**
Integrate source code from [kynkaat/guidance-interval](https://github.com/kynkaat/guidance-interval) into `huggingface/diffusers` in `callbacks.py`.
**Additional context.**
Although, we have a [pr#7761](https://github.com/huggingface/diffusers/pull/7761) provide a so... | closed | completed | false | 4 | [] | [] | 2025-08-20T11:56:13Z | 2025-08-20T22:09:07Z | 2025-08-20T22:09:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Bili-Sakura | 108,928,239 | U_kgDOBn4c7w | User | false |
huggingface/diffusers | 1,313,466,303 | I_kwDOHa8MBc5OSee_ | 122 | https://github.com/huggingface/diffusers/issues/122 | https://api.github.com/repos/huggingface/diffusers/issues/122 | raise the problem No module named 'exceptions' When train the model | Traceback (most recent call last):
File "scripts/train.py", line 1, in <module>
import diffuser.utils as utils
File "/mnt/d/RL/UCSC/diffuser/diffuser/utils/__init__.py", line 4, in <module>
from .setup import *
File "/mnt/d/RL/UCSC/diffuser/diffuser/utils/setup.py", line 6, in <module>
from tap ... | closed | completed | false | 6 | [] | [] | 2022-07-21T15:47:57Z | 2022-09-13T15:34:34Z | 2022-09-13T15:34:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Lynex-01 | 64,463,528 | MDQ6VXNlcjY0NDYzNTI4 | User | false |
huggingface/diffusers | 1,442,408,617 | I_kwDOHa8MBc5V-Wip | 1,220 | https://github.com/huggingface/diffusers/issues/1220 | https://api.github.com/repos/huggingface/diffusers/issues/1220 | FID on CIFAR10 does not match literature | ### Describe the bug
I used the unconditional training script from examples to train a model on CIFAR10, using default parameters (except using 1024 steps). After around 250k updates (batch size 128, 4 GPUs), the samples look decent but when doing sampling with DDIM with 128 steps and eta=0, the sampled images give ... | closed | completed | false | 13 | [
"bug",
"stale"
] | [] | 2022-11-09T16:38:31Z | 2023-07-18T10:52:15Z | 2023-01-21T15:03:13Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lukovnikov | 1,732,910 | MDQ6VXNlcjE3MzI5MTA= | User | false |
huggingface/diffusers | 3,340,146,034 | I_kwDOHa8MBc7HFpVy | 12,202 | https://github.com/huggingface/diffusers/issues/12202 | https://api.github.com/repos/huggingface/diffusers/issues/12202 | [Features] Planning to support TensorRT for Flux Pipelines | 🚀 Hi team,
I’m planning to code all Flux Pipelines with TensorRT support.
My plan is to create a new subfolder: `flux-tensorrt` inside the `examples` folder.
Pipelines to be supported:
```
FluxPipeline
FluxImg2ImgPipeline
FluxInpaintPipeline
FluxFillPipeline
FluxKontextPipeline
FluxKontextInpaintPipeline
```
✅ Init... | open | null | false | 3 | [
"stale"
] | [] | 2025-08-21T03:23:19Z | 2026-01-09T15:12:56Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vuongminh1907 | 119,489,204 | U_kgDOBx9CtA | User | false |
huggingface/diffusers | 3,341,209,171 | I_kwDOHa8MBc7HJs5T | 12,208 | https://github.com/huggingface/diffusers/issues/12208 | https://api.github.com/repos/huggingface/diffusers/issues/12208 | Transformer2DModel onnx convertation problems | Hi! I am trying to convert Transformer2DModel to onnx and cannot solve some problems.
I am trying to export UNet model with next architecture
```python
Transformer2DModel(
(pos_embed): PatchEmbed(
(proj): Conv2d(96, 1584, kernel_size=(2, 2), stride=(2, 2))
)
(transformer_blocks): ModuleList(
(0-23): 24 x ... | open | null | false | 1 | [
"stale"
] | [] | 2025-08-21T10:30:50Z | 2026-01-09T15:12:49Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Voveka98 | 34,947,202 | MDQ6VXNlcjM0OTQ3MjAy | User | false |
huggingface/diffusers | 3,341,742,683 | I_kwDOHa8MBc7HLvJb | 12,210 | https://github.com/huggingface/diffusers/issues/12210 | https://api.github.com/repos/huggingface/diffusers/issues/12210 | [Feature Request] Settability of Multiple Attention Types in Attention Dispatcher | Really great release, especially in terms of usability for other attention types natively!
**Is your feature request related to a problem? Please describe.**
AFAIU, currently in Attention Dispatcher, we cannot set one attention type to `attn1` and another one to `attn2` at the `transformer` level. In its original impl... | closed | completed | false | 1 | [] | [] | 2025-08-21T13:21:25Z | 2025-08-23T13:20:56Z | 2025-08-21T15:20:34Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tolgacangoz | 46,008,593 | MDQ6VXNlcjQ2MDA4NTkz | User | false |
huggingface/diffusers | 3,344,856,231 | I_kwDOHa8MBc7HXnSn | 12,216 | https://github.com/huggingface/diffusers/issues/12216 | https://api.github.com/repos/huggingface/diffusers/issues/12216 | Qwen-Image-Edit Inferior Results Compared to ComfyUI | ### Describe the bug
I am trying to do multi-image editing with Qwen-Image-Edit (It is a simplified version of [this](https://x.com/hellorob/status/1958197227135906087)). The ComfyUI workflow and diffusers script are shared below for reproducibility. I am using the same (unquantized) models and the same parameters.
T... | open | null | false | 21 | [
"bug",
"stale"
] | [] | 2025-08-22T09:42:53Z | 2026-06-10T04:29:12Z | null | NONE | null | 20260610T043530Z | 2026-06-10T04:35:30Z | enesmsahin | 36,894,859 | MDQ6VXNlcjM2ODk0ODU5 | User | false |
huggingface/diffusers | 3,346,692,898 | I_kwDOHa8MBc7Hensi | 12,221 | https://github.com/huggingface/diffusers/issues/12221 | https://api.github.com/repos/huggingface/diffusers/issues/12221 | [Looking for community contribution] support DiffSynth Controlnet in diffusers | ### Model/Pipeline/Scheduler description
Hi!
We want to add first party support for DiffSynth controlnet in diffusers, and we are looking for some help from the community!
Let me know if you're interested!
### Open source status
- [x] The model implementation is available.
