repo stringclasses 1
value | github_id int64 1.27B 4.95B | github_node_id stringlengths 18 24 | number int64 8 14.3k | 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 2026-07-22 10:32:58 | updated_at stringdate 2022-06-12 22:18:01 2026-07-22 10:32:58 | closed_at stringdate 2022-06-12 22:18:01 2026-07-21 13:17:28 ⌀ | author_association stringclasses 3
values | milestone_title stringclasses 0
values | snapshot_id stringclasses 149
values | extracted_at stringdate 2026-04-07 13:34:13 2026-07-22 10:35:16 | 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 | 1,447,013,749 | I_kwDOHa8MBc5WP611 | 1,268 | https://github.com/huggingface/diffusers/issues/1268 | https://api.github.com/repos/huggingface/diffusers/issues/1268 | [RFC] Proposal to change API of `ConfigMixin` and `SchedulerMixin`. | There are a couple of API problems currently related to the `from_config` method. This is a proposal to change the behavior of `from_config` and add a new `load_config` method instead.
## 1. Loading schedulers
Currently one has to load schedulers as follows:
```python
from diffusers import DDIMScheduler
sc... | closed | completed | false | 5 | [] | [] | 2022-11-13T17:32:38Z | 2022-11-15T17:15:15Z | 2022-11-15T17:15:15Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 3,640,076,970 | I_kwDOHa8MBc7Y9yqq | 12,683 | https://github.com/huggingface/diffusers/issues/12683 | https://api.github.com/repos/huggingface/diffusers/issues/12683 | [FEATURE] [WAN2.2] Add support to train LoRA for WanVacePipeline | I can't train LoRA (aka PEFT) model, using only diffusers.
I know about DiffSync, but it has different code and some broken copabilities.
I have dataset:
In each folder 4 files:
`orig.mp4` - original video (1280*720)
`cropped.mp4` - cropped video with gray frames
`mask.png` - mask for outpaint
`prompt.txt` - positive ... | closed | completed | true | 1 | [
"stale"
] | [] | 2025-11-18T21:21:51Z | 2026-01-10T01:49:26Z | 2026-01-10T01:49:26Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | SlimRG | 39,348,033 | MDQ6VXNlcjM5MzQ4MDMz | User | false |
huggingface/diffusers | 3,648,587,250 | I_kwDOHa8MBc7ZeQXy | 12,689 | https://github.com/huggingface/diffusers/issues/12689 | https://api.github.com/repos/huggingface/diffusers/issues/12689 | LoRA adapter weights are not loaded from checkpoints | ### Describe the bug
examples/text_to_image/train_text_to_image_lora.py file has a serious bug.
The training pipeline saves the accelerator states and the lora adapters correctly while saving checkpoint.
But the checkpoint loading code **skips loading lora adapters**.
I am raising a PR very soon...
Please look int... | closed | completed | false | 0 | [
"bug"
] | [] | 2025-11-20T18:42:48Z | 2025-11-28T06:26:40Z | 2025-11-28T06:26:40Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | SurAyush | 145,102,142 | U_kgDOCKYVPg | User | false |
huggingface/diffusers | 3,652,667,588 | I_kwDOHa8MBc7Zt0jE | 12,694 | https://github.com/huggingface/diffusers/issues/12694 | https://api.github.com/repos/huggingface/diffusers/issues/12694 | Feature Request: Enhanced Model Metadata Extraction and Export Utility | # Feature Request: Enhanced Model Metadata Extraction and Export Utility
## Is your feature request related to a problem? Please describe.
Currently, when working with diffusion models in the diffusers library, there's no standardized way to extract and export comprehensive model metadata. This creates several challe... | closed | completed | false | 3 | [] | [] | 2025-11-21T18:18:26Z | 2025-11-22T05:45:56Z | 2025-11-22T05:45:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jitmisra | 169,401,406 | U_kgDOChjcPg | User | false |
huggingface/diffusers | 3,652,933,863 | I_kwDOHa8MBc7Zu1jn | 12,695 | https://github.com/huggingface/diffusers/issues/12695 | https://api.github.com/repos/huggingface/diffusers/issues/12695 | Hunyuan v1.5 Support | # Hunyuan v1.5
https://github.com/user-attachments/assets/fd068bf8-0798-4d29-8067-ed5c42c8da52
**Is your feature request related to a problem? Please describe.**
HunyuanVideo-1.5 is the newest state-of-the-art open-source model. Would love to see support in Diffusers!
**Describe the solution you'd like.**
A text to ... | closed | completed | false | 3 | [
"stale"
] | [] | 2025-11-21T20:02:19Z | 2026-01-10T01:53:09Z | 2026-01-10T01:53:08Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghunkins | 12,562,057 | MDQ6VXNlcjEyNTYyMDU3 | User | false |
huggingface/diffusers | 3,653,774,988 | I_kwDOHa8MBc7ZyC6M | 12,697 | https://github.com/huggingface/diffusers/issues/12697 | https://api.github.com/repos/huggingface/diffusers/issues/12697 | Magcache Support. | Id like to see magcache acceleration supported if possible. My own testing shows it to cut video generation times in half with almost no quality loss at all. It seems to be considerably better quality than something like teacache or more efficient attention mechanisms.
https://github.com/Zehong-Ma/MagCache
Kandinsky5... | closed | completed | false | 3 | [
"contributions-welcome",
"advanced",
"performance",
"diffusers-mvp"
] | [] | 2025-11-22T03:34:50Z | 2026-02-04T08:15:14Z | 2026-02-04T08:15:14Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Meatfucker | 74,834,323 | MDQ6VXNlcjc0ODM0MzIz | User | false |
huggingface/diffusers | 3,654,213,950 | I_kwDOHa8MBc7ZzuE- | 12,699 | https://github.com/huggingface/diffusers/issues/12699 | https://api.github.com/repos/huggingface/diffusers/issues/12699 | the quality of video downgrade very much when calling pipe for 50 different prompts in wan 2.2 with lighting lora | I need to call WanImageToVideoPipeline 50 times for 50 different prompts to generate a long video. In each call, I delete the pipe and reconstruct and use torch.cuda.empty_cache() and torch.cuda.synchronize() to clean the gpu, but not helpful. The video downgrade very much in 1:30. https://huggingface.co/datasets/exact... | closed | completed | true | 4 | [
"stale"
] | [] | 2025-11-22T08:21:34Z | 2026-01-10T01:57:49Z | 2026-01-10T01:57:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chaowenguo | 229,142,208 | U_kgDODahuwA | User | false |
huggingface/diffusers | 3,658,501,843 | I_kwDOHa8MBc7aEE7T | 12,706 | https://github.com/huggingface/diffusers/issues/12706 | https://api.github.com/repos/huggingface/diffusers/issues/12706 | [Feature] Ulysses Attention for any sequence length w/o padding | ## 🤖UAA: Ulysses Anything Attention
<div id="ulysses-anything-attention"></div>
We have implemented the **[📚UAA: Ulysses Anything Attention](https://github.com/vipshop/cache-dit/blob/main/docs/User_Guide.md#uaa-ulysses-anything-attention)**: An Ulysses Attention that supports **arbitrary sequence length** with ✅**... | closed | completed | false | 8 | [
"roadmap"
] | [
"DefTruth"
] | 2025-11-24T11:42:00Z | 2026-02-02T11:34:34Z | 2026-02-02T11:34:34Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DefTruth | 31,974,251 | MDQ6VXNlcjMxOTc0MjUx | User | false |
huggingface/diffusers | 3,659,783,058 | I_kwDOHa8MBc7aI9uS | 12,709 | https://github.com/huggingface/diffusers/issues/12709 | https://api.github.com/repos/huggingface/diffusers/issues/12709 | Meituan Longcat Video | ### Model/Pipeline/Scheduler description
https://huggingface.co/meituan-longcat/LongCat-Video
https://github.com/meituan-longcat/LongCat-Video
https://meituan-longcat.github.io/LongCat-Video/
Video generation model. Supports T2V, I2V and video continuation
### Open source status
- [x] The model implementation is av... | open | reopened | false | 6 | [] | [] | 2025-11-24T17:11:07Z | 2026-04-01T08:59:05Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alerikaisattera | 73,682,764 | MDQ6VXNlcjczNjgyNzY0 | User | false |
huggingface/diffusers | 1,447,082,676 | I_kwDOHa8MBc5WQLq0 | 1,271 | https://github.com/huggingface/diffusers/issues/1271 | https://api.github.com/repos/huggingface/diffusers/issues/1271 | Unable to correctly install dependencies for Dreambooth example on GCP user or managed notebooks | ### Describe the bug
I am attempting to simply run the [Dreambooth training example](https://github.com/huggingface/diffusers/tree/main/examples/dreambooth) on a GCP Vertex AI workbench notebook. I have tried their managed notebook and user-managed notebooks with the same issue. However, I can not seem to get the de... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2022-11-13T21:01:39Z | 2023-03-17T06:13:16Z | 2023-01-21T15:03:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DevonPeroutky | 36,947,377 | MDQ6VXNlcjM2OTQ3Mzc3 | User | false |
huggingface/diffusers | 3,663,940,874 | I_kwDOHa8MBc7aY00K | 12,712 | https://github.com/huggingface/diffusers/issues/12712 | https://api.github.com/repos/huggingface/diffusers/issues/12712 | FLUX.2-dev-bnb-4bit Missing | ### Describe the bug
<img width="461" height="282" alt="Image" src="https://github.com/user-attachments/assets/4365a617-05ee-43bd-bc95-5f59d0750f45" />
The `diffusers/FLUX.2-dev-bnb-4bit` model is missing from Huggingface! Maybe the repo is accidentally set to private?
