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 2026-06-27 10:39:13 | closed_at stringdate 2022-06-12 22:18:01 2026-06-27 10:39:13 ⌀ | author_association stringclasses 3
values | milestone_title stringclasses 0
values | snapshot_id stringclasses 108
values | extracted_at stringdate 2026-04-07 13:34:13 2026-06-27 11:35:13 | 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 | 3,485,530,744 | I_kwDOHa8MBc7PwPp4 | 12,437 | https://github.com/huggingface/diffusers/issues/12437 | https://api.github.com/repos/huggingface/diffusers/issues/12437 | Support OmniSVG | Id love to see diffusers integration for this new and promising looking SVG model. They are already hosting the weights on huggingface.
https://github.com/OmniSVG/OmniSVG/tree/main
https://huggingface.co/OmniSVG/OmniSVG | closed | completed | false | 2 | [] | [] | 2025-10-06T03:08:07Z | 2025-10-07T05:47:00Z | 2025-10-07T05:47:00Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Meatfucker | 74,834,323 | MDQ6VXNlcjc0ODM0MzIz | User | false |
huggingface/diffusers | 3,488,458,367 | I_kwDOHa8MBc7P7aZ_ | 12,441 | https://github.com/huggingface/diffusers/issues/12441 | https://api.github.com/repos/huggingface/diffusers/issues/12441 | Support Wan2.2-Animate | [Wan2.2-Animate-14B](https://humanaigc.github.io/wan-animate), it's a unified model for character animation and replacement, with holistic movement and expression replication.
https://github.com/user-attachments/assets/351227d0-4edc-4f6c-9bf9-053e53f218e4
We would like open to the community, if anyone is interested, ... | closed | completed | false | 0 | [
"help wanted",
"contributions-welcome"
] | [] | 2025-10-06T18:08:21Z | 2025-11-13T02:52:32Z | 2025-11-13T02:52:32Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | asomoza | 5,442,875 | MDQ6VXNlcjU0NDI4NzU= | User | false |
huggingface/diffusers | 3,489,571,069 | I_kwDOHa8MBc7P_qD9 | 12,444 | https://github.com/huggingface/diffusers/issues/12444 | https://api.github.com/repos/huggingface/diffusers/issues/12444 | [train_sana_sprint] Custom save_hook overwrites trained transformer with frozen model during checkpointing | ### Describe the bug
In the train_sana_sprint_diffusers.py example script, the custom save_model_hook incorrectly saves the wrong model. The isinstance() check is too broad because both the trained transformer and the frozen pretrained_model are instances of the same class. This causes the hook to save the trained mod... | open | null | false | 3 | [
"bug",
"stale"
] | [] | 2025-10-07T00:52:02Z | 2026-01-09T15:09:05Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | riteshrm | 109,967,474 | U_kgDOBo34cg | User | false |
huggingface/diffusers | 3,499,870,539 | I_kwDOHa8MBc7Qm8lL | 12,457 | https://github.com/huggingface/diffusers/issues/12457 | https://api.github.com/repos/huggingface/diffusers/issues/12457 | CineScale | ### Model/Pipeline/Scheduler description
HighRes training free WAN 2.1
### 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://github.com/Eyeline-Labs/CineScale | closed | completed | false | 0 | [] | [] | 2025-10-09T16:36:52Z | 2025-10-09T16:37:50Z | 2025-10-09T16:37:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bhack | 1,710,528 | MDQ6VXNlcjE3MTA1Mjg= | User | false |
huggingface/diffusers | 3,501,262,607 | I_kwDOHa8MBc7QsQcP | 12,458 | https://github.com/huggingface/diffusers/issues/12458 | https://api.github.com/repos/huggingface/diffusers/issues/12458 | [Qwen-image-edit] Batch Inference Issue / Feature Request | ### Describe the bug
I am trying to run the batch inference for 'Qwen-Image-Edit-2509', but it will raise inference error
```
AttributeError Traceback (most recent call last)
.venv/lib/python3.10/site-packages/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_plus.py:633) height = heigh... | open | reopened | false | 13 | [
"stale",
"advanced"
] | [] | 2025-10-10T03:48:32Z | 2026-06-04T12:57:51Z | null | NONE | null | 20260604T133522Z | 2026-06-04T13:35:22Z | Django-Jiang | 43,953,876 | MDQ6VXNlcjQzOTUzODc2 | User | false |
huggingface/diffusers | 3,501,339,658 | I_kwDOHa8MBc7QsjQK | 12,459 | https://github.com/huggingface/diffusers/issues/12459 | https://api.github.com/repos/huggingface/diffusers/issues/12459 | FLUX.1-Kontext-dev batch inference throughput issue | ### Describe the bug
I am trying to use FLUX.1-Kontext-dev in batch inference, but the inference time for batch is linear to batch size, e.g., time of batch size 2 ~= 2* time of batch size 1. This makes me confused, did I get something wrong or this is the issue with this pipeline?
### Reproduction
```
import torch
... | open | null | false | 6 | [
"bug",
"stale"
] | [] | 2025-10-10T04:40:01Z | 2026-02-03T15:17:25Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Django-Jiang | 43,953,876 | MDQ6VXNlcjQzOTUzODc2 | User | false |
huggingface/diffusers | 1,444,222,137 | I_kwDOHa8MBc5WFRS5 | 1,246 | https://github.com/huggingface/diffusers/issues/1246 | https://api.github.com/repos/huggingface/diffusers/issues/1246 | [Dreambooth Example] Attempting to unscale FP16 gradients. | ### Describe the bug
I had the training script working fine but then I updated diffusers to 0.7.2 and now I get the following error:
```sh
Traceback (most recent call last):
File "/tmp/pycharm_project_990/train_dreambooth.py", line 938, in <module>
main(args)
File "/tmp/pycharm_project_990/train_dreambo... | closed | completed | false | 31 | [
"bug",
"training"
] | [
"pcuenca",
"williamberman",
"patil-suraj"
] | 2022-11-10T16:39:07Z | 2024-10-10T22:06:39Z | 2023-01-13T10:28:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jpiabrantes | 2,369,107 | MDQ6VXNlcjIzNjkxMDc= | User | false |
huggingface/diffusers | 3,504,120,567 | I_kwDOHa8MBc7Q3KL3 | 12,465 | https://github.com/huggingface/diffusers/issues/12465 | https://api.github.com/repos/huggingface/diffusers/issues/12465 | Remote VAE feedback | ### Did you like the remote VAE solution?