- [x] The model weights are available (... | open | null | false | 8 | [
"help wanted",
"Good second issue",
"contributions-welcome"
] | [] | 2025-08-22T20:49:18Z | 2026-03-15T05:54:47Z | null | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 3,346,741,986 | I_kwDOHa8MBc7Hezri | 12,222 | https://github.com/huggingface/diffusers/issues/12222 | https://api.github.com/repos/huggingface/diffusers/issues/12222 | [Contribution welcome] adding a fast test for Qwen-Image Controlnet Pipeline | We are looking for help from community to add a fast time for this PR
https://github.com/huggingface/diffusers/pull/12215
You can add a file under this folder:
https://github.com/huggingface/diffusers/tree/main/tests/pipelines/qwenimage
You can reference other tests we added for qwee pipelines [example](https://git... | closed | completed | false | 6 | [
"good first issue",
"help wanted",
"contributions-welcome"
] | [] | 2025-08-22T21:04:50Z | 2025-08-25T01:58:59Z | 2025-08-25T01:58:59Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 3,348,307,931 | I_kwDOHa8MBc7Hkx_b | 12,224 | https://github.com/huggingface/diffusers/issues/12224 | https://api.github.com/repos/huggingface/diffusers/issues/12224 | Option to cache text embedding vectors | **Is your feature request related to a problem? Please describe.**
Text to image models require generation of text embeddings as a first step. While this step is relatively quick if VRAM amount is not a constraint but it often is and therefore the text model(s) need to loaded, unload, then transformer loaded which lead... | closed | completed | true | 4 | [
"stale"
] | [] | 2025-08-23T14:28:36Z | 2026-01-09T23:10:20Z | 2026-01-09T23:10:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | innokean | 115,506,081 | U_kgDOBuJ7oQ | User | false |
huggingface/diffusers | 3,349,442,604 | I_kwDOHa8MBc7HpHAs | 12,227 | https://github.com/huggingface/diffusers/issues/12227 | https://api.github.com/repos/huggingface/diffusers/issues/12227 | [Feature] Integrate 'CHORDS: Diffusion Sampling Accelerator with Multi-core Hierarchical ODE Solvers' | I find a latest, general, training-free, and model-agnostic acceleration strategy via multicore parallelism for image generation and video generation.
[Source Code](https://github.com/hanjq17/CHORDS)
```bibtex
@misc{hanCHORDSDiffusionSampling2025,
title = {{{CHORDS}}: {{Diffusion Sampling Accelerator}} with {{Multi... | open | null | false | 1 | [
"stale"
] | [] | 2025-08-24T11:43:42Z | 2026-01-09T15:12:35Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Bili-Sakura | 108,928,239 | U_kgDOBn4c7w | User | false |
huggingface/diffusers | 3,351,863,273 | I_kwDOHa8MBc7HyV_p | 12,235 | https://github.com/huggingface/diffusers/issues/12235 | https://api.github.com/repos/huggingface/diffusers/issues/12235 | Support fun control variants for Wan | It would be cool to support the control variants of Wan: https://huggingface.co/collections/alibaba-pai/wan22-fun-68958eabec343b948f1225c5.
https://huggingface.co/alibaba-pai/Wan2.2-Fun-A14B-Control has some very nice examples.
Checking with @a-r-r-o-w if this is already in the works. | open | null | false | 4 | [
"stale",
"contributions-welcome"
] | [] | 2025-08-25T13:28:04Z | 2026-02-03T15:18:49Z | null | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 3,354,390,997 | I_kwDOHa8MBc7H7_HV | 12,239 | https://github.com/huggingface/diffusers/issues/12239 | https://api.github.com/repos/huggingface/diffusers/issues/12239 | Support for InfiniteTalk | ### Model/Pipeline/Scheduler description
https://huggingface.co/MeiGen-AI/InfiniteTalk is a wonderful audio driven video generation model and can also support infinite frame , which is based on wan2.1. The demo and user's workflow is also awesome. some examples: https://www.runninghub.cn/ai-detail/195843862495620301... | open | null | false | 1 | [
"help wanted",
"New pipeline/model",
"contributions-welcome"
] | [] | 2025-08-26T06:57:43Z | 2025-09-05T00:18:46Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | supermeng | 3,438,462 | MDQ6VXNlcjM0Mzg0NjI= | User | false |
huggingface/diffusers | 3,355,426,976 | I_kwDOHa8MBc7H_8Cg | 12,241 | https://github.com/huggingface/diffusers/issues/12241 | https://api.github.com/repos/huggingface/diffusers/issues/12241 | WAN2.1 FLF2V: Incorrect MASK Creation???? | Hello! I think that it is maybe error. (Or not, please explain it for me!!)
In **WanImageToVideoPipeline** class in `pipline_wan_i2v.py`,
<img width="868" height="243" alt="Image" src="https://github.com/user-attachments/assets/8108a9e9-8632-44a1-93b8-abd9ae6a22cd" />
(the code is the part of `prepare_latents` funct... | open | null | false | 2 | [
"stale"
] | [] | 2025-08-26T12:23:09Z | 2026-01-09T15:12:24Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | KyujinHan | 98,331,298 | U_kgDOBdxqog | User | false |
huggingface/diffusers | 3,364,061,337 | I_kwDOHa8MBc7Ig4CZ | 12,253 | https://github.com/huggingface/diffusers/issues/12253 | https://api.github.com/repos/huggingface/diffusers/issues/12253 | stabilityai/stable-diffusion-2-base is missing fp16 files' so that the correct variant file can be added | FutureWarning: You are loading the variant fp16 from stabilityai/stable-diffusion-2-base via `revision='fp16'`. This behavior is deprecated and will be removed in diffusers v1. One should use `variant='fp16'` instead. However, it appears that stabilityai/stable-diffusion-2-base currently does not have the required var... | closed | completed | false | 2 | [
"stale"
] | [] | 2025-08-28T17:05:30Z | 2026-01-09T23:21:58Z | 2026-01-09T23:21:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aryaIndia-dev | 229,145,099 | U_kgDODah6Cw | User | false |
huggingface/diffusers | 3,364,294,649 | I_kwDOHa8MBc7Ihw_5 | 12,255 | https://github.com/huggingface/diffusers/issues/12255 | https://api.github.com/repos/huggingface/diffusers/issues/12255 | [BUG] Misleading ValueError when subclassing StableDiffusionImg2ImgPipeline with a mismatched __init__ signature | ### Describe the bug
When subclassing diffusers.StableDiffusionImg2ImgPipeline, if the subclass's __init__ signature does not include the requires_safety_checker: bool = True argument, the default .from_pretrained() loader raises a confusing and indirect ValueError.
The official documentation for StableDiffusionImg2I... | closed | completed | false | 2 | [
"bug"
] | [] | 2025-08-28T18:31:14Z | 2025-08-30T07:41:16Z | 2025-08-30T07:41:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | BoostZhu | 148,531,048 | U_kgDOCNpnaA | User | false |
huggingface/diffusers | 3,365,892,623 | I_kwDOHa8MBc7In3IP | 12,257 | https://github.com/huggingface/diffusers/issues/12257 | https://api.github.com/repos/huggingface/diffusers/issues/12257 | [Looking for community contribution] support Wan 2.2 S2V: an audio-driven cinematic video generation model | We're super excited about the Wan 2.2 S2V (Speech-to-Video) model and want to get it integrated into Diffusers! This would be an amazing addition, and we're looking for experienced community contributors to help make this happen.
- **Project Page**: https://humanaigc.github.io/wan-s2v-webpage/
- **Source Code**: htt... | open | null | false | 4 | [
"help wanted",
"Good second issue",
"contributions-welcome"
] | [] | 2025-08-29T08:04:43Z | 2026-04-03T17:57:58Z | null | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 3,377,328,450 | I_kwDOHa8MBc7JTfFC | 12,270 | https://github.com/huggingface/diffusers/issues/12270 | https://api.github.com/repos/huggingface/diffusers/issues/12270 | From 0.35.0: DLL load failed while importing libtriton: A dynamic link library (DLL) initialization routine failed. | ### Describe the bug
From 0.35.0, Triton seems to be required to run Diffusers, however no matter what version or whl I dig up, I end up with the above-mentioned error. Downgrading to 0.34.0, it works as it used to.
The error has stalled the development of my Diffusers based free add-on for Blender: Pallaidium. And ... | open | null | false | 13 | [
"bug",
"stale"
] | [] | 2025-09-02T20:01:24Z | 2026-02-03T15:18:41Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 3,378,427,050 | I_kwDOHa8MBc7JXrSq | 12,273 | https://github.com/huggingface/diffusers/issues/12273 | https://api.github.com/repos/huggingface/diffusers/issues/12273 | New HunyuanVideoI2V PR changes to transformer_hunyuan_video.py | Hello!