Huge thank you to the team for the Day 0 suppor... | closed | completed | false | 2 | [
"bug"
] | [] | 2025-11-25T17:24:10Z | 2025-11-25T17:33:27Z | 2025-11-25T17:32:33Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghunkins | 12,562,057 | MDQ6VXNlcjEyNTYyMDU3 | User | false |
huggingface/diffusers | 3,666,454,182 | I_kwDOHa8MBc7aiaam | 12,719 | https://github.com/huggingface/diffusers/issues/12719 | https://api.github.com/repos/huggingface/diffusers/issues/12719 | how to use quantization and device_map=balance to run qwen-image on kaggle T4 * 2 | ```python3
!python3 -m pip install -U diffusers peft bitsandbytes protobuf
import diffusers, torch, math
qwen = diffusers.QwenImagePipeline.from_pretrained('Qwen/Qwen-Image', quantization_config=diffusers.PipelineQuantizationConfig(quant_backend='bitsandbytes_4bit', quant_kwargs={'load_in_4bit':True, 'bnb_4bit_quant_ty... | open | null | false | 3 | [
"stale"
] | [] | 2025-11-26T08:35:46Z | 2026-02-03T15:15:31Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chaowenguo | 229,142,208 | U_kgDODahuwA | User | false |
huggingface/diffusers | 1,447,085,601 | I_kwDOHa8MBc5WQMYh | 1,272 | https://github.com/huggingface/diffusers/issues/1272 | https://api.github.com/repos/huggingface/diffusers/issues/1272 | --mixed_precision="fp16" works only with accellerate config | ### Describe the bug
when setting the parameters --mixed_precision="fp16" the system prompts the message "--mixed_precision is set to its default value to no" despite of declaring the parameters. The message disappear if I launch accelerate config and set mixed precision from there.
### Reproduction
_No response_... | closed | completed | false | 5 | [
"bug"
] | [
"patil-suraj"
] | 2022-11-13T21:08:26Z | 2022-11-22T12:35:25Z | 2022-11-22T12:35:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | spuliz | 15,948,985 | MDQ6VXNlcjE1OTQ4OTg1 | User | false |
huggingface/diffusers | 3,666,820,769 | I_kwDOHa8MBc7ajz6h | 12,720 | https://github.com/huggingface/diffusers/issues/12720 | https://api.github.com/repos/huggingface/diffusers/issues/12720 | how to quantization wan 2.2 vace after loading lora? | ```python3
diffusers.WanVACEPipeline.from_pretrained('linoyts/Wan2.2-VACE-Fun-14B-diffusers', vae=diffusers.AutoencoderKLWan.from_pretrained('linoyts/Wan2.2-VACE-Fun-14B-diffusers', subfolder='vae', torch_dtype=torch.float32), torch_dtype=torch.bfloat16, quantization_config=diffusers.PipelineQuantizationConfig(quant_ba... | closed | completed | true | 2 | [
"stale"
] | [] | 2025-11-26T10:11:38Z | 2026-01-10T02:03:43Z | 2026-01-10T02:03:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chaowenguo | 229,142,208 | U_kgDODahuwA | User | false |
huggingface/diffusers | 3,667,397,297 | I_kwDOHa8MBc7amAqx | 12,722 | https://github.com/huggingface/diffusers/issues/12722 | https://api.github.com/repos/huggingface/diffusers/issues/12722 | How to run qwen-image in kaggle gpu T4 * 2 successfully? | ```python3
!python3 -m pip install -U diffusers peft bitsandbytes
import diffusers, torch, math
qwen = diffusers.QwenImagePipeline.from_pretrained('Qwen/Qwen-Image', torch_dtype=torch.float16, low_cpu_mem_usage=True, quantization_config=diffusers.PipelineQuantizationConfig(quant_backend='bitsandbytes_4bit', quant_kwarg... | open | null | false | 2 | [
"stale"
] | [] | 2025-11-26T12:53:30Z | 2026-01-09T15:04:28Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chaowenguo | 229,142,208 | U_kgDODahuwA | User | false |
huggingface/diffusers | 3,669,640,750 | I_kwDOHa8MBc7aukYu | 12,728 | https://github.com/huggingface/diffusers/issues/12728 | https://api.github.com/repos/huggingface/diffusers/issues/12728 | torch.compile + channels_last support for Wan 2.2 (T2V / I2V) fails with RuntimeError + Dynamo Unsupported behavior | ### Describe the bug
## Hi, I am trying to optimize Wan 2.2 T2V / I2V inference speed on a single RTX 4090, using:
1 Wan2.2 (Diffusers)
2 LightX2V LoRA
3 flash attention
4 group offload (Diffusers 0.30+)
5 torch.compile(mode="max-autotune", fullgraph=True) / torch.channels_last (as recommended in the docs)
## My goa... | open | null | false | 9 | [
"bug",
"stale"
] | [] | 2025-11-27T02:30:40Z | 2026-02-08T15:06:00Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Passenger12138 | 47,623,700 | MDQ6VXNlcjQ3NjIzNzAw | User | false |
huggingface/diffusers | 1,447,091,091 | I_kwDOHa8MBc5WQNuT | 1,273 | https://github.com/huggingface/diffusers/issues/1273 | https://api.github.com/repos/huggingface/diffusers/issues/1273 | [Dreambooth] Training script mean loss calculation question | Hello,
I was playing with Dreambooth training and noticed a thing [here](https://github.com/huggingface/diffusers/blob/0025626cd9feca260e05f295f272af4f7b9ce44a/examples/dreambooth/train_dreambooth.py#L606)
Why did the author use this way of calculation mean of the loss? Is there any particular technical need to ... | closed | completed | false | 6 | [
"training"
] | [
"patil-suraj"
] | 2022-11-13T21:22:42Z | 2022-12-30T13:32:39Z | 2022-12-30T13:32:39Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zetyquickly | 25,350,960 | MDQ6VXNlcjI1MzUwOTYw | User | false |
huggingface/diffusers | 3,669,802,489 | I_kwDOHa8MBc7avL35 | 12,730 | https://github.com/huggingface/diffusers/issues/12730 | https://api.github.com/repos/huggingface/diffusers/issues/12730 | Debugging snippet still remains in Flux2Pipeline | It looks like this code—probably used for debugging—is still in the repository. If it wasn’t meant to be kept, shouldn’t it be removed?
https://github.com/huggingface/diffusers/blob/e6d46123091afd58281dc7487c0f6b67055683b9/src/diffusers/pipelines/flux2/pipeline_flux2.py#L864
| closed | completed | false | 2 | [] | [] | 2025-11-27T04:19:10Z | 2025-12-01T05:55:21Z | 2025-12-01T05:55:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nuclearboy95 | 15,180,201 | MDQ6VXNlcjE1MTgwMjAx | User | false |
huggingface/diffusers | 3,669,819,209 | I_kwDOHa8MBc7avP9J | 12,732 | https://github.com/huggingface/diffusers/issues/12732 | https://api.github.com/repos/huggingface/diffusers/issues/12732 | diffusers inference pipeline error for Z-Image | ### Describe the bug
after I pip install diffusers from source
I run the following code"
import torch
from diffusers import ZImagePipeline,
# 1. Load the pipeline
# Use bfloat16 for optimal performance on supported GPUs
pipe = ZImagePipeline.from_pretrained(
"Tongyi-MAI/Z-Image-Turbo",
torch_dtype=torch.bfl... | closed | completed | false | 2 | [
"bug"
] | [] | 2025-11-27T04:29:14Z | 2025-11-27T06:06:59Z | 2025-11-27T06:06:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MengHao666 | 47,114,466 | MDQ6VXNlcjQ3MTE0NDY2 | User | false |
huggingface/diffusers | 3,670,103,623 | I_kwDOHa8MBc7awVZH | 12,735 | https://github.com/huggingface/diffusers/issues/12735 | https://api.github.com/repos/huggingface/diffusers/issues/12735 | Sam3D support | Hello, there was an amazing release earlier this week of Sam3D: https://ai.meta.com/sam3d/
Is it planned to support it?
Thanks in advance! | closed | completed | false | 2 | [] | [] | 2025-11-27T06:54:59Z | 2025-11-27T07:50:16Z | 2025-11-27T07:50:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fabvio | 10,325,202 | MDQ6VXNlcjEwMzI1MjAy | User | false |
huggingface/diffusers | 3,672,558,378 | I_kwDOHa8MBc7a5ssq | 12,738 | https://github.com/huggingface/diffusers/issues/12738 | https://api.github.com/repos/huggingface/diffusers/issues/12738 | ZImagePipeline fails with batch sizes above 1 with Assertion Error | ### Describe the bug
If you try to use ZImagePipeline with batch sizes above 1, it fails with an Assertion error
```
### Reproduction
import torch
from diffusers import ZImagePipeline
# 1. Load the pipeline
# Use bfloat16 for optimal performance on supported GPUs
pipe = ZImagePipeline.from_pretrained(
"Tongyi-... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2025-11-27T17:43:57Z | 2026-01-10T02:04:10Z | 2026-01-10T02:04:10Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Meatfucker | 74,834,323 | MDQ6VXNlcjc0ODM0MzIz | User | false |
huggingface/diffusers | 1,447,467,832 | I_kwDOHa8MBc5WRps4 | 1,274 | https://github.com/huggingface/diffusers/issues/1274 | https://api.github.com/repos/huggingface/diffusers/issues/1274 | num_images_per_prompt=num_samples FAILS after latest commit in SD pipelines | ### Describe the bug
This used to wwork:
images = pipe(
prompt=prompt,
image=image,
mask_image=mask_image,
guidance_scale=guidance_scale,
generator=generator,
num_images_per_prompt=num_samples,
num_inference_steps =50,
height=height,
width=width,
).images
Now i... | closed | completed | false | 4 | [
"bug"
] | [
"patil-suraj"
] | 2022-11-14T06:24:51Z | 2022-11-14T21:03:12Z | 2022-11-14T21:03:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghpkishore | 26,677,859 | MDQ6VXNlcjI2Njc3ODU5 | User | false |
huggingface/diffusers | 3,675,670,463 | I_kwDOHa8MBc7bFke_ | 12,742 | https://github.com/huggingface/diffusers/issues/12742 | https://api.github.com/repos/huggingface/diffusers/issues/12742 | Zero tensor in transformer_z_image | ### Describe the bug
In [src/diffusers/models/transformers/transformer_z_image.py](https://github.com/huggingface/diffusers/blob/b010a8ce0c5b5288045045f9f79c496899e80b5a/src/diffusers/models/transformers/transformer_z_image.py#L511), there exists a image_padding_len for padding image to be multiple of SEQ_MULTI_OF. H... | closed | completed | false | 1 | [
"bug"
] | [] | 2025-11-28T18:47:49Z | 2025-12-02T22:51:25Z | 2025-12-02T22:51:25Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lime-j | 14,124,305 | MDQ6VXNlcjE0MTI0MzA1 | User | false |
huggingface/diffusers | 3,676,461,764 | I_kwDOHa8MBc7bIlrE | 12,745 | https://github.com/huggingface/diffusers/issues/12745 | https://api.github.com/repos/huggingface/diffusers/issues/12745 | ZImagePipeline Add Lora | Currently, ZImagePipeline does not support applying LoRA models. This limits the flexibility of the pipeline when trying to use specific styles or customized models.