Absolutely, it is amazing for us AMD users. Until they figure out whatever bug is causing super long vae decode/encodes with AMD cards.. This solution is perfect. Who knows if its even a bug, its been happening for quite a while now unfortunately.
### What can be improved ab... | open | null | false | 2 | [
"stale"
] | [] | 2025-10-10T18:43:34Z | 2026-01-09T15:08:47Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Qurtison | 21,269,537 | MDQ6VXNlcjIxMjY5NTM3 | User | false |
huggingface/diffusers | 3,505,854,276 | I_kwDOHa8MBc7Q9xdE | 12,470 | https://github.com/huggingface/diffusers/issues/12470 | https://api.github.com/repos/huggingface/diffusers/issues/12470 | Add update_unexpected_keys and update_expected_keys APIs to DiffusersQuantizer | **Is your feature request related to a problem? Please describe.**
I am the author of SDNQ quantizer and trying to add native support for pre-quantized models to ModelMixin.from_pretrained.
I need to update the expected and unexpected keys so Diffusers will pass the quantization scales / zero_points as a parameter to... | closed | completed | false | 1 | [] | [] | 2025-10-11T13:43:31Z | 2025-11-29T19:50:16Z | 2025-11-29T19:50:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Disty0 | 47,277,141 | MDQ6VXNlcjQ3Mjc3MTQx | User | false |
huggingface/diffusers | 3,510,417,414 | I_kwDOHa8MBc7RPLgG | 12,477 | https://github.com/huggingface/diffusers/issues/12477 | https://api.github.com/repos/huggingface/diffusers/issues/12477 | Add VideoInpaintPipeline for temporally-consistent diffusion-based video inpainting | ## Feature request
Introduce a new pipeline to extend existing image inpainting capabilities (StableDiffusionInpaintPipeline) to videos. The goal is to provide a native, GPU-optimized API within Diffusers that performs temporally coherent video inpainting instead of independent per-frame processing.
## Motivation
Cur... | open | null | false | 10 | [
"stale"
] | [] | 2025-10-13T15:04:38Z | 2026-02-03T15:17:16Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Aki-07 | 95,642,646 | U_kgDOBbNkFg | User | false |
huggingface/diffusers | 3,513,516,354 | I_kwDOHa8MBc7RbAFC | 12,480 | https://github.com/huggingface/diffusers/issues/12480 | https://api.github.com/repos/huggingface/diffusers/issues/12480 | [Bug] TypeError: CLIPTextModel.__init__() got unexpected keyword argument 'offload_state_dict' (Latest Diffs/Transf) | ### Describe the bug
This is the final, complete bug report draft, incorporating the specific version numbers that led to the dependency conflict.
You can copy and paste the entire text below into a new issue on the Hugging Face diffusers GitHub Issues page: https://github.com/huggingface/diffusers/issues.
📝 Bug Re... | closed | completed | false | 1 | [
"bug"
] | [] | 2025-10-14T11:34:34Z | 2025-11-07T12:47:42Z | 2025-11-07T12:47:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lingbin1964 | 59,083,258 | MDQ6VXNlcjU5MDgzMjU4 | User | false |
huggingface/diffusers | 3,513,524,098 | I_kwDOHa8MBc7RbB-C | 12,481 | https://github.com/huggingface/diffusers/issues/12481 | https://api.github.com/repos/huggingface/diffusers/issues/12481 | Bug] TypeError: CLIPTextModel.__init__() got unexpected keyword argument 'offload_state_dict' (Latest Diffs/Transf) | ### Describe the bug
This is the final, complete bug report draft, incorporating the specific version numbers that led to the dependency conflict.
You can copy and paste the entire text below into a new issue on the Hugging Face diffusers GitHub Issues page: https://github.com/huggingface/diffusers/issues.
📝 Bug Re... | closed | completed | false | 0 | [
"bug"
] | [] | 2025-10-14T11:37:14Z | 2025-10-14T11:43:21Z | 2025-10-14T11:43:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lingbin1964 | 59,083,258 | MDQ6VXNlcjU5MDgzMjU4 | User | false |
huggingface/diffusers | 3,513,528,654 | I_kwDOHa8MBc7RbDFO | 12,482 | https://github.com/huggingface/diffusers/issues/12482 | https://api.github.com/repos/huggingface/diffusers/issues/12482 | TypeError: CLIPTextModel got unexpected kwarg 'offload_state_dict' | ### Describe the bug
This is the final, complete bug report draft, incorporating the specific version numbers that led to the dependency conflict.
You can copy and paste the entire text below into a new issue on the Hugging Face diffusers GitHub Issues page: https://github.com/huggingface/diffusers/issues.
📝 Bug Re... | closed | completed | false | 0 | [
"bug"
] | [] | 2025-10-14T11:38:49Z | 2025-10-14T11:42:48Z | 2025-10-14T11:42:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lingbin1964 | 59,083,258 | MDQ6VXNlcjU5MDgzMjU4 | User | false |
huggingface/diffusers | 3,513,530,357 | I_kwDOHa8MBc7RbDf1 | 12,483 | https://github.com/huggingface/diffusers/issues/12483 | https://api.github.com/repos/huggingface/diffusers/issues/12483 | Dependency Conflict: diffusers 0.35.1 throws TypeError with transformers 4.57.0 | ### Describe the bug
This is the final, complete bug report draft, incorporating the specific version numbers that led to the dependency conflict.
You can copy and paste the entire text below into a new issue on the Hugging Face diffusers GitHub Issues page: https://github.com/huggingface/diffusers/issues.