The [HunyuanVideoI2V PR](https://github.com/huggingface/diffusers/commit/8907a70a366c96b2322656f57b24e442ea392c7b) broke down the HunyuanVideo example code in xDiT causing an error at this line:
https://github.com/xdit-project/xDiT/blob/main/examples/hunyuan_video_usp_example.py#L79
```
temb = self.ti... | closed | completed | false | 5 | [
"stale"
] | [] | 2025-09-03T06:03:44Z | 2026-01-09T23:51:53Z | 2026-01-09T23:51:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eppaneamd | 186,286,759 | U_kgDOCxqCpw | User | false |
huggingface/diffusers | 3,378,868,226 | I_kwDOHa8MBc7JZXAC | 12,276 | https://github.com/huggingface/diffusers/issues/12276 | https://api.github.com/repos/huggingface/diffusers/issues/12276 | The image is blurry. | How to solve image blurriness during fine-tuning? | closed | completed | false | 1 | [
"stale"
] | [] | 2025-09-03T08:29:38Z | 2026-01-09T23:26:35Z | 2026-01-09T23:26:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sucessfullys | 152,789,949 | U_kgDOCRtjvQ | User | false |
huggingface/diffusers | 1,442,778,588 | I_kwDOHa8MBc5V_w3c | 1,228 | https://github.com/huggingface/diffusers/issues/1228 | https://api.github.com/repos/huggingface/diffusers/issues/1228 | Improve Dreambooth example documentation | The current [Dreambooth example](https://github.com/huggingface/diffusers/tree/main/examples/dreambooth) is difficult to follow.
We're told to:
> Initialize an Accelerate environment with:
accelerate config
However, "accelerate config" asks utterly baffling questions (to a neophyte), like
> "Do you want to ... | closed | completed | false | 12 | [] | [
"williamberman",
"patil-suraj"
] | 2022-11-09T21:38:26Z | 2022-12-06T02:26:32Z | 2022-12-06T02:26:32Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | exo-pla-net | 47,930,742 | MDQ6VXNlcjQ3OTMwNzQy | User | false |
huggingface/diffusers | 3,381,423,677 | I_kwDOHa8MBc7JjG49 | 12,282 | https://github.com/huggingface/diffusers/issues/12282 | https://api.github.com/repos/huggingface/diffusers/issues/12282 | Questions about train_text_to_image_lora_sdxl.py | ### Describe the bug
In this code chunk:
```
# Adapted from pipelines.StableDiffusionXLPipeline.encode_prompt
def encode_prompt(batch, text_encoders, tokenizers, proportion_empty_prompts, caption_column, is_train=True):
prompt_embeds_list = []
prompt_batch = batch[caption_column]
captions = []
for cap... | closed | duplicate | false | 0 | [
"bug"
] | [] | 2025-09-03T21:59:48Z | 2025-09-03T22:24:21Z | 2025-09-03T22:24:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Light-yzc | 61,189,042 | MDQ6VXNlcjYxMTg5MDQy | User | false |
huggingface/diffusers | 3,383,908,654 | I_kwDOHa8MBc7Jslku | 12,284 | https://github.com/huggingface/diffusers/issues/12284 | https://api.github.com/repos/huggingface/diffusers/issues/12284 | Attention test seems uncompatible with latest xformers | https://github.com/huggingface/diffusers/blob/764b62473ac10afd9c52e6b3f3f528f719bc7a34/src/diffusers/models/attention.py#L244
It seems that the latest xformers has the memory_efficient_attention function under xops.ops and not xops in this case.
xops.__version__
'0.0.32.post2'
Maybe "import xformers as xops" should ... | open | null | false | 6 | [
"stale"
] | [] | 2025-09-04T13:54:09Z | 2026-02-03T15:18:37Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jovp | 11,956,862 | MDQ6VXNlcjExOTU2ODYy | User | false |
huggingface/diffusers | 3,388,747,701 | I_kwDOHa8MBc7J_C-1 | 12,292 | https://github.com/huggingface/diffusers/issues/12292 | https://api.github.com/repos/huggingface/diffusers/issues/12292 | Why use torch.repeat instead of torch.repeat_interleave in train_dreambooth_lora_sdxl | ### Describe the bug
### in `train_dreambooth_lora_sdxl.py`
### you can see those codes:
```python
if not args.train_text_encoder:
unet_added_conditions = {
"time_ids": add_time_ids,
"text_embeds": unet_add_text_embeds.repeat(elems_to_repeat_tex... | open | null | false | 3 | [
"bug",
"stale"
] | [] | 2025-09-05T20:52:41Z | 2026-02-03T15:18:32Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Light-yzc | 61,189,042 | MDQ6VXNlcjYxMTg5MDQy | User | false |
huggingface/diffusers | 3,389,168,319 | I_kwDOHa8MBc7KApq_ | 12,294 | https://github.com/huggingface/diffusers/issues/12294 | https://api.github.com/repos/huggingface/diffusers/issues/12294 | [Qwen-image] encoder_hidden_states_mask is not used | ### Describe the bug
During text encoding stage, the text encoder will produce encoder_hidden_states_mask to mask some regions. It usually happens during training with batch size more than 1, to make sure that the produce text embedding has the same sequence length.
However, when I check the following code, it seems ... | closed | completed | false | 26 | [
"bug"
] | [] | 2025-09-06T01:39:06Z | 2026-02-02T04:59:45Z | 2026-02-02T04:59:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JingyeChen-Canva | 222,557,404 | U_kgDODUP03A | User | false |
huggingface/diffusers | 3,390,083,196 | I_kwDOHa8MBc7KEJB8 | 12,295 | https://github.com/huggingface/diffusers/issues/12295 | https://api.github.com/repos/huggingface/diffusers/issues/12295 | Qwen Image pipeline, Various | ### Describe the bug
First of all, thank you for providing this pipeline! It makes it much easier to build tools on top of these models.
While doing that, I ran across various issues, big and small. I'll list them here - let me know if some of them should be a separate issue.
1.