I would like to request native support for LoRA in ZImagePipeline. Please enable the pipeline to apply LoRA weights. | closed | completed | false | 4 | [
"stale"
] | [] | 2025-11-29T07:29:45Z | 2026-01-10T02:05:30Z | 2026-01-10T02:05:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | snz2 | 21,010,696 | MDQ6VXNlcjIxMDEwNjk2 | User | false |
huggingface/diffusers | 3,676,585,960 | I_kwDOHa8MBc7bJD_o | 12,746 | https://github.com/huggingface/diffusers/issues/12746 | https://api.github.com/repos/huggingface/diffusers/issues/12746 | Flux2 Test error cannot import name 'AutoProcessor' from 'transformers' | diffusers 0.36.0.dev0
transformers 4.57.3
bitsandbytes 0.48.2
Could you help me answer it @sayakpaul | closed | completed | false | 2 | [] | [] | 2025-11-29T10:08:42Z | 2025-12-10T08:39:34Z | 2025-12-10T08:39:34Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Honey-666 | 76,274,942 | MDQ6VXNlcjc2Mjc0OTQy | User | false |
huggingface/diffusers | 3,676,789,341 | I_kwDOHa8MBc7bJ1pd | 12,748 | https://github.com/huggingface/diffusers/issues/12748 | https://api.github.com/repos/huggingface/diffusers/issues/12748 | Please support GGUF format models for Z-Image-Turbo | **Is your feature request related to a problem? Please describe.**
Z-Image currently does not support loading from GGUF format models attempting to do so raises the following error
```
transformer = ZImageTransformer2DModel.from_single_file(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Vol... | closed | completed | false | 0 | [] | [] | 2025-11-29T14:07:27Z | 2025-12-04T14:52:49Z | 2025-12-04T14:52:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Vargol | 62,868 | MDQ6VXNlcjYyODY4 | User | false |
huggingface/diffusers | 3,676,866,410 | I_kwDOHa8MBc7bKIdq | 12,749 | https://github.com/huggingface/diffusers/issues/12749 | https://api.github.com/repos/huggingface/diffusers/issues/12749 | An Idea for partial sequential cpu offloading | Just an idea.
It's not a problem or anything...
I've be using a custom offload for my potato GPU. Maybe there is another way to do it or so...
In short, I've being using sequential offloading for a long time, when I enable it it use a minimal of VRAM, however I know It could use more VRAM to do less IO, so I created... | closed | completed | true | 1 | [
"stale"
] | [] | 2025-11-29T15:41:10Z | 2026-01-10T02:06:01Z | 2026-01-10T02:06:01Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rodjjo | 10,155,053 | MDQ6VXNlcjEwMTU1MDUz | User | false |
huggingface/diffusers | 3,677,037,254 | I_kwDOHa8MBc7bKyLG | 12,752 | https://github.com/huggingface/diffusers/issues/12752 | https://api.github.com/repos/huggingface/diffusers/issues/12752 | Z-Image img2img and inpainting pipeline | ### Model/Pipeline/Scheduler description
I have written and opened a PR for implementation of img2img pipeline for Z-Image.
#12751
<img width="1355" height="1024" alt="Image" src="https://github.com/user-attachments/assets/a3138ce8-522c-4a6d-9ee4-94b2bf1a9992" />
### Open source status
- [x] The model implementati... | open | null | false | 9 | [] | [] | 2025-11-29T19:37:27Z | 2026-01-21T00:28:25Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CalamitousFelicitousness | 63,501,758 | MDQ6VXNlcjYzNTAxNzU4 | User | false |
huggingface/diffusers | 3,677,383,595 | I_kwDOHa8MBc7bMGur | 12,754 | https://github.com/huggingface/diffusers/issues/12754 | https://api.github.com/repos/huggingface/diffusers/issues/12754 | `from_pipe` converts pipelines to float32 by default | ### Describe the bug
Pipelines passed to `from_pipe()` are converted to float32 unless `torch_dtype` is specified, leading to higher memory usage and slower inference.
### Reproduction
```python
import torch
from diffusers import StableDiffusionPipeline, StableDiffusionImg2ImgPipeline
pipe = StableDiffusionPipeline.... | open | null | false | 13 | [
"bug"
] | [] | 2025-11-30T05:06:34Z | 2026-02-12T22:00:47Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | missionfloyd | 4,073,789 | MDQ6VXNlcjQwNzM3ODk= | User | false |
huggingface/diffusers | 3,677,625,577 | I_kwDOHa8MBc7bNBzp | 12,755 | https://github.com/huggingface/diffusers/issues/12755 | https://api.github.com/repos/huggingface/diffusers/issues/12755 | Flux 2: The shape of the latent argument is undocumented | The shape of the `latent` argument is not, as users might guess, `(1, 16, h//8, w//8)`, but `(1, 128, h//16, w//16)`.
This should be documented. | open | null | false | 2 | [
"stale"
] | [] | 2025-11-30T10:39:51Z | 2026-01-09T15:04:03Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rolux | 152,646 | MDQ6VXNlcjE1MjY0Ng== | User | false |
huggingface/diffusers | 1,448,093,004 | I_kwDOHa8MBc5WUCVM | 1,276 | https://github.com/huggingface/diffusers/issues/1276 | https://api.github.com/repos/huggingface/diffusers/issues/1276 | During inpainting, scaling latent before concat break some schedulers. | ### Describe the bug
In pipeline_stable_diffusion_inpaint.py there is this two lines:
```
latent_model_input = torch.cat([latent_model_input, mask, masked_image_latents], dim=1)
latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
```
It seen to break some schedulers, setting scale first s... | closed | completed | false | 1 | [
"bug"
] | [] | 2022-11-14T13:40:31Z | 2022-11-18T11:32:18Z | 2022-11-18T11:32:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | BurguerJohn | 32,472,975 | MDQ6VXNlcjMyNDcyOTc1 | User | false |
huggingface/diffusers | 3,678,985,925 | I_kwDOHa8MBc7bSN7F | 12,760 | https://github.com/huggingface/diffusers/issues/12760 | https://api.github.com/repos/huggingface/diffusers/issues/12760 | `StateManager`, `Hook`, `State`: missing context setting (cond or uncond) in pipelines | In diffusion denoising model, we usually use `classifier free guidance` to control qualities. With [State Manager](https://github.com/huggingface/diffusers/blob/152f7ca357c066c4af3d1a58cdf17662ef5a2f87/src/diffusers/hooks/hooks.py#L42), we want each context hold different `state`, so we have to set which context when c... | open | null | false | 4 | [
"stale",
"contributions-welcome"
] | [] | 2025-12-01T02:48:37Z | 2026-02-08T15:05:54Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | toilaluan | 92,072,154 | U_kgDOBXzo2g | User | false |
huggingface/diffusers | 3,679,054,407 | I_kwDOHa8MBc7bSepH | 12,761 | https://github.com/huggingface/diffusers/issues/12761 | https://api.github.com/repos/huggingface/diffusers/issues/12761 | function TemplatedUlyssesAttentionBackward returned an incorrect number of gradients (expected 12, got 11) | ### Describe the bug
https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_dispatch.py#L1245
https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_dispatch.py#L1337
when using 'flash' as attention_backend, lack one 'None' from return.
### Reproduction
~
### L... | closed | completed | false | 0 | [
"bug"
] | [] | 2025-12-01T03:24:59Z | 2025-12-02T03:22:58Z | 2025-12-02T03:22:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiarenyf | 12,405,578 | MDQ6VXNlcjEyNDA1NTc4 | User | false |
huggingface/diffusers | 3,680,541,593 | I_kwDOHa8MBc7bYJuZ | 12,764 | https://github.com/huggingface/diffusers/issues/12764 | https://api.github.com/repos/huggingface/diffusers/issues/12764 | When will the img2img pipeline of FLUX.2-dev be released? | I see that the current version(0.36.0-dev) only updated the text-to-image pipeline for Flux2. We are looking forward to the update of the image-to-image pipeline!
| closed | completed | true | 2 | [
"stale"
] | [] | 2025-12-01T11:25:35Z | 2026-01-10T02:14:27Z | 2026-01-10T02:14:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | guanxyu | 104,199,757 | U_kgDOBjX2TQ | User | false |
huggingface/diffusers | 3,684,570,462 | I_kwDOHa8MBc7bnhVe | 12,769 | https://github.com/huggingface/diffusers/issues/12769 | https://api.github.com/repos/huggingface/diffusers/issues/12769 | Z-Image Turbo Controlnet Union is out | Worth considering?
https://huggingface.co/alibaba-pai/Z-Image-Turbo-Fun-Controlnet-Union | closed | completed | false | 3 | [] | [] | 2025-12-02T07:49:38Z | 2025-12-17T19:44:22Z | 2025-12-17T19:44:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 3,684,767,307 | I_kwDOHa8MBc7boRZL | 12,771 | https://github.com/huggingface/diffusers/issues/12771 | https://api.github.com/repos/huggingface/diffusers/issues/12771 | DPMSolverMultistepScheduler out-of-bounds error when using beta_schedule="squaredcos_cap_v2" + use_karras_sigmas=True | ### Describe the bug
Hi, I might be misunderstanding something here — I’m not very experienced with submitting issues to Diffusers yet, so please feel free to correct me if I’ve overlooked anything.