📝 Bug Re... | closed | completed | false | 3 | [
"bug"
] | [] | 2025-10-14T11:39:24Z | 2025-11-03T05:14:59Z | 2025-11-03T05:14:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lingbin1964 | 59,083,258 | MDQ6VXNlcjU5MDgzMjU4 | User | false |
huggingface/diffusers | 3,513,534,944 | I_kwDOHa8MBc7RbEng | 12,484 | https://github.com/huggingface/diffusers/issues/12484 | https://api.github.com/repos/huggingface/diffusers/issues/12484 | persistent conflicts | ### Describe the bug
This is the final, complete bug report draft, incorporating the specific version numbers that led to the dependency conflict.
You can copy and paste the entire text below into a new issue on the Hugging Face diffusers GitHub Issues page: https://github.com/huggingface/diffusers/issues.
📝 Bug Re... | closed | completed | false | 0 | [
"bug"
] | [] | 2025-10-14T11:41:00Z | 2025-10-14T11:42:22Z | 2025-10-14T11:42:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lingbin1964 | 59,083,258 | MDQ6VXNlcjU5MDgzMjU4 | User | false |
huggingface/diffusers | 3,515,618,841 | I_kwDOHa8MBc7RjBYZ | 12,485 | https://github.com/huggingface/diffusers/issues/12485 | https://api.github.com/repos/huggingface/diffusers/issues/12485 | How to enable Context Parallelism for training | Hi @a-r-r-o-w , I would like to ask you for tips on using Context Parallelism for distributed training.
**Is your feature request related to a problem? Please describe.**
Here is the minimal code for adapting Context Parallelism into diffusion model training
```python
# Diffusers Version: 0.36.0.dev0
from diffusers.m... | closed | completed | false | 0 | [] | [] | 2025-10-14T21:48:35Z | 2025-10-15T20:33:30Z | 2025-10-15T20:33:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | liming-ai | 73,068,772 | MDQ6VXNlcjczMDY4Nzcy | User | false |
huggingface/diffusers | 3,516,242,195 | I_kwDOHa8MBc7RlZkT | 12,489 | https://github.com/huggingface/diffusers/issues/12489 | https://api.github.com/repos/huggingface/diffusers/issues/12489 | [Bug] CUDA kernel compatibility error (sm_120) — Diffusers ControlNet pipeline fails with “no kernel image is available” | ### Describe the bug
CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call.
NVIDIA GeForce RTX 5070 with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA ca... | closed | completed | false | 3 | [
"bug"
] | [] | 2025-10-15T03:59:15Z | 2025-10-29T07:38:33Z | 2025-10-29T07:38:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lingbin1964 | 59,083,258 | MDQ6VXNlcjU5MDgzMjU4 | User | false |
huggingface/diffusers | 1,444,520,583 | I_kwDOHa8MBc5WGaKH | 1,249 | https://github.com/huggingface/diffusers/issues/1249 | https://api.github.com/repos/huggingface/diffusers/issues/1249 | AttributeError: 'numpy.ndarray' object has no attribute 'device' [OnnxPipeline] [Schedulers DDPM, Euler, Euler-A] | ### Describe the bug
[PR 636](https://github.com/huggingface/diffusers/pull/636) and [PR 944](https://github.com/huggingface/diffusers/pull/944) added the [Euler ](https://moon-ci-docs.huggingface.co/docs/diffusers/pr_1244/en/api/schedulers#diffusers.EulerDiscreteScheduler) and [Euler-A](https://moon-ci-docs.hugging... | closed | completed | false | 3 | [
"bug"
] | [
"anton-l"
] | 2022-11-10T20:29:49Z | 2022-11-17T19:44:07Z | 2022-11-17T15:37:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | averad | 640,619 | MDQ6VXNlcjY0MDYxOQ== | User | false |
huggingface/diffusers | 3,518,438,971 | I_kwDOHa8MBc7Rtx47 | 12,492 | https://github.com/huggingface/diffusers/issues/12492 | https://api.github.com/repos/huggingface/diffusers/issues/12492 | module transformers has no attribute CLIPFeatureExtractor | ### System Info
latest main
### Who can help?
@SunMarc
### Information
- [ ] The official example scripts
- [ ] My own modified scripts
### Tasks
- [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...)
- [ ] My own task or dataset (give details below)
### Reproduction
```python
f... | closed | completed | false | 3 | [
"bug"
] | [
"sayakpaul",
"dg845"
] | 2025-10-15T08:26:05Z | 2025-11-03T05:02:54Z | 2025-11-03T05:02:53Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiqing-feng | 107,918,818 | U_kgDOBm614g | User | false |
huggingface/diffusers | 3,519,002,638 | I_kwDOHa8MBc7Rv7gO | 12,494 | https://github.com/huggingface/diffusers/issues/12494 | https://api.github.com/repos/huggingface/diffusers/issues/12494 | i think there is something wrong with new/latest scripts. RuntimeError: mat1 and mat2 shapes cannot be multiplied (2x1536 and 768x3072) | ### Describe the bug
i got "RuntimeError: mat1 and mat2 shapes cannot be multiplied (2x1536 and 768x3072)" when i ran "train_dreambooth_lora_flux_advanced.py" from the latest version of diffusers or v0.35.1 but the problem solved when i downgraded the version to v0.31.0 including all the dependencies. i ran the script... | open | null | false | 3 | [
"bug",
"stale"
] | [] | 2025-10-15T17:45:53Z | 2026-01-09T15:08:28Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | gerylavin | 53,449,230 | MDQ6VXNlcjUzNDQ5MjMw | User | false |
huggingface/diffusers | 3,524,033,659 | I_kwDOHa8MBc7SDHx7 | 12,499 | https://github.com/huggingface/diffusers/issues/12499 | https://api.github.com/repos/huggingface/diffusers/issues/12499 | Wan 2.2 I2V condition shape mismatch | ### Describe the bug
RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 21 but got size 81 for tensor number 1 in the list
(cu129)➜ pipeline git:(dev) ✗ python3 run_wan_2.2_i2v.py
torch.Size([1, 1, 4, 90, 68])
torch.Size([1, 1, 84, 90, 68]) 4
torch.Size([1, 21, 4, 90, 68])
torch.Size([1, 4... | closed | completed | false | 12 | [
"bug"
] | [] | 2025-10-17T01:02:39Z | 2025-10-22T17:53:14Z | 2025-10-22T17:53:14Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DefTruth | 31,974,251 | MDQ6VXNlcjMxOTc0MjUx | User | false |
huggingface/diffusers | 1,313,683,598 | I_kwDOHa8MBc5OTTiO | 125 | https://github.com/huggingface/diffusers/issues/125 | https://api.github.com/repos/huggingface/diffusers/issues/125 | Launch training with accelerate launch arg error | Launching training with `accelerate launch` raise argument errors
```
Traceback (most recent call last):
File "/usr/local/bin/accelerate-launch", line 8, in <module>
sys.exit(main())
File "/usr/local/lib/python3.7/dist-packages/accelerate/commands/launch.py", line 690, in main
launch_command(args)
... | closed | completed | false | 3 | [] | [] | 2022-07-21T19:07:52Z | 2022-07-22T14:59:41Z | 2022-07-22T11:49:12Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | radames | 102,277 | MDQ6VXNlcjEwMjI3Nw== | User | false |
huggingface/diffusers | 3,525,966,585 | I_kwDOHa8MBc7SKfr5 | 12,501 | https://github.com/huggingface/diffusers/issues/12501 | https://api.github.com/repos/huggingface/diffusers/issues/12501 | VAE Decoder next(iter(..)) causes graph break | ### Describe the bug
When doing torch.compile with the decoder part of the VAE, a graph break gets hit. This makes it impossible to compile the model as a whole.