Here:
https://github.com/huggingface/... | open | null | false | 9 | [
"bug",
"stale"
] | [] | 2025-09-06T12:25:31Z | 2026-02-03T15:18:29Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dxqb | 183,307,934 | U_kgDOCu0Ong | User | false |
huggingface/diffusers | 3,390,723,598 | I_kwDOHa8MBc7KGlYO | 12,296 | https://github.com/huggingface/diffusers/issues/12296 | https://api.github.com/repos/huggingface/diffusers/issues/12296 | Pixar-style red race car with big eyes and smile, slipping on banana peel, spinning, then dancing funny on racetrack, smooth animation, bright colors, cinematic lighting |  | closed | completed | false | 1 | [
"stale"
] | [] | 2025-09-06T18:54:24Z | 2026-01-09T23:37:29Z | 2026-01-09T23:37:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | finofirdaus01-web | 230,763,234 | U_kgDODcEq4g | User | false |
huggingface/diffusers | 3,392,824,403 | I_kwDOHa8MBc7KOmRT | 12,299 | https://github.com/huggingface/diffusers/issues/12299 | https://api.github.com/repos/huggingface/diffusers/issues/12299 | SD3 failing with tensor mismatch when using negative prompt embeddings & num_images_per_prompt > 1 | ### Describe the bug
Hi there
I am attempting to get around the 77 CLIP token limit when using SD3 (Medium). To do this, I am using a package called [sd_embed](https://github.com/xhinker/sd_embed/tree/main) which chunks up my positive and negative prompts and embeds them. Rather than passing the prompt strings to the... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2025-09-08T06:36:35Z | 2026-01-09T23:38:54Z | 2026-01-09T23:38:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | danielamassiceti | 15,345,596 | MDQ6VXNlcjE1MzQ1NTk2 | User | false |
huggingface/diffusers | 1,313,470,512 | I_kwDOHa8MBc5OSfgw | 123 | https://github.com/huggingface/diffusers/issues/123 | https://api.github.com/repos/huggingface/diffusers/issues/123 | Add an "About" to the repo | It would be nice to add an about section to this repo:
 | closed | completed | false | 3 | [] | [] | 2022-07-21T15:51:07Z | 2022-07-21T17:36:05Z | 2022-07-21T17:36:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | feconroses | 10,597,153 | MDQ6VXNlcjEwNTk3MTUz | User | false |
huggingface/diffusers | 3,393,153,841 | I_kwDOHa8MBc7KP2sx | 12,300 | https://github.com/huggingface/diffusers/issues/12300 | https://api.github.com/repos/huggingface/diffusers/issues/12300 | add OminiControl | Source Code: https://github.com/Yuanshi9815/OminiControl
Paper: http://arxiv.org/abs/2411.15098
```bibtex
@article{tan2025ominicontrol,
title={OminiControl: Minimal and Universal Control for Diffusion Transformer},
author={Tan, Zhenxiong and Liu, Songhua and Yang, Xingyi and Xue, Qiaochu and Wang, Xinchao},
book... | closed | completed | false | 1 | [] | [] | 2025-09-08T08:25:43Z | 2025-12-30T12:30:42Z | 2025-12-30T12:30:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Bili-Sakura | 108,928,239 | U_kgDOBn4c7w | User | false |
huggingface/diffusers | 3,401,621,567 | I_kwDOHa8MBc7KwKA_ | 12,308 | https://github.com/huggingface/diffusers/issues/12308 | https://api.github.com/repos/huggingface/diffusers/issues/12308 | Support `flash-attn` kernel support for non-Hopper GPUs | **Is your feature request related to a problem? Please describe.**
In #12236 , flash-attn-3 support was introduced for diffusers using the kernel implementation. I believe that adding the [kernels-community/flash-attn](https://huggingface.co/kernels-community/flash-attn) can help users who don't have access to Hopper G... | closed | completed | false | 8 | [
"contributions-welcome"
] | [] | 2025-09-10T09:10:22Z | 2025-12-04T08:10:05Z | 2025-12-04T08:10:05Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ParagEkbote | 69,567,729 | MDQ6VXNlcjY5NTY3NzI5 | User | false |
huggingface/diffusers | 1,442,819,000 | I_kwDOHa8MBc5V_6u4 | 1,231 | https://github.com/huggingface/diffusers/issues/1231 | https://api.github.com/repos/huggingface/diffusers/issues/1231 | K-LMS scheduler in Flax | ### Describe the bug
Hello everyone! So far very happy with the Flax implementation, great work!
However, I've been struggling to get K-LMS working. Here's what I've tried so far:
```python
import jax
import numpy as np
from flax.jax_utils import replicate
from flax.training.common_utils import shard
from t... | closed | completed | false | 2 | [
"bug",
"stale"
] | [
"pcuenca"
] | 2022-11-09T22:10:56Z | 2022-12-18T15:03:12Z | 2022-12-18T15:03:12Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mar-muel | 19,345,805 | MDQ6VXNlcjE5MzQ1ODA1 | User | false |
huggingface/diffusers | 3,402,996,008 | I_kwDOHa8MBc7K1Zko | 12,311 | https://github.com/huggingface/diffusers/issues/12311 | https://api.github.com/repos/huggingface/diffusers/issues/12311 | Flux DEV - bad output quality (bad seeds?) | # Flux.1 [DEV] Model Quality Issues Report
I don't know if I'm the only one with this problem, because I wasn't able to find any other mention of it. But some of the outputs from Flux.1 [DEV] model are of low quality. This either means blurry, or low resolution (the real image resolution is correct though).
I can see... | open | null | false | 7 | [
"stale"
] | [] | 2025-09-10T15:50:19Z | 2026-02-03T15:18:22Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Mahrkeenerh | 43,829,749 | MDQ6VXNlcjQzODI5NzQ5 | User | false |
huggingface/diffusers | 3,407,359,985 | I_kwDOHa8MBc7LGC_x | 12,319 | https://github.com/huggingface/diffusers/issues/12319 | https://api.github.com/repos/huggingface/diffusers/issues/12319 | Broken group offloading using block_level | ### Describe the bug
pr #12283 introduced group offloading on a pipeline
but its currently broken when using offload_type="block_level" as it does not handle vae module correctly
and result is failure with input vs weight device mismatch
using `offload_type="leaf_level"` it seems to work
### Reproduction
```... | closed | completed | false | 10 | [
"bug",
"contributions-welcome",
"group-offloading"
] | [] | 2025-09-11T16:52:41Z | 2025-12-05T13:24:06Z | 2025-12-05T13:24:06Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 1,442,883,214 | I_kwDOHa8MBc5WAKaO | 1,232 | https://github.com/huggingface/diffusers/issues/1232 | https://api.github.com/repos/huggingface/diffusers/issues/1232 | Dreambooth: cosine_with_restarts learning rate schedule won't restart | ### Describe the bug
Hello,
During log observation of a three dreambooth training logs found that `--lr_scheduler="cosine_with_restarts" lead to a cosine-alike lr change on a tensorboard.
<img width="386" alt="Screen Shot 2022-11-09 at 2 50 45 PM" src="https://user-images.githubusercontent.com/25350960/200959317... | closed | completed | false | 12 | [
"bug",
"stale"
] | [
"pcuenca",
"patil-suraj"
] | 2022-11-09T22:54:27Z | 2023-02-25T02:51:43Z | 2023-01-06T21:43:01Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zetyquickly | 25,350,960 | MDQ6VXNlcjI1MzUwOTYw | User | false |
huggingface/diffusers | 3,408,505,993 | I_kwDOHa8MBc7LKayJ | 12,321 | https://github.com/huggingface/diffusers/issues/12321 | https://api.github.com/repos/huggingface/diffusers/issues/12321 | HunyuanImage-2.1 support | New T2I model, would you mind to support it?
https://huggingface.co/tencent/HunyuanImage-2.1
https://github.com/Tencent-Hunyuan/HunyuanImage-2.1
Many thanks in advance. | closed | completed | false | 1 | [] | [] | 2025-09-12T01:24:08Z | 2025-10-24T08:31:14Z | 2025-10-24T08:31:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kk3dmax | 46,808,324 | MDQ6VXNlcjQ2ODA4MzI0 | User | false |
huggingface/diffusers | 3,410,729,447 | I_kwDOHa8MBc7LS5nn | 12,323 | https://github.com/huggingface/diffusers/issues/12323 | https://api.github.com/repos/huggingface/diffusers/issues/12323 | Add guidance free method as GFT | Source code: https://github.com/thu-ml/GFT
Paper: Chen, Huayu, Kai Jiang, Kaiwen Zheng, Jianfei Chen, Hang Su, and Jun Zhu. “Visual Generation Without Guidance,” 2025. ICML, https://openreview.net/forum?id=gM6aboVgTO.