When using DPMSolverMultistepScheduler with the combination:
```python
DPMSolverMultistepScheduler(
beta_schedule=... | open | null | false | 4 | [
"bug",
"stale"
] | [] | 2025-12-02T08:46:24Z | 2026-02-04T15:14:53Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | scp3125 | 64,680,608 | MDQ6VXNlcjY0NjgwNjA4 | User | false |
huggingface/diffusers | 3,684,900,877 | I_kwDOHa8MBc7boyAN | 12,772 | https://github.com/huggingface/diffusers/issues/12772 | https://api.github.com/repos/huggingface/diffusers/issues/12772 | How to convert diffusers model to wan2.2 format | I see convert_wan_to_diffusers.py in diffusers repo, but no convert_diffusers_to_wan.py. Do you have plan to upload a convert scripts?
| closed | completed | true | 2 | [
"stale"
] | [] | 2025-12-02T09:19:29Z | 2026-01-10T02:17:07Z | 2026-01-10T02:17:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wikiwen | 38,649,098 | MDQ6VXNlcjM4NjQ5MDk4 | User | false |
huggingface/diffusers | 3,686,823,178 | I_kwDOHa8MBc7bwHUK | 12,776 | https://github.com/huggingface/diffusers/issues/12776 | https://api.github.com/repos/huggingface/diffusers/issues/12776 | Flux2: Tensor tuples can cause issues for checkpointing | ### Describe the bug
The modulations calculated here...
https://github.com/huggingface/diffusers/blob/edf36f5128abf3e6ecf92b5145115514363c58e6/src/diffusers/models/transformers/transformer_flux2.py#L716
...return tuples of Tensors:
https://github.com/huggingface/diffusers/blob/edf36f5128abf3e6ecf92b5145115514363c58e6... | closed | completed | false | 16 | [
"bug"
] | [] | 2025-12-02T17:20:54Z | 2026-06-27T10:39:13Z | 2026-06-27T10:39:13Z | CONTRIBUTOR | null | 20260627T113513Z | 2026-06-27T11:35:13Z | dxqb | 183,307,934 | U_kgDOCu0Ong | User | false |
huggingface/diffusers | 3,688,022,201 | I_kwDOHa8MBc7b0sC5 | 12,778 | https://github.com/huggingface/diffusers/issues/12778 | https://api.github.com/repos/huggingface/diffusers/issues/12778 | AttributeError: type object 'Flux2ImageProcessor' has no attribute 'image_processor' | ### Describe the bug
I'm trying to run train_dreambooth_flux2_img2img.py from the DreamBooth examples, but I encountered this error. I inspected the Flux2ImageProcessor code in diffusers.pipelines.flux2.image_processor, and it turns out that Flux2ImageProcessor does not have an image_processor attribute. The same is t... | closed | completed | false | 3 | [
"bug"
] | [] | 2025-12-03T00:46:25Z | 2026-01-10T02:22:56Z | 2026-01-10T02:22:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hannalaguilar | 55,287,797 | MDQ6VXNlcjU1Mjg3Nzk3 | User | false |
huggingface/diffusers | 3,688,162,032 | I_kwDOHa8MBc7b1OLw | 12,779 | https://github.com/huggingface/diffusers/issues/12779 | https://api.github.com/repos/huggingface/diffusers/issues/12779 | Flux2 dreambooth: AttributeError: 'Namespace' object has no attribute 'validation_image_path' | ### Describe the bug
In train_dreambooth_lora_flux2_img2img.py on line 1422, there is a variable named validation_image_path, but shouldn’t it actually be validation_image?
` validation_image = load_image(args.validation_image_path).convert("RGB")
AttributeError: 'Namespace' object has no attribute 'validation_imag... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2025-12-03T02:01:07Z | 2026-01-10T02:24:02Z | 2026-01-10T02:24:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hannalaguilar | 55,287,797 | MDQ6VXNlcjU1Mjg3Nzk3 | User | false |
huggingface/diffusers | 3,689,364,089 | I_kwDOHa8MBc7b5zp5 | 12,781 | https://github.com/huggingface/diffusers/issues/12781 | https://api.github.com/repos/huggingface/diffusers/issues/12781 | Impossible to log into Huggingface/Diffusers Discord | ### Describe the bug
When trying to verify my Discord/Huggingface account, no matter what I do, I end up with this message:
<img width="512" height="217" alt="Image" src="https://github.com/user-attachments/assets/d1d0f18b-c80f-4862-abde-fb49ee505ddd" />
Has the HF Discord died? If that is the case, what alternative... | closed | completed | false | 4 | [
"bug"
] | [] | 2025-12-03T09:42:55Z | 2025-12-04T15:11:42Z | 2025-12-04T15:11:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 3,690,811,960 | I_kwDOHa8MBc7b_VI4 | 12,783 | https://github.com/huggingface/diffusers/issues/12783 | https://api.github.com/repos/huggingface/diffusers/issues/12783 | Sage Attention sm90 causes confetti/noisy output | ### Describe the bug
When using
_sage_qk_int8_pv_fp8_cuda_sm90 as the attention backend on WAN2.2 I2V I notice that the output is broken:
<img width="1259" height="708" alt="Image" src="https://github.com/user-attachments/assets/89739db5-8602-45b3-bdec-907a2a6d593c" />
It works fine with _flash_3_hub and _sage_qk_i... | open | null | false | 3 | [
"bug"
] | [
"DN6"
] | 2025-12-03T15:53:07Z | 2026-02-20T03:42:40Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dylanprins | 30,610,302 | MDQ6VXNlcjMwNjEwMzAy | User | false |
huggingface/diffusers | 3,691,671,611 | I_kwDOHa8MBc7cCnA7 | 12,786 | https://github.com/huggingface/diffusers/issues/12786 | https://api.github.com/repos/huggingface/diffusers/issues/12786 | [PRX] Improving export to Tensorrt | **Is your feature request related to a problem? Please describe.**
When I export PRX model to ONNX and try to build TensorRT Engine it fails because `Col2Im` operator is not supported. This `Col2Im` is common when using `torch.nn.functional.fold` and `torch.nn.functional.unfold`. Those functions are used in [img2seq f... | closed | completed | false | 0 | [] | [] | 2025-12-03T19:51:02Z | 2025-12-08T12:12:18Z | 2025-12-08T12:12:18Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | WaterKnight1998 | 41,203,448 | MDQ6VXNlcjQxMjAzNDQ4 | User | false |
huggingface/diffusers | 1,314,016,109 | I_kwDOHa8MBc5OUktt | 128 | https://github.com/huggingface/diffusers/issues/128 | https://api.github.com/repos/huggingface/diffusers/issues/128 | how to conditional diffusion with two classes label | null | closed | completed | false | 2 | [] | [] | 2022-07-22T01:22:40Z | 2022-08-16T08:41:30Z | 2022-07-22T12:08:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Johnson-yue | 10,268,274 | MDQ6VXNlcjEwMjY4Mjc0 | User | false |
huggingface/diffusers | 1,448,398,501 | I_kwDOHa8MBc5WVM6l | 1,280 | https://github.com/huggingface/diffusers/issues/1280 | https://api.github.com/repos/huggingface/diffusers/issues/1280 | DDPM pipeline with black and white 1-channel image 'L" | I am trying to generate synthetic images starting from a black and white 1-channel training data set using the DDPMPipeline.
This is my UNet2D model
```
model = UNet2DModel(
sample_size=config.image_size, # the target image resolution
in_channels = 1, #3, # the number of input channels, 3 for RGB im... | closed | completed | false | 4 | [
"stale"
] | [] | 2022-11-14T16:46:28Z | 2023-12-16T15:49:49Z | 2022-12-24T15:03:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zepmck | 73,830,233 | MDQ6VXNlcjczODMwMjMz | User | false |
huggingface/diffusers | 3,709,127,884 | I_kwDOHa8MBc7dFMzM | 12,809 | https://github.com/huggingface/diffusers/issues/12809 | https://api.github.com/repos/huggingface/diffusers/issues/12809 | Kandinsky5TimeEmbeddings hardcodes 'cuda' in @torch.autocast decorator, causing warning on non-CUDA systems | Body:
### Describe the bug
When importing `diffusers` on a non-CUDA system (e.g., Apple Silicon Mac with MPS), a warning is
emitted:
/torch/amp/autocast_mode.py:270: UserWarning: User provided device_type of 'cuda', but CUDA is not
available. Disabling
warnings.warn(
This occurs because `Kandinsky5... | closed | completed | false | 6 | [] | [] | 2025-12-09T04:12:40Z | 2026-01-08T19:00:59Z | 2026-01-08T19:00:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | knd0331 | 24,603,107 | MDQ6VXNlcjI0NjAzMTA3 | User | false |
huggingface/diffusers | 1,448,547,750 | I_kwDOHa8MBc5WVxWm | 1,281 | https://github.com/huggingface/diffusers/issues/1281 | https://api.github.com/repos/huggingface/diffusers/issues/1281 | what is the meaning of parameter: "num_class_images" | What is the `num_class_images `parameter used for? I see that in some examples it is 50, sometimes it is 200. In the source code it is said that: "Minimal class images for prior preservation loss. If not have enough images, additional images will be sampled with class_prompt."
I still do not fully grasp it. For exampl... | closed | completed | false | 7 | [] | [
"williamberman",
"patil-suraj"
] | 2022-11-14T18:32:22Z | 2022-12-06T01:47:42Z | 2022-12-06T01:47:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | himmetozcan | 44,242,024 | MDQ6VXNlcjQ0MjQyMDI0 | User | false |
huggingface/diffusers | 3,713,702,763 | I_kwDOHa8MBc7dWptr | 12,817 | https://github.com/huggingface/diffusers/issues/12817 | https://api.github.com/repos/huggingface/diffusers/issues/12817 | Implement Selective Attention for Memory-Efficient Inference | # [Feature Request] Implement Selective Attention for Memory-Efficient Inference
**Is your feature request related to a problem? Please describe.**
Yes, it is hard to generate longer videos or process high-resolution sequences with video diffusion models (CogVideoX, Mochi, LTX, Hunyuan Video, Wan 2.2). These models f... | closed | completed | false | 5 | [
"stale"
] | [
"cdutr"
] | 2025-12-10T05:50:36Z | 2026-01-09T15:25:20Z | 2026-01-09T15:25:20Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | cdutr | 13,859,415 | MDQ6VXNlcjEzODU5NDE1 | User | false |
huggingface/diffusers | 3,717,368,522 | I_kwDOHa8MBc7dkorK | 12,818 | https://github.com/huggingface/diffusers/issues/12818 | https://api.github.com/repos/huggingface/diffusers/issues/12818 | train_controlnet | Using the official training script, an error occurs and the fill50k dataset cannot be loaded.