And these are the exact lines from `diffusers` library that cause the graph break:
`diffusers/src/diffusers/models/autoencoders/vae.py:287-289`
```
... | closed | completed | false | 4 | [
"bug"
] | [] | 2025-10-17T12:27:08Z | 2025-10-30T03:09:42Z | 2025-10-30T03:09:42Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ppadjinTT | 169,346,642 | U_kgDOChgGUg | User | false |
huggingface/diffusers | 3,528,758,057 | I_kwDOHa8MBc7SVJMp | 12,504 | https://github.com/huggingface/diffusers/issues/12504 | https://api.github.com/repos/huggingface/diffusers/issues/12504 | freeU ComplexHalf warning | ### Describe the bug
got this warning with sd-turbo+fp16+freeU. just wanted to know if this has actual impact
### Reproduction
```py
import torch
from diffusers import AutoPipelineForText2Image
pipe = AutoPipelineForText2Image.from_pretrained('stabilityai/sd-turbo', torch_dtype=torch.float16, variant='fp16', use_saf... | closed | completed | false | 1 | [
"bug",
"stale"
] | [] | 2025-10-18T12:15:44Z | 2026-04-21T11:45:12Z | 2026-04-21T11:45:12Z | NONE | null | 20260421T123535Z | 2026-04-21T12:35:35Z | eyaler | 4,436,747 | MDQ6VXNlcjQ0MzY3NDc= | User | false |
huggingface/diffusers | 3,529,094,468 | I_kwDOHa8MBc7SWbVE | 12,505 | https://github.com/huggingface/diffusers/issues/12505 | https://api.github.com/repos/huggingface/diffusers/issues/12505 | tgate error on img2img sd/sdxl pipelines | ### Describe the bug
I minimally modified the working example in https://huggingface.co/docs/diffusers/main/en/optimization/tgate?pipelines=Stable+Diffusion+XL to use img2img pipeline which results in the first error, and also tried with SD which gave the second error. both pipelines work when calling without the tgat... | open | null | false | 1 | [
"bug",
"stale"
] | [] | 2025-10-18T18:46:21Z | 2026-01-09T15:08:18Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eyaler | 4,436,747 | MDQ6VXNlcjQ0MzY3NDc= | User | false |
huggingface/diffusers | 1,444,556,882 | I_kwDOHa8MBc5WGjBS | 1,251 | https://github.com/huggingface/diffusers/issues/1251 | https://api.github.com/repos/huggingface/diffusers/issues/1251 | RePaintPipeline always returns black image when pipe.to("mps") is called | ### Describe the bug
When I try to do the inpainting with RePaintPipeline on the MPS backend it only returns black image. It works fine with CPU, but very slow.
### Reproduction
```python
scheduler = RePaintScheduler.from_config("google/ddpm-ema-celebahq-256")
pipe = RePaintPipeline.from_pretrained("google/ddpm-em... | open | reopened | false | 12 | [
"bug",
"stale",
"wip"
] | [
"anton-l"
] | 2022-11-10T21:00:50Z | 2023-03-13T10:48:12Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | GaidamakUA | 2,983,027 | MDQ6VXNlcjI5ODMwMjc= | User | false |
huggingface/diffusers | 1,444,657,589 | I_kwDOHa8MBc5WG7m1 | 1,252 | https://github.com/huggingface/diffusers/issues/1252 | https://api.github.com/repos/huggingface/diffusers/issues/1252 | [In-paint] How to generate mask example? | HI there, given a image, is there a effective way to generate mask example that covers the main object? Thank you. | closed | completed | false | 2 | [] | [] | 2022-11-10T22:36:44Z | 2022-11-13T08:09:32Z | 2022-11-11T19:42:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | elricwan | 33,402,371 | MDQ6VXNlcjMzNDAyMzcx | User | false |
huggingface/diffusers | 3,540,184,029 | I_kwDOHa8MBc7TAuvd | 12,529 | https://github.com/huggingface/diffusers/issues/12529 | https://api.github.com/repos/huggingface/diffusers/issues/12529 | Wan2.2 TI2V-5B Tiled VAE Tensor size mismatch | ### Describe the bug
When running Wan 2.2 TI2V 5B with tiled VAE, the VAE step is unsuccessful, resulting in:
```RuntimeError: The size of tensor a (2) must match the size of tensor b (4) at non-singleton dimension 2```
### Reproduction
```shell
import torch
import numpy as np
from diffusers import WanPipeline, Auto... | closed | completed | false | 3 | [
"bug"
] | [] | 2025-10-22T10:15:05Z | 2025-10-30T02:35:51Z | 2025-10-30T02:35:51Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CalamitousFelicitousness | 63,501,758 | MDQ6VXNlcjYzNTAxNzU4 | User | false |
huggingface/diffusers | 1,444,917,493 | I_kwDOHa8MBc5WH7D1 | 1,253 | https://github.com/huggingface/diffusers/issues/1253 | https://api.github.com/repos/huggingface/diffusers/issues/1253 | Caching Pipes | ### Describe the bug
It used to possible to cache stable Diffusion pipes to load them faster, and more efficiently. But development in diffusers has got in the way of this working smoothly.