| open | null | false | 1 | [
"stale",
"consider-for-modular-diffusers"
] | [] | 2025-09-12T14:28:18Z | 2026-01-09T15:11:15Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Bili-Sakura | 108,928,239 | U_kgDOBn4c7w | User | false |
huggingface/diffusers | 3,414,016,624 | I_kwDOHa8MBc7LfcJw | 12,326 | https://github.com/huggingface/diffusers/issues/12326 | https://api.github.com/repos/huggingface/diffusers/issues/12326 | Script to convert a SDXL PEFT trained lora to original format | I've trained a SDXL lora using the PEFT library. It works perfectly fine using the diffusers library but it isn't compatible with comfyui. I am wondering if anyone has made any conversion script for that. I've tried to do a conversion script on my side but I am getting tensor dimension mismatches.
| open | null | false | 1 | [
"stale"
] | [] | 2025-09-13T19:12:15Z | 2026-01-09T15:11:12Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | frutiemax92 | 142,428,698 | U_kgDOCH1KGg | User | false |
huggingface/diffusers | 3,416,066,058 | I_kwDOHa8MBc7LnQgK | 12,329 | https://github.com/huggingface/diffusers/issues/12329 | https://api.github.com/repos/huggingface/diffusers/issues/12329 | WanImageToVideoPipeline loads single-file transformer as Wan2.1 instead of Wan2.2 | ### Describe the bug
When overwriting a transformer loaded as a single file in WanImageToVideoPipeline, it gets loaded as Wan 2.1 instead of Wan 2.2, causing a bug in the encoding process. If you set pipe.transformer.config.image_dim = None after loading, it fixes the issue.
More details: https://huggingface.co/Wan-A... | closed | completed | false | 3 | [
"bug",
"stale"
] | [
"DN6"
] | 2025-09-15T02:45:22Z | 2026-01-09T23:44:08Z | 2026-01-09T23:44:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alfredplpl | 3,625,196 | MDQ6VXNlcjM2MjUxOTY= | User | false |
huggingface/diffusers | 3,421,100,258 | I_kwDOHa8MBc7L6dji | 12,334 | https://github.com/huggingface/diffusers/issues/12334 | https://api.github.com/repos/huggingface/diffusers/issues/12334 | a bug when i use enable_xformers_memory_efficient_attention() | ### Describe the bug
This problem might stem from version incompatibility.
I solved this problem on my own temporarily:
I change the code in ./diffusers/src/diffusers/models/attention.py line 244
`_ = xops.memory_efficient_attention(q, q, q)`
to
`
try: _ = xops.memory_efficient_attention(q, q, q)
except: _ = xops.o... | closed | completed | false | 0 | [
"bug"
] | [
"DN6"
] | 2025-09-16T08:32:18Z | 2025-09-22T07:07:42Z | 2025-09-22T07:07:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zouzhekang | 54,238,776 | MDQ6VXNlcjU0MjM4Nzc2 | User | false |
huggingface/diffusers | 3,422,063,848 | I_kwDOHa8MBc7L-Izo | 12,338 | https://github.com/huggingface/diffusers/issues/12338 | https://api.github.com/repos/huggingface/diffusers/issues/12338 | `AutoencoderDC` bug with `pipe.enable_vae_slicing()` and decoding multiple images | ### Describe the bug
When using the Sana_Sprint_1.6B_1024px and the SANA1.5_4.8B_1024px models, I cannot enable VAE slicing when generating multiple images. I guess this issue will affect the rest of the Sana model and pipeline configurations because they all use the same `AutoencoderDC` model.
I traced the issue to ... | closed | completed | false | 0 | [
"bug"
] | [] | 2025-09-16T12:23:29Z | 2025-09-22T06:55:35Z | 2025-09-22T06:55:35Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mingyi456 | 65,432,252 | MDQ6VXNlcjY1NDMyMjUy | User | false |
huggingface/diffusers | 3,427,320,328 | I_kwDOHa8MBc7MSMII | 12,344 | https://github.com/huggingface/diffusers/issues/12344 | https://api.github.com/repos/huggingface/diffusers/issues/12344 | Qwen Image: txt_seq_lens is redundant and not used | ### Describe the bug
Here https://github.com/huggingface/diffusers/blob/efb7a299af46d739dec6a57a5d2814165fba24b5/src/diffusers/pipelines/qwenimage/pipeline_qwenimage.py#L675 the parameter `txt_seq_lens` is calculated.
It is set to the length of the text attention mask for each batch sample - but both the attention ma... | closed | completed | false | 8 | [
"bug"
] | [] | 2025-09-17T18:16:59Z | 2026-01-12T08:15:11Z | 2026-01-12T08:15:10Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dxqb | 183,307,934 | U_kgDOCu0Ong | User | false |
huggingface/diffusers | 3,431,485,666 | I_kwDOHa8MBc7MiFDi | 12,348 | https://github.com/huggingface/diffusers/issues/12348 | https://api.github.com/repos/huggingface/diffusers/issues/12348 | WanImageToVideoPipeline produces tensor mismatch at some resolutions. | ### Describe the bug
It is unclear what resolutions WanImageToVideoPipeline supports. It says 720p in the docs but 1280 x 720 will produce a tensor mismatch. 1280x704 works.
It says multiples of 16, but theres gotta be more than that to it, evidenced by the above issue. Some other resolutions Ive tried are 640x624, w... | open | null | false | 5 | [
"bug",
"stale"
] | [] | 2025-09-18T18:23:25Z | 2026-02-03T15:18:11Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Meatfucker | 74,834,323 | MDQ6VXNlcjc0ODM0MzIz | User | false |
huggingface/diffusers | 3,436,413,761 | I_kwDOHa8MBc7M04NB | 12,355 | https://github.com/huggingface/diffusers/issues/12355 | https://api.github.com/repos/huggingface/diffusers/issues/12355 | Support for SVD Training | Thank you so much for your valuable contributions.
I am interested in using SVD for training purposes, and I would greatly appreciate it if you could let me know when the SVD training code will be released in the examples. | closed | completed | true | 1 | [
"stale"
] | [] | 2025-09-20T01:54:12Z | 2026-01-09T23:45:05Z | 2026-01-09T23:45:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Hiahia1369 | 166,591,135 | U_kgDOCe36nw | User | false |
huggingface/diffusers | 3,437,029,804 | I_kwDOHa8MBc7M3Oms | 12,358 | https://github.com/huggingface/diffusers/issues/12358 | https://api.github.com/repos/huggingface/diffusers/issues/12358 | Add Lumina-DiMOO as a pipeline | ### Model/Pipeline/Scheduler description
Lumina-DiMOO is a unified multimodal model built on fully discrete diffusion. It supports a wide range of tasks, including text-to-image, image-to-image (editing, subject-driven generation, inpainting), and image understanding. It also outperforms existing open-source multimoda... | open | null | false | 1 | [
"stale"
] | [] | 2025-09-20T11:03:38Z | 2026-01-09T15:10:37Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | qianyu-dlut | 73,575,386 | MDQ6VXNlcjczNTc1Mzg2 | User | false |
huggingface/diffusers | 3,438,020,195 | I_kwDOHa8MBc7M7AZj | 12,359 | https://github.com/huggingface/diffusers/issues/12359 | https://api.github.com/repos/huggingface/diffusers/issues/12359 | Chroma pipeline documentation bug regarding the `guidance_scale` parameter | ### Describe the bug
From my understanding, Chroma is a retrained and dedistilled version of the Flux architecture, so it uses true CFG, unlike Flux. I can indeed confirm that this is true by tracing through the source code.