Traceback (most recent call last):
File "/root/autodl-tmp/diffusers/examples/controlnet/train_controlnet.py", line 1197, in <module>
main(args)
File "/root/autodl-tmp/diffusers/examples/controlnet/train_controlnet.py",... | closed | completed | false | 2 | [] | [] | 2025-12-11T01:38:35Z | 2026-01-15T00:55:53Z | 2025-12-11T05:23:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Mandarin1126 | 223,811,304 | U_kgDODVcW6A | User | false |
huggingface/diffusers | 3,717,888,228 | I_kwDOHa8MBc7dmnjk | 12,821 | https://github.com/huggingface/diffusers/issues/12821 | https://api.github.com/repos/huggingface/diffusers/issues/12821 | train_controlnet | from diffusers import StableDiffusionControlNetPipeline, ControlNetModel, UniPCMultistepScheduler
from diffusers.utils import load_image
import torch
base_model_path = "stable-diffusion-v1-5/stable-diffusion-v1-5"
controlnet_path = "./output/checkpoint-12500/controlnet/"
controlnet = ControlNetModel.from_pretrained(c... | open | null | false | 1 | [
"stale"
] | [] | 2025-12-11T05:53:15Z | 2026-01-10T15:02:41Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Mandarin1126 | 223,811,304 | U_kgDODVcW6A | User | false |
huggingface/diffusers | 3,717,996,262 | I_kwDOHa8MBc7dnB7m | 12,823 | https://github.com/huggingface/diffusers/issues/12823 | https://api.github.com/repos/huggingface/diffusers/issues/12823 | How to use quantizer after pipeline loaded? | How to use quantizer after pipeline loaded?
- Currently
```python
# Quantization occurs at load time.
pipe = QwenImagePipeline.from_pretrained(
(
args.model_path
if args.model_path is not None
else os.environ.get(
"QWEN_IMAGE_DIR",
"Qwen/Qwen-Image",
)
... | closed | completed | false | 3 | [] | [] | 2025-12-11T06:32:38Z | 2026-01-10T16:05:45Z | 2026-01-10T16:05:44Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DefTruth | 31,974,251 | MDQ6VXNlcjMxOTc0MjUx | User | false |
huggingface/diffusers | 3,718,166,188 | I_kwDOHa8MBc7dnras | 12,826 | https://github.com/huggingface/diffusers/issues/12826 | https://api.github.com/repos/huggingface/diffusers/issues/12826 | cannot import name 'export_to_video' from 'diffusers | ### Describe the bug
tried different versions of diffusers but getting this error, trying to run the https://github.com/SkyworkAI/SkyReels-V2
### Reproduction
tried different versions of diffusers but getting this error, trying to run the https://github.com/SkyworkAI/SkyReels-V2
### Logs
```shell
tried different v... | closed | completed | false | 4 | [
"bug"
] | [] | 2025-12-11T07:31:21Z | 2026-01-10T16:03:59Z | 2026-01-10T16:03:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Arslan-Mehmood1 | 51,626,734 | MDQ6VXNlcjUxNjI2NzM0 | User | false |
huggingface/diffusers | 3,727,270,927 | I_kwDOHa8MBc7eKaQP | 12,838 | https://github.com/huggingface/diffusers/issues/12838 | https://api.github.com/repos/huggingface/diffusers/issues/12838 | Merge Loras for FLUX | The issue is based on https://huggingface.co/docs/diffusers/main/using-diffusers/merge_loras
Is there a similar procedure for merging loras for FLUX models? The guide seems to be specific for UNet based methods. I'm working on FLUX-dev and I would like to perform a linear merge of my loras. | open | null | false | 1 | [
"stale"
] | [] | 2025-12-14T12:39:41Z | 2026-01-13T15:04:18Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shrikrishnalolla | 52,192,798 | MDQ6VXNlcjUyMTkyNzk4 | User | false |
huggingface/diffusers | 1,448,646,576 | I_kwDOHa8MBc5WWJew | 1,284 | https://github.com/huggingface/diffusers/issues/1284 | https://api.github.com/repos/huggingface/diffusers/issues/1284 | Increase the batch size cause longer sample time | ### Describe the bug
I tried to use the DDPMPipeline to generate images and test the model's inference time.
However, I found when I increased the batch size, the sampling time takes longer time. For example, when I set the number of steps as 51, and batch size at 2, the total sample time is `4.72`s on the RTX 2080 G... | closed | completed | false | 2 | [
"bug"
] | [] | 2022-11-14T19:51:54Z | 2022-11-18T16:22:32Z | 2022-11-18T16:22:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jsw-zorro | 31,890,568 | MDQ6VXNlcjMxODkwNTY4 | User | false |
huggingface/diffusers | 3,729,031,076 | I_kwDOHa8MBc7eRH-k | 12,840 | https://github.com/huggingface/diffusers/issues/12840 | https://api.github.com/repos/huggingface/diffusers/issues/12840 | Fix the warning `torch_dtype` is deprecated | ## Description
When loading `QwenImageEditPipeline` (or other pipelines with Transformers models) using `torch_dtype` parameter, a deprecation warning is displayed:
```python
pipeline = QwenImageEditPipeline.from_pretrained(
"Qwen/Qwen-Image-Edit",
torch_dtype=torch.bfloat16,
)
```
Warning message:
```
`torch... | closed | completed | false | 0 | [] | [] | 2025-12-15T06:57:09Z | 2026-01-09T03:05:27Z | 2026-01-09T03:05:27Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | msdsm | 66,203,251 | MDQ6VXNlcjY2MjAzMjUx | User | false |
huggingface/diffusers | 3,729,054,967 | I_kwDOHa8MBc7eRNz3 | 12,843 | https://github.com/huggingface/diffusers/issues/12843 | https://api.github.com/repos/huggingface/diffusers/issues/12843 | [Z-Image] ONNX export fails due to pad_sequence / unbind / dynamic-shape ops in ZImageTransformer2DModel (request for export-friendly path) | ### Describe the bug
### Problem: ONNX export succeeds, but exported model is not truly dynamic (size-switching fails)
The ONNX export itself does NOT fail. The exported ONNX model looks valid and can run if the inference image size matches the export-time size.
However, Z-Image in PyTorch supports running at multip... | open | null | false | 1 | [
"bug",
"stale"
] | [] | 2025-12-15T07:04:58Z | 2026-01-14T15:04:40Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | er6y | 58,937,228 | MDQ6VXNlcjU4OTM3MjI4 | User | false |
huggingface/diffusers | 3,741,698,349 | I_kwDOHa8MBc7fBckt | 12,859 | https://github.com/huggingface/diffusers/issues/12859 | https://api.github.com/repos/huggingface/diffusers/issues/12859 | load_lora_weights(), unload_lora_weights these doesn't work with LongCatImagePipeline. | ### Model/Pipeline/Scheduler description
load_lora_weights(), unload_lora_weights these doesn't work with LongCatImagePipeline. Please add lora support for this model in diffusers.
<img width="784" height="141" alt="Image" src="https://github.com/user-attachments/assets/ebf55ed2-26d9-4cb2-b771-050cc3993cbf" />
### O... | open | null | false | 3 | [
"stale"
] | [] | 2025-12-18T06:40:40Z | 2026-01-17T15:02:40Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ovi054 | 59,303,481 | MDQ6VXNlcjU5MzAzNDgx | User | false |
huggingface/diffusers | 3,750,457,982 | I_kwDOHa8MBc7fi3J- | 12,864 | https://github.com/huggingface/diffusers/issues/12864 | https://api.github.com/repos/huggingface/diffusers/issues/12864 | Please create a Diffusers SKILL for Claude and a Diffusers Plugin for Claude Code | Please create a Diffusers SKILL for Claude and a Diffusers Plugin for Claude Code. It is very hard to navigate the library, and having Claude to guide us would make things much faster. | open | null | false | 1 | [] | [] | 2025-12-20T21:52:41Z | 2026-01-20T16:25:14Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Emasoft | 713,559 | MDQ6VXNlcjcxMzU1OQ== | User | false |
huggingface/diffusers | 1,448,852,327 | I_kwDOHa8MBc5WW7tn | 1,287 | https://github.com/huggingface/diffusers/issues/1287 | https://api.github.com/repos/huggingface/diffusers/issues/1287 | Update UNet1D Tests to not use Mish | **Is your feature request related to a problem? Please describe.**
Currently the `UNet1DTests` heavily rely on mish because they were based on a re-implementation of a model. This is not needed generally, so it would be good to add another 1d model just for tests so we can delete a lot of the tests that look like this... | closed | completed | false | 5 | [
"good first issue"
] | [] | 2022-11-14T22:19:39Z | 2023-07-19T12:44:46Z | 2023-07-19T12:44:46Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | natolambert | 10,695,622 | MDQ6VXNlcjEwNjk1NjIy | User | false |
huggingface/diffusers | 3,752,963,400 | I_kwDOHa8MBc7fsa1I | 12,874 | https://github.com/huggingface/diffusers/issues/12874 | https://api.github.com/repos/huggingface/diffusers/issues/12874 | StabilityAI just take down their SD2 from huggingface | https://huggingface.co/spaces/stabilityai/stable-diffusion-3.5-large/discussions/648#6918c1f1bc616beaadd34376
https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/17204#comment-composer-heading | closed | completed | true | 2 | [] | [] | 2025-12-22T09:16:13Z | 2025-12-22T14:04:33Z | 2025-12-22T14:04:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Bili-Sakura | 108,928,239 | U_kgDOBn4c7w | User | false |
huggingface/diffusers | 3,754,791,216 | I_kwDOHa8MBc7fzZEw | 12,878 | https://github.com/huggingface/diffusers/issues/12878 | https://api.github.com/repos/huggingface/diffusers/issues/12878 | This new version of diffusers causes problems when doing inference with Flux 1.D and Wan2.2 Text To Image (explanation below) | ### Describe the bug
A few days ago I decided to upgrade Torch (cuda:gpu) from v2.7.1 to 2.8.0. In addition I also decided to update diffusers from v0.35.0 to v0.36.0. I updated additional components as well due to the Torch upgrade as well as compatibility with the new version of diffusers.