- If a model is already downloaded, it still checks HF model conf, and this process is needlessly slow
- It Is now not pos... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2022-11-11T04:33:54Z | 2022-12-21T15:03:14Z | 2022-12-21T15:03:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | WASasquatch | 1,151,589 | MDQ6VXNlcjExNTE1ODk= | User | false |
huggingface/diffusers | 3,543,012,933 | I_kwDOHa8MBc7TLhZF | 12,532 | https://github.com/huggingface/diffusers/issues/12532 | https://api.github.com/repos/huggingface/diffusers/issues/12532 | add train example for qwen-edit-plus | null | open | null | false | 1 | [
"stale"
] | [] | 2025-10-23T03:40:34Z | 2026-01-09T15:07:48Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | akk-123 | 98,469,560 | U_kgDOBd6GuA | User | false |
huggingface/diffusers | 3,543,521,911 | I_kwDOHa8MBc7TNdp3 | 12,533 | https://github.com/huggingface/diffusers/issues/12533 | https://api.github.com/repos/huggingface/diffusers/issues/12533 | Hooks conflicts: Context Parallelism and CPU Offload | ### Describe the bug
Enable cpu offload before enabling parallelism will raise shape error after first pipe call. It seems a bug of diffusers that cpu offload is not fully compatible with context parallelism, visa versa.
- cpu offload `before` context parallelism (not work)
```python
pipe.enable_model_cpu_offload(d... | open | null | false | 5 | [
"bug",
"roadmap",
"context-parallel"
] | [
"yiyixuxu"
] | 2025-10-23T07:33:37Z | 2026-03-10T03:36:33Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DefTruth | 31,974,251 | MDQ6VXNlcjMxOTc0MjUx | User | false |
huggingface/diffusers | 3,544,438,974 | I_kwDOHa8MBc7TQ9i- | 12,535 | https://github.com/huggingface/diffusers/issues/12535 | https://api.github.com/repos/huggingface/diffusers/issues/12535 | Wan2.2 i2v Latest Lightning LoRA not loading | ### Describe the bug
The latest lightning LoRA for Wan2.2 i2v fails to load.
Link to LoRA: [Latest](https://huggingface.co/lightx2v/Wan2.2-Distill-Loras/tree/main)
### Reproduction
```shell
model_id = "Wan-AI/Wan2.2-I2V-A14B-Diffusers"
pipe = WanImageToVideoPipeline.from_pretrained(
model_id,
torch_dtype=t... | closed | completed | false | 0 | [
"bug"
] | [] | 2025-10-23T12:17:00Z | 2025-10-28T03:25:25Z | 2025-10-28T03:25:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ljk1291 | 3,118,493 | MDQ6VXNlcjMxMTg0OTM= | User | false |
huggingface/diffusers | 3,544,453,298 | I_kwDOHa8MBc7TRBCy | 12,536 | https://github.com/huggingface/diffusers/issues/12536 | https://api.github.com/repos/huggingface/diffusers/issues/12536 | AssertionError in Context Parallelism during WanImageToVideoPipeline inference: Tensor size along sharding dimension not divisible by mesh size | Hi! @a-r-r-o-w ,I would like to ask you about my error on using Context Parallelism for inference.
### Issue Description
#### Environment
- **Diffusers**: 0.36.0.dev0
#### Problem Description
I'm trying to run image-to-video generation using `WanImageToVideoPipeline` with model quantization (qfloat8_e4m3fn via Quant... | open | null | false | 6 | [
"stale",
"context-parallel"
] | [] | 2025-10-23T12:21:47Z | 2026-02-03T15:16:48Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | leeguandong | 23,144,950 | MDQ6VXNlcjIzMTQ0OTUw | User | false |
huggingface/diffusers | 3,545,062,927 | I_kwDOHa8MBc7TTV4P | 12,538 | https://github.com/huggingface/diffusers/issues/12538 | https://api.github.com/repos/huggingface/diffusers/issues/12538 | Misallignment in Rotary Frequencies | ### Describe the bug
In `WanRotaryPosEmbed` ([link](https://github.com/huggingface/diffusers/blob/9c3b58dcf16ebd027fd3d85ec703ad5a142b1e1e/src/diffusers/models/transformers/transformer_wan.py#L349)), we are splitting the `attention_head_dim` into the different dimensions in different ways in `__init__` and `forward`.... | closed | completed | false | 2 | [
"bug",
"contributions-welcome"
] | [] | 2025-10-23T14:54:49Z | 2025-11-11T21:45:38Z | 2025-11-11T21:45:38Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Warvito | 4,493,399 | MDQ6VXNlcjQ0OTMzOTk= | User | false |
huggingface/diffusers | 1,445,393,232 | I_kwDOHa8MBc5WJvNQ | 1,254 | https://github.com/huggingface/diffusers/issues/1254 | https://api.github.com/repos/huggingface/diffusers/issues/1254 | How to - Image2Image with (Image + Text) Guided! | I've an **image with target object**. And I like to generate new image with **text-guided background** with the **target object** in a natural view. In short
- An image with target object (i.e. car).
- Complex backgrouond removed by **Model A**.
- Then pass it to **Model B** with **text / prompt**.