However, currently the documentation for the `guidance_scale` parameter in the `ChromaPipelin... | closed | completed | false | 1 | [
"bug"
] | [] | 2025-09-21T08:34:15Z | 2025-09-22T20:04:15Z | 2025-09-22T20:04:15Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mingyi456 | 65,432,252 | MDQ6VXNlcjY1NDMyMjUy | User | false |
huggingface/diffusers | 3,440,567,805 | I_kwDOHa8MBc7NEuX9 | 12,367 | https://github.com/huggingface/diffusers/issues/12367 | https://api.github.com/repos/huggingface/diffusers/issues/12367 | Support for Qwen-Image Eligen Lora like Diffsynth Studio. | **Is your feature request related to a problem? Please describe.**
This Feature will really help to make workflows without using Diffsynth.
**Describe the solution you'd like.**
same implementation but natively in Diffusers.
**Describe alternatives you've considered.**
Diffsynth Studio is the only alternative.
**Add... | open | null | false | 1 | [
"stale"
] | [] | 2025-09-22T11:48:59Z | 2026-01-09T15:10:33Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | krigeta | 75,309,361 | MDQ6VXNlcjc1MzA5MzYx | User | false |
huggingface/diffusers | 3,441,124,306 | I_kwDOHa8MBc7NG2PS | 12,368 | https://github.com/huggingface/diffusers/issues/12368 | https://api.github.com/repos/huggingface/diffusers/issues/12368 | Add WanImage2ImagePipeline | [Wan-AI/Wan2.2-T2V-A14B](https://huggingface.co/Wan-AI/Wan2.2-I2V-A14B) supported via `WanPipeline` is not only one of the best video models, its also widely popular as *text-to-image* model
(basically using it to produce video with single frame) due to its very good photo-realistic output.
Ask is to extend support ... | open | null | false | 4 | [
"contributions-welcome"
] | [
"devanshi00"
] | 2025-09-22T13:53:21Z | 2026-02-20T14:11:39Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 3,441,160,684 | I_kwDOHa8MBc7NG_Hs | 12,369 | https://github.com/huggingface/diffusers/issues/12369 | https://api.github.com/repos/huggingface/diffusers/issues/12369 | Wan 2.1 CineScale - up to 4K | ### Model/Pipeline/Scheduler description
Wan 2.1 up to 4k
https://eyeline-labs.github.io/CineScale/
### Open source status
- [x] The model implementation is available.
- [x] The model weights are available (Only relevant if addition is not a scheduler).
### Provide useful links for the implementation
https://gith... | open | null | false | 1 | [
"stale",
"contributions-welcome",
"consider-for-modular-diffusers"
] | [] | 2025-09-22T14:01:53Z | 2026-01-09T15:10:27Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bhack | 1,710,528 | MDQ6VXNlcjE3MTA1Mjg= | User | false |
huggingface/diffusers | 3,443,719,768 | I_kwDOHa8MBc7NQv5Y | 12,373 | https://github.com/huggingface/diffusers/issues/12373 | https://api.github.com/repos/huggingface/diffusers/issues/12373 | Unlearn API for Diffusers | **Is your feature request related to a problem? Please describe.**
Currently, it appears there in no unlearn API that is part of diffusers library specifically in the context of T2I models. Researchers always need to fork existing repos released by various works , plumb them for execution. This is not straight forward.... | open | null | false | 1 | [
"stale"
] | [] | 2025-09-23T05:17:54Z | 2026-01-09T15:10:17Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ksasi | 7,089,391 | MDQ6VXNlcjcwODkzOTE= | User | false |
huggingface/diffusers | 3,444,353,824 | I_kwDOHa8MBc7NTKsg | 12,375 | https://github.com/huggingface/diffusers/issues/12375 | https://api.github.com/repos/huggingface/diffusers/issues/12375 | What kernels should we integrate in Diffusers? | Now that we have an [integration](https://github.com/huggingface/diffusers/pull/12236) with the `kernels` lib to use Flash Attention 3 (FA3), it'd be nice to gather community interest about which kernels we should try to incorporate in the library through the [`kernels` lib](https://github.com/huggingface/kernels/). FA... | open | null | false | 10 | [
"stale",
"performance"
] | [] | 2025-09-23T09:03:13Z | 2026-02-03T15:18:04Z | null | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,443,257,131 | I_kwDOHa8MBc5WBlsr | 1,238 | https://github.com/huggingface/diffusers/issues/1238 | https://api.github.com/repos/huggingface/diffusers/issues/1238 | [Flax] Crazy Loss 😱 | ### Describe the bug
https://github.com/huggingface/diffusers/blob/2e980ac9a0c03200cbe52e78e23f648d17e97b9c/examples/dreambooth/train_dreambooth_flax.py#L617
https://github.com/huggingface/diffusers/blob/2e980ac9a0c03200cbe52e78e23f648d17e97b9c/examples/text_to_image/train_text_to_image_flax.py#L525
I am getting... | closed | completed | false | 0 | [
"bug"
] | [] | 2022-11-10T05:00:15Z | 2022-11-10T06:09:49Z | 2022-11-10T06:09:49Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | camenduru | 54,370,274 | MDQ6VXNlcjU0MzcwMjc0 | User | false |
huggingface/diffusers | 3,450,008,513 | I_kwDOHa8MBc7NovPB | 12,386 | https://github.com/huggingface/diffusers/issues/12386 | https://api.github.com/repos/huggingface/diffusers/issues/12386 | Implement missing features on ModularPipeline | as i'm looking to take advantage of new `ModularPipeline` ask is to implement some currently missing features
my use case is to convert existing loaded model using standard pipeline into modular pipeline. that functionality was provided via #11915 and is now working.
first minor obstacle is that modular pipeline does... | open | null | false | 4 | [
"stale",
"roadmap"
] | [] | 2025-09-24T15:49:23Z | 2026-02-03T15:17:57Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 1,443,282,113 | I_kwDOHa8MBc5WBrzB | 1,239 | https://github.com/huggingface/diffusers/issues/1239 | https://api.github.com/repos/huggingface/diffusers/issues/1239 | Accept License while downloading as a y/n option in cmd terminal | **Is your feature request related to a problem? Please describe.**
When I wanted to "git clone https://huggingface.co/runwayml/stable-diffusion-v1-5", it failed cause I hadn't accepted the permission in HF for the repo.
**Describe the solution you'd like**
Instead of asking me to go to the web to accept it, w... | closed | completed | false | 4 | [] | [] | 2022-11-10T05:31:50Z | 2022-11-16T10:52:56Z | 2022-11-16T10:52:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghpkishore | 26,677,859 | MDQ6VXNlcjI2Njc3ODU5 | User | false |
huggingface/diffusers | 3,456,704,505 | I_kwDOHa8MBc7OCR_5 | 12,396 | https://github.com/huggingface/diffusers/issues/12396 | https://api.github.com/repos/huggingface/diffusers/issues/12396 | Using leaf_level together with delete_adapters can get some errors. | ### Describe the bug
I want to dynamically control LoRA, but when I use CPU offloading in leaf_level mode, an error occurs after calling delete_adapters.I think it's caused by LoRA's hook.