llama-stack
transformers
h... | closed | completed | false | 4 | [
"bug"
] | [] | 2025-12-22T18:27:38Z | 2025-12-23T17:49:00Z | 2025-12-23T17:49:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ukaprch | 107,368,096 | U_kgDOBmZOoA | User | false |
huggingface/diffusers | 1,448,900,239 | I_kwDOHa8MBc5WXHaP | 1,288 | https://github.com/huggingface/diffusers/issues/1288 | https://api.github.com/repos/huggingface/diffusers/issues/1288 | Explore NVIDIA/TransformerEngine for speed/efficiency | **Is your feature request related to a problem? Please describe.**
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
**Describe the solution you'd like**
- https://github.com/NVIDIA/TransformerEngine
- > A library for accelerating Transformer models on NVIDIA GPUs, i... | closed | completed | false | 2 | [
"stale"
] | [] | 2022-11-14T23:06:44Z | 2022-12-24T15:03:15Z | 2022-12-24T15:03:15Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | 0xdevalias | 753,891 | MDQ6VXNlcjc1Mzg5MQ== | User | false |
huggingface/diffusers | 3,755,520,700 | I_kwDOHa8MBc7f2LK8 | 12,880 | https://github.com/huggingface/diffusers/issues/12880 | https://api.github.com/repos/huggingface/diffusers/issues/12880 | Allow users to pass in a custom device_mesh to the enable_parallelism method in the diffusers library |
`def enable_parallelism(
self,
*,
config: Union[ParallelConfig, ContextParallelConfig],
cp_plan: Optional[Dict[str, ContextParallelModelPlan]] = None,
mesh: Optional[DeviceMesh] = None, # Add this parameter
):`
If users have already initialized a device mesh for other parallelism strategies (FSDP... | closed | completed | false | 8 | [
"roadmap"
] | [] | 2025-12-22T23:16:56Z | 2026-03-11T11:12:52Z | 2026-03-11T11:12:52Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pthombre | 204,920,687 | U_kgDODDbXbw | User | false |
huggingface/diffusers | 3,755,748,396 | I_kwDOHa8MBc7f3Cws | 12,881 | https://github.com/huggingface/diffusers/issues/12881 | https://api.github.com/repos/huggingface/diffusers/issues/12881 | Is that a bug of prompt2prompt pipeline with replace word pormpt? | ### Describe the bug
It performance the same when return different cross attention map, is implement error or just the problem with prompt2prompt.
### Reproduction
Use stable-diffusion-2-1:
`images = pipe(["A turtle playing with a ball", "A monkey playing with a ball"],
generator=torch.Generator("cu... | open | null | false | 1 | [
"bug",
"stale"
] | [] | 2025-12-23T01:55:06Z | 2026-01-22T15:06:49Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lincion | 87,743,668 | MDQ6VXNlcjg3NzQzNjY4 | User | false |
huggingface/diffusers | 3,756,844,330 | I_kwDOHa8MBc7f7OUq | 12,884 | https://github.com/huggingface/diffusers/issues/12884 | https://api.github.com/repos/huggingface/diffusers/issues/12884 | Compatibility issues regarding checkpoint/VAE dependency conflicts when Diffusers load Civitai LoRA | Hello everyone, I'm currently learning to use diffusers and would like to ask all my friends a question. I saw a good lora on Civitai, but this lora has requirements for checkpoint and vea. So I downloaded both models as the author requested. However, when I ran the following code, an error occurred.
The specific code ... | closed | completed | true | 1 | [] | [] | 2025-12-23T10:11:27Z | 2025-12-23T13:41:47Z | 2025-12-23T13:41:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hhhFuture | 184,748,857 | U_kgDOCwMLOQ | User | false |
huggingface/diffusers | 3,760,404,590 | I_kwDOHa8MBc7gIzhu | 12,887 | https://github.com/huggingface/diffusers/issues/12887 | https://api.github.com/repos/huggingface/diffusers/issues/12887 | z-image-turbo from_single_file errors | ### Describe the bug
z-image-turbo support for loading transformer using `from_single_file` was added via #12756
and it works for *some* fine-tunes. however, most of z-image-turbo finetunes found on civitai fail as they contain some empty blocks.
note that use of `quantization_config` works around the issue as quant... | closed | completed | false | 2 | [
"bug"
] | [] | 2025-12-24T14:07:04Z | 2026-01-02T16:59:25Z | 2026-01-02T16:59:25Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 3,761,526,992 | I_kwDOHa8MBc7gNFjQ | 12,889 | https://github.com/huggingface/diffusers/issues/12889 | https://api.github.com/repos/huggingface/diffusers/issues/12889 | Question about qwen-image-edit-2511 loading warning | When loading the model qwen-image-edit-2511 using the diffusers library, I encounter the following warning:
The config attributes {'zero_cond_t': True} were passed to QwenImageTransformer2DModel, but are not expected and will be ignored. Please verify your config.json configuration file.
This suggests that the zero_c... | closed | completed | false | 2 | [] | [] | 2025-12-25T07:06:28Z | 2025-12-25T08:56:28Z | 2025-12-25T08:56:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wizardbob | 55,169,618 | MDQ6VXNlcjU1MTY5NjE4 | User | false |
huggingface/diffusers | 3,762,562,762 | I_kwDOHa8MBc7gRCbK | 12,891 | https://github.com/huggingface/diffusers/issues/12891 | https://api.github.com/repos/huggingface/diffusers/issues/12891 | not support load Qwen-Image-Edit-2511 gguf or fp8 safetensors？ |
```
import torch
from diffusers import QwenImageTransformer2DModel, GGUFQuantizationConfig
ckpt_id = r"Qwen/Qwen-Image-Edit-2511"
transformer1 = QwenImageTransformer2DModel.from_single_file(
r"qwen-image-edit-2511-Q2_K.gguf",
quantization_config=GGUFQuantizationConfig(compute_dtype=torch.bfloat16),
torch... | closed | completed | false | 8 | [] | [] | 2025-12-26T02:26:44Z | 2026-02-04T11:39:42Z | 2026-01-02T07:29:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhaoyun0071 | 35,762,050 | MDQ6VXNlcjM1NzYyMDUw | User | false |
huggingface/diffusers | 3,763,732,711 | I_kwDOHa8MBc7gVgDn | 12,893 | https://github.com/huggingface/diffusers/issues/12893 | https://api.github.com/repos/huggingface/diffusers/issues/12893 | Z-Image text sequence length issue | ### Describe the bug
I think there might be an issue with calculating the sequence length of `cap_feat` (which is the text encoder output), and masking it accordingly.
I'm going to use code links from *before* the omni commit, because it's easier to read - but the issue seems exist in both, before and after the omni ... | closed | completed | false | 8 | [
"bug"
] | [] | 2025-12-26T16:52:11Z | 2026-03-24T17:29:21Z | 2026-03-24T17:29:21Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dxqb | 183,307,934 | U_kgDOCu0Ong | User | false |
huggingface/diffusers | 3,764,217,135 | I_kwDOHa8MBc7gXWUv | 12,895 | https://github.com/huggingface/diffusers/issues/12895 | https://api.github.com/repos/huggingface/diffusers/issues/12895 | StableDiffusion3Pipeline crashes with AttributeError: 'NoneType' object has no attribute 'ndim' when using num_images_per_prompt with batched prompts | ### Describe the bug
When using **StableDiffusion3Pipeline** with **batched prompts** (`prompt` as a list) and passing the parameter `num_images_per_prompt` (even when set to `1`), the pipeline crashes with:
```
AttributeError: 'NoneType' object has no attribute 'ndim'
```
This error does **not** occur:
* when `num... | open | null | false | 2 | [
"stale"
] | [] | 2025-12-27T02:04:42Z | 2026-01-26T15:05:54Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FredyRivera-dev | 116,135,955 | U_kgDOBuwYEw | User | false |
huggingface/diffusers | 3,764,247,783 | I_kwDOHa8MBc7gXdzn | 12,896 | https://github.com/huggingface/diffusers/issues/12896 | https://api.github.com/repos/huggingface/diffusers/issues/12896 | The same WAN 2.2 i2v script (Almost) works in CUDA, bur fails in MPS | ### Describe the bug
Hi
I am trying WAN 2.2 I2V in diffusers. I struggled with the terrible result for hours, and finally find that the example code from (https://huggingface.co/Wan-AI/Wan2.2-I2V-A14B-Diffusers) behaves differently. below are the result from mps and cuda. and the code are almost the same except devic... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2025-12-27T02:57:36Z | 2026-01-26T21:25:58Z | 2026-01-26T21:25:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kerufu | 45,701,634 | MDQ6VXNlcjQ1NzAxNjM0 | User | false |
huggingface/diffusers | 3,766,777,754 | I_kwDOHa8MBc7ghHea | 12,899 | https://github.com/huggingface/diffusers/issues/12899 | https://api.github.com/repos/huggingface/diffusers/issues/12899 | Training script of z-image controlnet? | Can diffusers provide training script of z-image controlnet? | open | null | false | 3 | [] | [] | 2025-12-29T08:30:09Z | 2026-02-05T15:14:04Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | universewill | 6,790,730 | MDQ6VXNlcjY3OTA3MzA= | User | false |
huggingface/diffusers | 1,449,090,500 | I_kwDOHa8MBc5WX13E | 1,290 | https://github.com/huggingface/diffusers/issues/1290 | https://api.github.com/repos/huggingface/diffusers/issues/1290 | [Community] DreamArtist : One-Shot Text-to-Image Generation For Diffusers | **Is your feature request related to a problem? Please describe.**
There is new, One-Shot Text-to-Image for Texual Inversion
[Twitter Link](https://twitter.com/nearcyan/status/1591885331837898753)
There is implementation for Automatic1111, can we have same for diffusers?