- Generate **... | closed | completed | false | 15 | [
"stale"
] | [] | 2022-11-11T12:17:39Z | 2023-05-08T15:04:15Z | 2023-05-08T15:04:15Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | innat | 17,668,390 | MDQ6VXNlcjE3NjY4Mzkw | User | false |
huggingface/diffusers | 3,549,609,259 | I_kwDOHa8MBc7Tkr0r | 12,540 | https://github.com/huggingface/diffusers/issues/12540 | https://api.github.com/repos/huggingface/diffusers/issues/12540 | SanaCombinedTimestepGuidanceEmbeddings do not work with SanaPipeline | ### Describe the bug
When building a `SanaTransformer2DModel`, it lets you which time_embed to use with `guidance_embeds`. The default choice is `AdaLayerNormSingle`, which works fine when calling the pipeline; the other choice, `SanaCombinedTimestepGuidanceEmbeddings`, does not work however.
### Reproduction
```
fr... | open | null | false | 3 | [
"bug",
"stale"
] | [] | 2025-10-24T13:50:15Z | 2026-01-09T15:07:35Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | frutiemax92 | 142,428,698 | U_kgDOCH1KGg | User | false |
huggingface/diffusers | 3,552,960,809 | I_kwDOHa8MBc7TxeEp | 12,542 | https://github.com/huggingface/diffusers/issues/12542 | https://api.github.com/repos/huggingface/diffusers/issues/12542 | add custom output_params to export_to_video | The export_to_video docstring mentions `output_params`: https://github.com/huggingface/diffusers/blob/v0.35.2/src/diffusers/utils/export_utils.py#L152, but this is not implemented, not allowing to pass custom params to ffmpeg.
| open | null | false | 3 | [
"stale"
] | [] | 2025-10-25T20:40:07Z | 2026-01-09T15:07:31Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eyaler | 4,436,747 | MDQ6VXNlcjQ0MzY3NDc= | User | false |
huggingface/diffusers | 3,554,609,775 | I_kwDOHa8MBc7T3wpv | 12,547 | https://github.com/huggingface/diffusers/issues/12547 | https://api.github.com/repos/huggingface/diffusers/issues/12547 | Fine tuning Dreambooth Flux Kontext I2I Error: the following arguments are required: --instance_prompt | ### Describe the bug
Hello HF team, @sayakpaul @bghira
I'm encountering a persistent issue when trying to fine-tune the black-forest-labs/FLUX.1-Kontext-dev model using the train_dreambooth_lora_flux_kontext.py script.
I am following the [official README instructions](https://github.com/huggingface/diffusers/blob/ma... | closed | completed | false | 7 | [
"bug"
] | [] | 2025-10-27T00:21:34Z | 2025-10-28T02:31:42Z | 2025-10-28T02:31:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MichaelMelgarejoFlorez | 218,021,471 | U_kgDODP6-Xw | User | false |
huggingface/diffusers | 1,445,455,176 | I_kwDOHa8MBc5WJ-VI | 1,255 | https://github.com/huggingface/diffusers/issues/1255 | https://api.github.com/repos/huggingface/diffusers/issues/1255 | cannot import name 'DPMSolverMultistepScheduler' from 'diffusers' | ### Describe the bug
because python3.7 ? 🙄
https://github.com/huggingface/diffusers/blob/045157a46fb16a21a5f37a5f3f3ad710895b680b/scripts/convert_original_stable_diffusion_to_diffusers.py#L33
```py
ImportError: cannot import name 'DPMSolverMultistepScheduler' from 'diffusers'
Traceback (most recent call las... | closed | completed | false | 2 | [
"bug"
] | [] | 2022-11-11T13:11:31Z | 2022-11-11T15:36:02Z | 2022-11-11T15:36:02Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | camenduru | 54,370,274 | MDQ6VXNlcjU0MzcwMjc0 | User | false |
huggingface/diffusers | 3,556,745,931 | I_kwDOHa8MBc7T_6LL | 12,550 | https://github.com/huggingface/diffusers/issues/12550 | https://api.github.com/repos/huggingface/diffusers/issues/12550 | Fal Flashpack | **Is your feature request related to a problem? Please describe.**
Safetensors loading seems it could be faster.
**Describe the solution you'd like.**
Recently fal has open sourced a new loading method for faster DIT and TE loading.
**Describe alternatives you've considered.**
[A clear and concise description of any... | closed | completed | false | 12 | [] | [] | 2025-10-27T13:17:16Z | 2026-04-13T15:38:14Z | 2026-04-13T15:38:14Z | NONE | null | 20260417T173542Z | 2026-04-17T17:35:42Z | vgabbo | 105,066,170 | U_kgDOBkMuug | User | false |
huggingface/diffusers | 3,560,069,027 | I_kwDOHa8MBc7UMlej | 12,553 | https://github.com/huggingface/diffusers/issues/12553 | https://api.github.com/repos/huggingface/diffusers/issues/12553 | Reason to move from OpenCV to ffmpeg | I see that `diffusers.utils.export_to_video()` encourages ffmpeg usage instead of OpenCV. Can you share the reason? I'm looking for a way to add video decoding to my project so I'm collecting arguments. | open | null | false | 12 | [
"stale"
] | [] | 2025-10-28T06:49:48Z | 2026-02-03T15:16:43Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Wovchena | 10,669,582 | MDQ6VXNlcjEwNjY5NTgy | User | false |
huggingface/diffusers | 3,566,330,754 | I_kwDOHa8MBc7UkeOC | 12,558 | https://github.com/huggingface/diffusers/issues/12558 | https://api.github.com/repos/huggingface/diffusers/issues/12558 | Please add QwenImageEditPlusPipeline to 0.36 | It's currently not in the roadmap.