### Reproduction
```python
import torch
from controlnet_aux import CannyDetector
from diffusers import FluxControlPipeline
from ... | closed | completed | false | 4 | [
"bug",
"contributions-welcome",
"group-offloading"
] | [] | 2025-09-26T09:46:34Z | 2025-12-03T11:16:33Z | 2025-12-03T11:16:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | songh11 | 75,419,275 | MDQ6VXNlcjc1NDE5Mjc1 | User | false |
huggingface/diffusers | 1,313,628,914 | I_kwDOHa8MBc5OTGLy | 124 | https://github.com/huggingface/diffusers/issues/124 | https://api.github.com/repos/huggingface/diffusers/issues/124 | [Documentation request] How to create a dataset for training | You do say [how to train a new model from a dataset](https://github.com/huggingface/diffusers/tree/main/examples). You don't say how to create a dataset.
Is it just a directory of images? Do they need to be sized somehow? Is there a subdirectory tree? Are there text files containing metadata? | closed | completed | false | 5 | [] | [
"anton-l"
] | 2022-07-21T18:14:50Z | 2022-07-21T19:45:01Z | 2022-07-21T19:45:01Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | StoneCypher | 77,482 | MDQ6VXNlcjc3NDgy | User | false |
huggingface/diffusers | 3,461,387,212 | I_kwDOHa8MBc7OUJPM | 12,401 | https://github.com/huggingface/diffusers/issues/12401 | https://api.github.com/repos/huggingface/diffusers/issues/12401 | Redundant lru_cache in QwenEmbedRope | https://github.com/huggingface/diffusers/blob/041501aea92919c9c7f36e189fc9cf7d865ebb96/src/diffusers/models/transformers/transformer_qwenimage.py#L238
`self.rope_cache` is equivalent to `lru_cache`, so `lru_cache` is no longer needed here. | closed | completed | false | 2 | [] | [] | 2025-09-28T08:47:52Z | 2025-10-20T01:41:59Z | 2025-10-20T01:41:59Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chenxiao111222 | 154,797,505 | U_kgDOCToFwQ | User | false |
huggingface/diffusers | 3,461,543,685 | I_kwDOHa8MBc7OUvcF | 12,402 | https://github.com/huggingface/diffusers/issues/12402 | https://api.github.com/repos/huggingface/diffusers/issues/12402 | Eliminate code ambiguity in models.transformers.FluxAttention | In the initialization of [FluxAttention](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/transformers/transformer_flux.py#L319), we determine whether to create to_out based on the parameter pre_only.
However, during inference, we decide whether to call to_out based on whether [encoder_hidden_sta... | open | null | false | 1 | [
"stale"
] | [] | 2025-09-28T10:20:41Z | 2026-01-09T15:09:47Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | j2gg0s | 17,896,478 | MDQ6VXNlcjE3ODk2NDc4 | User | false |
huggingface/diffusers | 3,462,549,840 | I_kwDOHa8MBc7OYlFQ | 12,404 | https://github.com/huggingface/diffusers/issues/12404 | https://api.github.com/repos/huggingface/diffusers/issues/12404 | Modular guiders should provide sane defaults | Modular guiders are a great concept, but right now they are nearly impossible to use due to sane lack of default values.
For example, `PerturbedAttentionGuidance` requires either `SkipLayerGuidance: List[int]` param or `perturbed_guidance_config: LayerSkipConfig` param
There are no chance that normal user would know a... | open | null | false | 2 | [
"stale",
"roadmap"
] | [] | 2025-09-28T21:21:29Z | 2026-02-03T15:17:50Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 3,463,752,523 | I_kwDOHa8MBc7OdKtL | 12,406 | https://github.com/huggingface/diffusers/issues/12406 | https://api.github.com/repos/huggingface/diffusers/issues/12406 | Generated image is cropped when using smaller output size to reduce VRAM and processing time in Flux Kontext | Use `max_area` in FluxKontextPipeline, default `max_area=1024**2`. I set `max_area=768**2`, and output is cropped
Original image:
Output with `max_area=1024**2`:
<img width="1248" height="832" alt="Image" src="https://github.... | open | null | false | 5 | [
"stale"
] | [] | 2025-09-29T08:07:42Z | 2026-02-03T15:17:46Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dangph-alala | 198,682,075 | U_kgDOC9el2w | User | false |
huggingface/diffusers | 3,469,520,845 | I_kwDOHa8MBc7OzK_N | 12,409 | https://github.com/huggingface/diffusers/issues/12409 | https://api.github.com/repos/huggingface/diffusers/issues/12409 | Diffusers crash due to hard-dependency on `torch._C._distributed_c10d` | ### Describe the bug
`BriaTransformer2DModel` introduces a hard dependency on `torch._C._distributed_c10d` which is NOT present in all **torch** variants/flavors.
Any `torch._C` import should be guarded with *if-available*, not a hard dependency on the entire library.
Without that, diffusers stop being a cross-platf... | closed | completed | false | 8 | [
"bug"
] | [
"sayakpaul"
] | 2025-09-30T13:36:56Z | 2025-10-10T14:33:00Z | 2025-10-10T14:33:00Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 3,469,605,517 | I_kwDOHa8MBc7OzfqN | 12,410 | https://github.com/huggingface/diffusers/issues/12410 | https://api.github.com/repos/huggingface/diffusers/issues/12410 | WanVACEPipeline does not work with low-noise stage only | ### Describe the bug
`WanVACEPipeline` has unnecessary/unsafe access to high-noise transformer module of the model which invalidates the ability to use low-noise stage only that was introduced in https://github.com/huggingface/diffusers/pull/12024
> https://github.com/huggingface/diffusers/blob/b59654544bbaf5c6040e67... | closed | completed | false | 1 | [
"bug"
] | [] | 2025-09-30T13:52:53Z | 2025-11-07T12:45:39Z | 2025-11-07T12:45:39Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 3,471,796,451 | I_kwDOHa8MBc7O72jj | 12,412 | https://github.com/huggingface/diffusers/issues/12412 | https://api.github.com/repos/huggingface/diffusers/issues/12412 | HunyuanImage-3.0 support | ### Model/Pipeline/Scheduler description
New T2I/I2I model released from Hunyuan (HunyuanImage-3.0)
Tencent just released HunyuanImage-3.0, which came quite fast after the 2.1 release:
- https://huggingface.co/tencent/HunyuanImage-3.0
- https://github.com/Tencent-Hunyuan/HunyuanImage-3.0
Would it be possible to add ... | open | null | false | 1 | [
"stale"
] | [] | 2025-10-01T02:47:18Z | 2026-01-09T15:09:30Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ryan-seungyong-lee | 74,466,088 | MDQ6VXNlcjc0NDY2MDg4 | User | false |
huggingface/diffusers | 3,472,897,753 | I_kwDOHa8MBc7PADbZ | 12,414 | https://github.com/huggingface/diffusers/issues/12414 | https://api.github.com/repos/huggingface/diffusers/issues/12414 | Validation of results with quantitative evaluation metrics - Text to Image LORA | Hello,
I used text to image with LORA to do fine tuning on my customized dataset and then tested model with prompts.
https://github.com/huggingface/diffusers/tree/main/examples/text_to_image
It seems it works fine, but I was wondering if there are some quantitative evaluation metrics to validate results.