[Github Code](https://github.com/7eu7d... | open | null | false | 8 | [
"enhancement",
"New pipeline/model",
"community-examples"
] | [] | 2022-11-15T03:14:15Z | 2023-03-18T18:47:44Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | adhikjoshi | 11,740,719 | MDQ6VXNlcjExNzQwNzE5 | User | false |
huggingface/diffusers | 3,766,914,351 | I_kwDOHa8MBc7gho0v | 12,900 | https://github.com/huggingface/diffusers/issues/12900 | https://api.github.com/repos/huggingface/diffusers/issues/12900 | [Bug]: OOM of Qwen-Image (2048px) generation using Qwen/Qwen-Image | ### Describe the bug
`python -m torch.utils.collect_env`
<details>
<summary>The output of python collect_env.py</summary>
PyTorch version: 2.9.0+cu128
Is debug build: False
CUDA used to build PyTorch: 12.8
ROCM used to build PyTorch: N/A
OS: Ubuntu 24.04.2 LTS (x86_64)
GCC version: (Ubuntu 13.3.0-6ubuntu2~24.04) 13... | closed | completed | false | 2 | [
"bug"
] | [] | 2025-12-29T09:33:42Z | 2025-12-29T15:23:04Z | 2025-12-29T15:23:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LJH-LBJ | 98,734,602 | U_kgDOBeKSCg | User | false |
huggingface/diffusers | 3,769,942,895 | I_kwDOHa8MBc7gtMNv | 12,903 | https://github.com/huggingface/diffusers/issues/12903 | https://api.github.com/repos/huggingface/diffusers/issues/12903 | [Feature Request] Add Implement of 'Kuramoto Orientation Diffusion' | Source code: https://github.com/KingJamesSong/OrientationDiffusion (the official implement use pure `torch`)
This paper introduces a nonlinear score-based generative framework that incorporates stochastic Kuramoto dynamics to model orientation-rich data. By formulating the forward process as synchronization and the re... | open | null | false | 1 | [
"stale"
] | [] | 2025-12-30T11:55:32Z | 2026-01-29T15:10:58Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Bili-Sakura | 108,928,239 | U_kgDOBn4c7w | User | false |
huggingface/diffusers | 3,774,178,783 | I_kwDOHa8MBc7g9WXf | 12,905 | https://github.com/huggingface/diffusers/issues/12905 | https://api.github.com/repos/huggingface/diffusers/issues/12905 | Wrong Z_Image guidance scale implementation | ### Describe the bug
With cfg implementation of Z_image: `pred = pos + current_guidance_scale * (pos - neg)`, the `do_classifier_free_guidance` function should be implement: `return self._guidance_scale > 0`, but currently in [diffusers code](https://github.com/huggingface/diffusers/blob/1cdb8723b85f1b427031e390e0bd0... | closed | completed | false | 10 | [
"bug",
"roadmap"
] | [] | 2026-01-01T06:04:15Z | 2026-02-26T01:08:13Z | 2026-02-26T01:08:13Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Trgtuan10 | 119,487,916 | U_kgDOBx89rA | User | false |
huggingface/diffusers | 3,774,543,041 | I_kwDOHa8MBc7g-vTB | 12,906 | https://github.com/huggingface/diffusers/issues/12906 | https://api.github.com/repos/huggingface/diffusers/issues/12906 | 사이버펑크 | ㅇㄴㅁㅁ | closed | completed | false | 1 | [] | [] | 2026-01-01T11:51:21Z | 2026-01-05T11:24:12Z | 2026-01-05T11:24:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tndud1573 | 252,154,925 | U_kgDODweULQ | User | false |
huggingface/diffusers | 3,774,563,440 | I_kwDOHa8MBc7g-0Rw | 12,907 | https://github.com/huggingface/diffusers/issues/12907 | https://api.github.com/repos/huggingface/diffusers/issues/12907 | 이미지 | ㅇㄴㅁ | closed | completed | false | 0 | [] | [] | 2026-01-01T12:09:01Z | 2026-01-01T12:18:43Z | 2026-01-01T12:18:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tndud1573 | 252,154,925 | U_kgDODweULQ | User | false |
huggingface/diffusers | 3,774,577,468 | I_kwDOHa8MBc7g-3s8 | 12,908 | https://github.com/huggingface/diffusers/issues/12908 | https://api.github.com/repos/huggingface/diffusers/issues/12908 | 15263 | null | closed | completed | false | 1 | [] | [] | 2026-01-01T12:21:35Z | 2026-01-02T13:33:37Z | 2026-01-02T13:33:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tndud1573 | 252,154,925 | U_kgDODweULQ | User | false |
huggingface/diffusers | 3,781,704,961 | I_kwDOHa8MBc7haD0B | 12,913 | https://github.com/huggingface/diffusers/issues/12913 | https://api.github.com/repos/huggingface/diffusers/issues/12913 | Is Lumina2Pipeline's mu calculation correct? | ### Describe the bug
Description
While reviewing the current main-branch implementation of pipeline_lumina2, I noticed a potential bug in the calculation of mu within the pipeline's __call__.
In the following section of the code:
https://github.com/huggingface/diffusers/blob/5ffb65803d0ddc5e3298c35df638ceed5e580922... | open | null | false | 2 | [
"bug",
"stale"
] | [] | 2026-01-05T14:30:01Z | 2026-02-04T15:14:09Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hwangdonghyun | 11,488,276 | MDQ6VXNlcjExNDg4Mjc2 | User | false |
huggingface/diffusers | 3,787,184,301 | I_kwDOHa8MBc7hu9it | 12,918 | https://github.com/huggingface/diffusers/issues/12918 | https://api.github.com/repos/huggingface/diffusers/issues/12918 | Is there any latest information about CVE-2025-14922? | Hugging Face Diffusers CogView4 Deserialization of Untrusted Data Remote Code Execution Vulnerability.
Is there any latest information about CVE-2025-14922 and the scope of the affected versions?
https://nvd.nist.gov/vuln/detail/CVE-2025-14922
https://www.zerodayinitiative.com/advisories/ZDI-25-1142/ | closed | completed | false | 2 | [] | [] | 2026-01-07T03:06:04Z | 2026-01-07T18:23:31Z | 2026-01-07T18:23:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | felnd | 18,362,602 | MDQ6VXNlcjE4MzYyNjAy | User | false |
huggingface/diffusers | 3,788,067,108 | I_kwDOHa8MBc7hyVEk | 12,920 | https://github.com/huggingface/diffusers/issues/12920 | https://api.github.com/repos/huggingface/diffusers/issues/12920 | Any plan on LTX-2? | LTX-2 is open sourced. But the original library is too heavily wrapped, making it less flexible. | closed | completed | false | 1 | [] | [] | 2026-01-07T09:52:21Z | 2026-01-07T09:56:27Z | 2026-01-07T09:56:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | gitlabspy | 61,307,585 | MDQ6VXNlcjYxMzA3NTg1 | User | false |
huggingface/diffusers | 3,791,524,372 | I_kwDOHa8MBc7h_hIU | 12,925 | https://github.com/huggingface/diffusers/issues/12925 | https://api.github.com/repos/huggingface/diffusers/issues/12925 | LTX-2 distilled checkpoint support | https://github.com/huggingface/diffusers/pull/12915 added support for LTX-2.
## What's supported?
* Single stage T2V and I2V
* Upsampling
Amongst other things, we don't support running inference with the [distilled checkpoint](https://huggingface.co/Lightricks/LTX-2/blob/main/ltx-2-19b-distilled.safetensors). This t... | closed | completed | false | 8 | [
"contributions-welcome",
"diffusers-mvp"
] | [] | 2026-01-08T06:32:32Z | 2026-01-29T03:35:45Z | 2026-01-29T03:35:45Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 3,791,542,713 | I_kwDOHa8MBc7h_lm5 | 12,926 | https://github.com/huggingface/diffusers/issues/12926 | https://api.github.com/repos/huggingface/diffusers/issues/12926 | LTX-2 condition pipeline | https://github.com/huggingface/diffusers/pull/12915 added support for LTX-2.
## What's supported?
* Single stage T2V and I2V
* Upsampling
Similar to the [LTX condition pipeline](https://huggingface.co/docs/diffusers/main/api/pipelines/ltx_video#diffusers.LTXConditionPipeline), it would be nice to have support for so... | closed | completed | false | 15 | [
"contributions-welcome",
"diffusers-mvp"
] | [] | 2026-01-08T06:36:38Z | 2026-03-05T08:42:56Z | 2026-03-05T08:42:56Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 3,797,292,165 | I_kwDOHa8MBc7iVhSF | 12,935 | https://github.com/huggingface/diffusers/issues/12935 | https://api.github.com/repos/huggingface/diffusers/issues/12935 | Please create a Huggingface Diffusers SKILL for Claude and a PLUGIN for Claude Code. | ### Feature request
Please create a Huggingface Diffusers SKILL for Claude and a PLUGIN for Claude Code.
### Motivation
It is very hard to navigate all the features of the Diffusers library. Letting Claude guide us would make things much faster.
### Your contribution
I'll help testing the SKILL on my mac M4. | open | null | false | 4 | [] | [] | 2026-01-09T15:08:22Z | 2026-02-10T15:24:36Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Emasoft | 713,559 | MDQ6VXNlcjcxMzU1OQ== | User | false |
huggingface/diffusers | 1,449,653,759 | I_kwDOHa8MBc5WZ_X_ | 1,295 | https://github.com/huggingface/diffusers/issues/1295 | https://api.github.com/repos/huggingface/diffusers/issues/1295 | Unable to run multiple prompts in stable diffusion fast jax.ipynb | ### Describe the bug
Adding multiple prompt lines as described in the blog https://huggingface.co/blog/stable_diffusion_jax
prompt_ids = pipeline.prepare_inputs(prompts)
prompt_ids = shard(prompt_ids)
images = pipeline(prompt_ids, p_params, rng, jit=True).images
images = images.reshape((images.shape[0], ) + imag... | closed | completed | false | 1 | [
"bug"
] | [] | 2022-11-15T11:49:31Z | 2022-11-18T11:34:20Z | 2022-11-18T11:34:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | G-force78 | 114,336,644 | U_kgDOBtCjhA | User | false |
huggingface/diffusers | 1,449,792,900 | I_kwDOHa8MBc5WahWE | 1,296 | https://github.com/huggingface/diffusers/issues/1296 | https://api.github.com/repos/huggingface/diffusers/issues/1296 | Paella | ### Model/Pipeline/Scheduler description
Fast Text-Conditional Discrete Denoising on Vector-Quantized Latent Spaces
abs: https://arxiv.org/abs/2211.07292
github: https://github.com/dome272/Paella
t2i model requiring less than 10 steps to sample high-fidelity images
### Open source status
- [X] The model im... | closed | completed | false | 4 | [
"New pipeline/model"
] | [] | 2022-11-15T13:33:26Z | 2023-04-19T15:23:35Z | 2023-04-19T15:23:34Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 3,803,046,021 | I_kwDOHa8MBc7ireCF | 12,967 | https://github.com/huggingface/diffusers/issues/12967 | https://api.github.com/repos/huggingface/diffusers/issues/12967 | GGUF remove de-quantizing and converting support, like comfyui | GGUF now needs compile in every running, very slow. Can you use the similar codes like comfyui-GGUF to help our low-vram users. GGUF is very important for us 🤗
https://github.com/city96/ComfyUI-GGUF | closed | completed | false | 18 | [] | [] | 2026-01-12T07:11:35Z | 2026-02-05T14:28:33Z | 2026-01-22T02:11:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zwukong | 39,182,231 | MDQ6VXNlcjM5MTgyMjMx | User | false |
huggingface/diffusers | 1,449,863,251 | I_kwDOHa8MBc5WayhT | 1,297 | https://github.com/huggingface/diffusers/issues/1297 | https://api.github.com/repos/huggingface/diffusers/issues/1297 | train_dreambooth.py: Option to delete the previous step saves when using --save_interval to save google drive space. | Storage space can be an issue when training models, especially when using a Google collab and saving the model to a google drive so it isn't lost when the collab disconnects. It would help if there was an option to delete the previous step saves (or only keep the most recent N saves).