https://github.com/orgs/huggingface/projects/61/views/1?layout=table | closed | completed | false | 1 | [] | [] | 2025-10-29T14:37:30Z | 2025-10-29T15:56:35Z | 2025-10-29T15:56:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yz342 | 85,502,134 | MDQ6VXNlcjg1NTAyMTM0 | User | false |
huggingface/diffusers | 3,566,429,419 | I_kwDOHa8MBc7Uk2Tr | 12,559 | https://github.com/huggingface/diffusers/issues/12559 | https://api.github.com/repos/huggingface/diffusers/issues/12559 | nms不存在！！！我在win系统下边实现这个examples/instruct_pix2pix的报错RuntimeError: operator torchvision::nms does not exist | ### Describe the bug
(img2img) PS C:\Study\Study_Documents\PyCharm\diffusers-main\examples\instruct_pix2pix> accelerate launch --mixed_precision="fp16" train_instruct_pix2pix.py `
>> --pretrained_model_name_or_path=$env:MODEL_NAME `
>> --dataset_name=$env:DATASET_ID `
>> --enable_xformers_memory_efficient_attent... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2025-10-29T14:57:58Z | 2026-01-10T00:30:41Z | 2026-01-10T00:30:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CSolaris | 223,991,768 | U_kgDODVnX2A | User | false |
huggingface/diffusers | 3,573,153,956 | I_kwDOHa8MBc7U-gCk | 12,564 | https://github.com/huggingface/diffusers/issues/12564 | https://api.github.com/repos/huggingface/diffusers/issues/12564 | [Proposals Welcome] Fal Flashpack integration for faster model loading | Hey! 👋
We've had a request to explore integrating Fal's Flashpack for faster DiT and Text Encoder loading (https://github.com/huggingface/diffusers/issues/12550). Before we jump into implementation, we wanted to open this up to the community to gather ideas and hear from anyone who's experimented with this.
We'd lov... | open | null | false | 4 | [
"help wanted",
"stale",
"contributions-welcome"
] | [] | 2025-10-31T02:25:55Z | 2026-02-03T15:16:35Z | null | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 3,573,372,024 | I_kwDOHa8MBc7U_VR4 | 12,565 | https://github.com/huggingface/diffusers/issues/12565 | https://api.github.com/repos/huggingface/diffusers/issues/12565 | Wan2.1 i2v running issue | ### Describe the bug
When I was trying to run the script which is python cache-dit/examples/pipelinerun_wan_2.2.py by the weight of Wan2.1-I2V-14B-720P-Diffusers, an error displayed.
<img width="1587" height="216" alt="Image" src="https://github.com/user-attachments/assets/916db755-4e0f-4084-b5c7-3c728375a6c9" />
Tr... | open | null | false | 10 | [
"bug",
"stale"
] | [] | 2025-10-31T04:34:34Z | 2026-02-03T15:16:32Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | luren55 | 97,496,981 | U_kgDOBc-vlQ | User | false |
huggingface/diffusers | 3,575,488,915 | I_kwDOHa8MBc7VHaGT | 12,567 | https://github.com/huggingface/diffusers/issues/12567 | https://api.github.com/repos/huggingface/diffusers/issues/12567 | generate noise when using Ulysses Attention | ### Describe the bug
When running **Wan2.1-T2V-14B-Diffusers** model with **Ulysses Attention**, the last few frames of the generated vedio are completely noisy.
### Reproduction
```
# infer_wan_t2v.py
import numpy as np
import torch
from diffusers import AutoencoderKLWan, WanPipeline, ContextParallelConfig, WanT... | open | null | false | 3 | [
"bug",
"stale"
] | [] | 2025-10-31T15:50:38Z | 2026-02-03T15:16:28Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TmacAaron | 83,484,024 | MDQ6VXNlcjgzNDg0MDI0 | User | false |
huggingface/diffusers | 3,576,990,063 | I_kwDOHa8MBc7VNIlv | 12,568 | https://github.com/huggingface/diffusers/issues/12568 | https://api.github.com/repos/huggingface/diffusers/issues/12568 | Context parallel bug when using QwenImage | Refer more here: https://github.com/huggingface/diffusers/pull/12443
More context under "Result: Single coherent image matching single-GPU output" [here](https://github.com/huggingface/diffusers/pull/12443#issue-3489494846).
Before proceeding with a PR, please reply to this issue mentioning the proposal you have in m... | open | null | false | 8 | [
"Good second issue",
"contributions-welcome"
] | [] | 2025-11-01T01:38:15Z | 2025-11-11T11:08:55Z | null | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 3,576,996,824 | I_kwDOHa8MBc7VNKPY | 12,569 | https://github.com/huggingface/diffusers/issues/12569 | https://api.github.com/repos/huggingface/diffusers/issues/12569 | [feature] implement TaylorSeer | Currently, we support PAB and FasterCache. More details: https://huggingface.co/docs/diffusers/main/en/optimization/cache.