Thanks in... | closed | completed | true | 3 | [
"stale"
] | [] | 2025-10-01T10:08:49Z | 2026-01-10T00:07:08Z | 2026-01-10T00:07:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | HripsimeS | 42,246,765 | MDQ6VXNlcjQyMjQ2NzY1 | User | false |
huggingface/diffusers | 3,473,028,740 | I_kwDOHa8MBc7PAjaE | 12,415 | https://github.com/huggingface/diffusers/issues/12415 | https://api.github.com/repos/huggingface/diffusers/issues/12415 | SVG 2 kernels | Can we support new sparse kernels in (Neurips 2025)
https://svg-project.github.io/v2/ | open | null | false | 1 | [
"stale"
] | [] | 2025-10-01T10:52:50Z | 2026-01-09T15:09:19Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bhack | 1,710,528 | MDQ6VXNlcjE3MTA1Mjg= | User | false |
huggingface/diffusers | 1,443,573,447 | I_kwDOHa8MBc5WCy7H | 1,242 | https://github.com/huggingface/diffusers/issues/1242 | https://api.github.com/repos/huggingface/diffusers/issues/1242 | Safe Latent Diffusion | ### Model/Pipeline/Scheduler description
We have just released a [paper](https://arxiv.org/abs/2211.05105) introducing Safe Latent Diffusion (SLD) for text2image diffusion models.
SLD mitigates the well known issue that models like Stable Diffusion that are trained on unfiltered, web-crawled datasets tend to suffe... | closed | completed | false | 4 | [
"stale"
] | [] | 2022-11-10T09:44:17Z | 2022-12-13T13:22:11Z | 2022-12-13T13:22:11Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | manuelbrack | 26,039,108 | MDQ6VXNlcjI2MDM5MTA4 | User | false |
huggingface/diffusers | 3,478,061,137 | I_kwDOHa8MBc7PTwBR | 12,422 | https://github.com/huggingface/diffusers/issues/12422 | https://api.github.com/repos/huggingface/diffusers/issues/12422 | Work WorkNCCL(SeqNum=2521, OpType=ALLREDUCE, NumelIn=160672064, NumelOut=160672064, Timeout(ms)=600000) timed out in blocking wait. | ### Describe the bug
It keeps getting stuck, I don't know what the problem is, one node has 8 H200
### Reproduction
```
accelerate launch --config_file "./accelerate_config.yaml" \
--use_deepspeed \
--deepspeed_config_file "./deepspeed_config.json" \
"diffusers/examples/dreambooth/train_dreambooth_flux.p... | open | null | false | 1 | [
"bug",
"stale"
] | [] | 2025-10-02T15:59:10Z | 2026-01-09T15:09:15Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hhy5861 | 8,274,977 | MDQ6VXNlcjgyNzQ5Nzc= | User | false |
huggingface/diffusers | 3,478,669,042 | I_kwDOHa8MBc7PWEby | 12,423 | https://github.com/huggingface/diffusers/issues/12423 | https://api.github.com/repos/huggingface/diffusers/issues/12423 | QwenImageTransformer2DModel.from_single_file | ### Describe the bug
`QwenImageTransformer2DModel.from_single_file` requires the following arguments:
```
QwenImageTransformer2DModel.from_single_file(
local_safetensors_file,
config="Qwen/Qwen-Image",
subfolder="transformer",
torch_dtype = torch.bfloat16
)
```
if `config` and `subfolder` are not provided, it... | closed | completed | false | 2 | [
"bug"
] | [] | 2025-10-02T19:28:56Z | 2025-10-03T09:23:04Z | 2025-10-03T09:23:03Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dxqb | 183,307,934 | U_kgDOCu0Ong | User | false |
huggingface/diffusers | 3,480,167,722 | I_kwDOHa8MBc7PbyUq | 12,427 | https://github.com/huggingface/diffusers/issues/12427 | https://api.github.com/repos/huggingface/diffusers/issues/12427 | UNet-only adapters will still cause problems with pipeline.set_adapters() after torch.compile because the compiled graph expects adapters to exist in all components. | 2025-10-03 07:47:56 - main - INFO - loading adapter detail_enhancer
No LoRA keys associated to CLIPTextModel found with the prefix='text_encoder'. This is safe to ignore if LoRA state dict didn't originally have any CLIPTextModel related params. You can also try specifying prefix=None to resolve the warning. Otherwise,... | open | null | false | 3 | [
"stale"
] | [] | 2025-10-03T07:55:01Z | 2026-02-03T15:17:33Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiangtaozh | 1,331,471 | MDQ6VXNlcjEzMzE0NzE= | User | false |
huggingface/diffusers | 3,484,230,809 | I_kwDOHa8MBc7PrSSZ | 12,431 | https://github.com/huggingface/diffusers/issues/12431 | https://api.github.com/repos/huggingface/diffusers/issues/12431 | [ko] Update outdated Korean documentation in docs/source/ko 🇰🇷<ko> | Hi! I noticed that the Korean translation (docs/source/ko) in the Diffusers documentation has not been updated for about 5 months.
Some of the translated MDX files are marked as “in progress” or partially translated.
I'd like to help update and improve the existing Korean documentation to match the latest English vers... | open | null | false | 2 | [
"stale"
] | [] | 2025-10-04T23:57:44Z | 2026-01-09T15:09:08Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | braintrue | 89,627,517 | MDQ6VXNlcjg5NjI3NTE3 | User | false |
huggingface/diffusers | 3,484,794,355 | I_kwDOHa8MBc7Ptb3z | 12,432 | https://github.com/huggingface/diffusers/issues/12432 | https://api.github.com/repos/huggingface/diffusers/issues/12432 | get_1d_sincos_pos_embed_from_grid uses float64 unavailable on MPS | ### Describe the bug
for some reason, positional embeds in pixart now require fp64, but this pattern has been discouraged as Mac users don't have fp64 at all - not even emulation is available currently.
### Reproduction
```py
def get_1d_sincos_pos_embed_from_grid(embed_dim, pos, output_type="np", flip_sin_to_cos=F... | closed | completed | false | 4 | [
"bug"
] | [] | 2025-10-05T13:10:25Z | 2025-10-23T10:48:09Z | 2025-10-23T10:48:09Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 3,485,032,724 | I_kwDOHa8MBc7PuWEU | 12,433 | https://github.com/huggingface/diffusers/issues/12433 | https://api.github.com/repos/huggingface/diffusers/issues/12433 | Documentation instructs installing diffusers from git causing version incompatibility with example requirements.txt (text to image lora) | ### Describe the bug
The official LoRA training tutorial/examples documentation instructs users to install diffusers from git, which installs the dev version. However, the requirements.txt files in the examples folders (examples/text_to_image/requirements.txt) specify outdated dependency versions like peft==0.7.0.
Th... | closed | completed | false | 5 | [
"bug"
] | [] | 2025-10-05T18:10:01Z | 2025-11-17T01:28:45Z | 2025-10-06T02:54:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | moha23 | 10,138,421 | MDQ6VXNlcjEwMTM4NDIx | User | false |
huggingface/diffusers | 3,485,440,338 | I_kwDOHa8MBc7Pv5lS | 12,436 | https://github.com/huggingface/diffusers/issues/12436 | https://api.github.com/repos/huggingface/diffusers/issues/12436 | TypeError: CLIPTextModel.__init__() got an unexpected keyword argument 'offload_state_dict' | ### Describe the bug
Getting this `TypeError: CLIPTextModel.__init__() got an unexpected keyword argument 'offload_state_dict'` when running a `DiffusionPipeline.from_pretrained`
### Reproduction
I was following this https://huggingface.co/docs/diffusers/main/en/installation for installation. This is what I have don... | closed | completed | false | 17 | [
"bug"
] | [] | 2025-10-06T02:00:04Z | 2025-11-01T19:18:32Z | 2025-10-07T08:21:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zillur01 | 54,721,606 | MDQ6VXNlcjU0NzIxNjA2 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.