Proposed Example:
```
accele... | closed | completed | false | 8 | [] | [
"pcuenca",
"patil-suraj"
] | 2022-11-15T14:19:10Z | 2023-02-16T08:30:35Z | 2023-02-16T08:30:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nbanyan | 30,988,728 | MDQ6VXNlcjMwOTg4NzI4 | User | false |
huggingface/diffusers | 3,811,801,396 | I_kwDOHa8MBc7jM3k0 | 12,975 | https://github.com/huggingface/diffusers/issues/12975 | https://api.github.com/repos/huggingface/diffusers/issues/12975 | AnimateDiffPipeline performance regression on CPU. | ### Describe the bug
regression PR: https://github.com/huggingface/diffusers/pull/11098 .
The [attn_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py#L2778) will turn the hidden_states into channel last (NHWC) layout, which will have great benefit on CPU when ru... | closed | completed | false | 5 | [
"bug",
"performance"
] | [] | 2026-01-14T06:36:09Z | 2026-01-16T03:18:54Z | 2026-01-16T03:18:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiqing-feng | 107,918,818 | U_kgDOBm614g | User | false |
huggingface/diffusers | 1,450,005,863 | I_kwDOHa8MBc5WbVVn | 1,298 | https://github.com/huggingface/diffusers/issues/1298 | https://api.github.com/repos/huggingface/diffusers/issues/1298 | Changes ~ 14 Nov 2022 have broken stablediffusionimg2img pipeline (at least mps, maybe more) | ### Describe the bug
A very recent update to diffusers has broken the stable diffusion img2img pipeline with the libs and archs below (mps). Have not yet tested other architectures. Likely from changes in pipeline_stable_diffusion_img2img.py .
I'm using nightly torch updates. I rolled torch back from 14 Nov to 1... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2022-11-15T15:53:20Z | 2023-01-03T15:04:12Z | 2023-01-03T15:04:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | NathanSiemers | 10,751,479 | MDQ6VXNlcjEwNzUxNDc5 | User | false |
huggingface/diffusers | 3,818,141,861 | I_kwDOHa8MBc7jlDil | 12,981 | https://github.com/huggingface/diffusers/issues/12981 | https://api.github.com/repos/huggingface/diffusers/issues/12981 | Problems loading LTX-2 transformer GGUFs | ### Describe the bug
Can't seem to load GGUFs for ltx-2. More specifically, `because time_embed.emb.timestep_embedder.linear_1.weight expected shape torch.Size([4096, 256]), but got torch.Size([4096, 512])`
### Reproduction
Running on commit 7f43cb1d7919e69d71516fb088db066f6fb7aa7d, using the weights from this repo:... | closed | completed | false | 7 | [
"bug"
] | [] | 2026-01-15T15:57:24Z | 2026-01-16T06:46:43Z | 2026-01-16T06:46:43Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AmericanPresidentJimmyCarter | 110,263,573 | U_kgDOBpJ9FQ | User | false |
huggingface/diffusers | 3,823,993,642 | I_kwDOHa8MBc7j7YMq | 12,989 | https://github.com/huggingface/diffusers/issues/12989 | https://api.github.com/repos/huggingface/diffusers/issues/12989 | how Flux.2 Klein support both T2I and TI2I | Currently, we already support Flux.2 Klein Text2Image Generation, I want to let it support image editing as well.
#12982 | open | null | false | 2 | [] | [] | 2026-01-16T23:28:01Z | 2026-01-23T04:31:45Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Bili-Sakura | 108,928,239 | U_kgDOBn4c7w | User | false |
huggingface/diffusers | 3,824,633,190 | I_kwDOHa8MBc7j90Vm | 12,990 | https://github.com/huggingface/diffusers/issues/12990 | https://api.github.com/repos/huggingface/diffusers/issues/12990 | [optimization] help us know which kernels we should integrate in Diffusers | This issue is for knowing which kernels we should integrate into the library through [`kernels`](https://github.com/huggingface/kernels/).
Currently, we leverage `kernels` for different attention backends (FA2, FA3, and SAGE). However, other layers can be optimized as well (RMS Norm, for example), depending on the mod... | open | null | false | 13 | [
"performance"
] | [] | 2026-01-17T07:41:01Z | 2026-02-24T09:08:25Z | null | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 3,828,116,084 | I_kwDOHa8MBc7kLGp0 | 12,995 | https://github.com/huggingface/diffusers/issues/12995 | https://api.github.com/repos/huggingface/diffusers/issues/12995 | Does `Cosmos-Predict2.5` pipeline generate video? or reconstruct input videos? | ### Describe the bug
Hi,
I tested `Cosmos-Predict2.5` Video2World, but the generated video is quite different from that generated with NVIDIA's original implementation.
It seems that the pipeline reconstructs the input video rather than generates the next frames.
Can you check whether the pipeline functioned proper... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-01-19T05:54:21Z | 2026-01-27T22:20:46Z | 2026-01-27T22:20:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rebel-dkhong | 190,372,430 | U_kgDOC1jaTg | User | false |
huggingface/diffusers | 3,828,709,692 | I_kwDOHa8MBc7kNXk8 | 12,997 | https://github.com/huggingface/diffusers/issues/12997 | https://api.github.com/repos/huggingface/diffusers/issues/12997 | Why does calculate_dimensions in qwen-image-edit require image height/width to be multiples of 32? | Hi, thanks for the great work!
I have a question regarding the logic in calculate_dimensions.
Currently, the image height and width are constrained to be multiples of 32.
From my understanding:
The VAE has a downsampling factor of 8, so the latent spatial size should require the input dimensions to be multiples of 8.
B... | open | null | false | 0 | [] | [] | 2026-01-19T09:04:23Z | 2026-01-19T09:04:23Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lucky-liuzhihong | 165,557,756 | U_kgDOCd41_A | User | false |
huggingface/diffusers | 1,272,010,750 | I_kwDOHa8MBc5L0Vf- | 13 | https://github.com/huggingface/diffusers/issues/13 | https://api.github.com/repos/huggingface/diffusers/issues/13 | Common API | How much of a common API do you expect there to be across different objects? For example, the `unet.py` file contains a `Trainer`, while other models seem to only rely on their `forward` method, but they have different APIs and signatures.
Same with pipelines, for example with the `BDDM` pipeline having the followin... | closed | completed | false | 9 | [] | [] | 2022-06-15T10:21:27Z | 2022-07-21T19:02:41Z | 2022-07-21T19:02:41Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LysandreJik | 30,755,778 | MDQ6VXNlcjMwNzU1Nzc4 | User | false |
huggingface/diffusers | 1,450,214,387 | I_kwDOHa8MBc5WcIPz | 1,300 | https://github.com/huggingface/diffusers/issues/1300 | https://api.github.com/repos/huggingface/diffusers/issues/1300 | Examples contain hardcoded model names | ### Describe the bug
Commit [c18941b](https://github.com/huggingface/diffusers/commit/c18941b01ad0ea6b07d020f353d81153c632a374) introduced hardcoded model names in the example scripts. Most of them should probably be replaced by `args.pretrained_model_name_or_path`.
### Reproduction
Run one of the scripts with... | closed | completed | false | 6 | [
"bug",
"stale"
] | [
"patrickvonplaten"
] | 2022-11-15T18:40:52Z | 2023-01-22T15:03:17Z | 2023-01-22T15:03:17Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | allo- | 1,482,341 | MDQ6VXNlcjE0ODIzNDE= | User | false |
huggingface/diffusers | 3,829,412,673 | I_kwDOHa8MBc7kQDNB | 13,000 | https://github.com/huggingface/diffusers/issues/13000 | https://api.github.com/repos/huggingface/diffusers/issues/13000 | RAE support |
Can we have support for Representation Autoencoders in diffusers? Since RAE already uses hf load for encoders, it will be much simpler if we can directly from_pretrained the whole autoencoder model.
Link: https://github.com/bytetriper/RAE
Paper: https://arxiv.org/abs/2510.11690
| closed | completed | false | 14 | [
"contributions-welcome"
] | [] | 2026-01-19T12:05:11Z | 2026-03-05T14:47:15Z | 2026-03-05T14:47:15Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bytetriper | 53,568,159 | MDQ6VXNlcjUzNTY4MTU5 | User | false |
huggingface/diffusers | 3,831,054,648 | I_kwDOHa8MBc7kWUE4 | 13,001 | https://github.com/huggingface/diffusers/issues/13001 | https://api.github.com/repos/huggingface/diffusers/issues/13001 | Shape mismatch when loading a Flux2.Klein GGUF | ### Describe the bug
When trying to load a Flux2.Klein GGUF:
`ValueError: double_stream_modulation_img.linear.weight has an expected quantized shape of: (18432, 3072), but received shape: torch.Size([18432, 6144])`
Reproducer and full error low below.
The inferred quantized shape of (18432, 3072) is the correct sha... | open | null | false | 4 | [
"bug"
] | [] | 2026-01-19T20:08:42Z | 2026-01-29T12:56:35Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dxqb | 183,307,934 | U_kgDOCu0Ong | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.