It'd be cool to support [Accelerating Diffusion Models with TaylorSeers](https://arxiv.org/abs/2503.06923) as it's quite neat and significantly better than others. Code is here:
https://github.co... | closed | completed | false | 15 | [
"contributions-welcome",
"performance"
] | [] | 2025-11-01T01:43:00Z | 2025-12-08T08:54:53Z | 2025-12-08T08:54:53Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,446,016,938 | I_kwDOHa8MBc5WMHeq | 1,257 | https://github.com/huggingface/diffusers/issues/1257 | https://api.github.com/repos/huggingface/diffusers/issues/1257 | Standardize on `init_image` for img2img, inpaint, and others | **What API design would you like to have changed or added to the library? Why?**
[img2img](https://github.com/huggingface/diffusers/blob/3d98dc763a645423f992b3546c4d9cdd97eb9d04/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py#L312) currently uses the `init_image` parameter, whereas [in... | closed | completed | false | 2 | [] | [
"pcuenca",
"patil-suraj"
] | 2022-11-11T21:24:18Z | 2022-12-01T15:55:24Z | 2022-12-01T15:55:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fboulnois | 16,434,376 | MDQ6VXNlcjE2NDM0Mzc2 | User | false |
huggingface/diffusers | 3,576,999,996 | I_kwDOHa8MBc7VNLA8 | 12,570 | https://github.com/huggingface/diffusers/issues/12570 | https://api.github.com/repos/huggingface/diffusers/issues/12570 | [feature] help us implement unified attention | Currently, we support ring and ulysses attention backends for context parallelism. More info: https://huggingface.co/docs/diffusers/main/en/api/parallel
Unified CP is a generalization of performing Ulysses and Ring together. Both those methods become special subcases of Unified attention. Paper: https://arxiv.org/abs/... | closed | completed | false | 20 | [
"contributions-welcome"
] | [] | 2025-11-01T01:45:29Z | 2026-01-13T03:47:30Z | 2026-01-13T03:47:30Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 3,578,403,434 | I_kwDOHa8MBc7VShpq | 12,572 | https://github.com/huggingface/diffusers/issues/12572 | https://api.github.com/repos/huggingface/diffusers/issues/12572 | Chroma inpainting pipeline | I'd like someone to add an inpaiting pipeline for Chroma
I created a working one here: [see the gist file](https://gist.github.com/rodjjo/1195ea079f395f8e6a81abcc2f13a141)
Working example:
```python
"""
ChromaInpaintPipeline implements a text-guided image inpainting pipeline for the lodestones/Chroma1-HD model,
base... | closed | completed | false | 18 | [] | [] | 2025-11-01T20:18:16Z | 2026-01-15T07:28:51Z | 2026-01-15T07:28:51Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rodjjo | 10,155,053 | MDQ6VXNlcjEwMTU1MDUz | User | false |
huggingface/diffusers | 3,579,092,824 | I_kwDOHa8MBc7VVJ9Y | 12,574 | https://github.com/huggingface/diffusers/issues/12574 | https://api.github.com/repos/huggingface/diffusers/issues/12574 | [Bug] WanVideoToVideoPipeline fails due to missing handling of control arguments required by underlying VACE transformer | ### Describe the bug
Problem:
WanVideoToVideoPipeline.__call__ does not expose control_hidden_states or control_hidden_states_scale in its signature.
However, the underlying TransformerWanVACE.forward() requires them and assumes they’re non-None.
This leads to a crash every time the pipeline is used:
```
pipe(video=m... | closed | completed | false | 11 | [
"bug"
] | [] | 2025-11-02T07:57:03Z | 2026-02-03T14:24:51Z | 2025-11-06T13:13:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Mathias5 | 7,868,427 | MDQ6VXNlcjc4Njg0Mjc= | User | false |
huggingface/diffusers | 3,579,459,482 | I_kwDOHa8MBc7VWjea | 12,575 | https://github.com/huggingface/diffusers/issues/12575 | https://api.github.com/repos/huggingface/diffusers/issues/12575 | [Chroma] Enabling attention backend checks causes error | ### Describe the bug
I am using [OneTrainer](https://github.com/Nerogar/OneTrainer) which uses diffusers to train a LoRA for Chroma. The latest diffusers introduces the concept of a selectable attention backend which can be enabled via the method documented below:
https://huggingface.co/docs/diffusers/main/en/optimiza... | closed | completed | false | 2 | [
"bug",
"contributions-welcome"
] | [] | 2025-11-02T11:37:56Z | 2025-11-12T06:55:29Z | 2025-11-12T06:55:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zzlol63 | 241,569,438 | U_kgDODmYOng | User | false |
huggingface/diffusers | 3,582,275,799 | I_kwDOHa8MBc7VhTDX | 12,580 | https://github.com/huggingface/diffusers/issues/12580 | https://api.github.com/repos/huggingface/diffusers/issues/12580 | Title: ModuleNotFoundError: Could not import module ‘CLIPTextModel’ in diffusers 0.36.0 + transformers 4.57.0 | I’m encountering the following error when using diffusers version 0.36.0 together with transformers version 4.57.0:
File "/<PATH>/diffusers/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py", line 19, in <module>
from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPV... | closed | completed | false | 0 | [] | [] | 2025-11-03T14:15:02Z | 2025-11-04T02:06:20Z | 2025-11-04T02:06:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xxxzzz200 | 107,396,573 | U_kgDOBma93Q | User | false |
huggingface/diffusers | 3,584,883,517 | I_kwDOHa8MBc7VrPs9 | 12,587 | https://github.com/huggingface/diffusers/issues/12587 | https://api.github.com/repos/huggingface/diffusers/issues/12587 | question about the condition usage of `context_pre_only` | Why not use the judgment 'not context_pre_only', as we know 'False' value is not None' in python.
https://github.com/huggingface/diffusers/blob/ac5a1e28fc9cc233863bcfb2abb9eef6807f156f/src/diffusers/models/attention_processor.py#L257
```python
>>> False is None
False
```
| closed | completed | false | 0 | [] | [] | 2025-11-04T06:07:21Z | 2025-11-04T06:13:04Z | 2025-11-04T06:13:04Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DefTruth | 31,974,251 | MDQ6VXNlcjMxOTc0MjUx | User | false |
huggingface/diffusers | 3,585,504,222 | I_kwDOHa8MBc7VtnPe | 12,588 | https://github.com/huggingface/diffusers/issues/12588 | https://api.github.com/repos/huggingface/diffusers/issues/12588 | Allowing other timestep_spacing heuristics for repaint scheduler. Leading in general is a problematic default. | **Is your feature request related to a problem? Please describe.**
I don't like that the repaint scheduler implements the `timestep_spacing` leading method by default, with no way to control it. The "leading" method is problematic for the average user: When the number of train timesteps isn't divisible by the num infer... | open | null | false | 2 | [
"stale"
] | [] | 2025-11-04T09:10:15Z | 2026-01-09T15:06:53Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MarkTension | 16,321,685 | MDQ6VXNlcjE2MzIxNjg1 | User | false |
huggingface/diffusers | 3,585,690,814 | I_kwDOHa8MBc7VuUy- | 12,589 | https://github.com/huggingface/diffusers/issues/12589 | https://api.github.com/repos/huggingface/diffusers/issues/12589 | [feature] implement TeaCache | Currently, we support PAB and FasterCache. More details: https://huggingface.co/docs/diffusers/main/en/optimization/cache.
It'd be cool to support [Timestep Embedding Tells: It's Time to Cache for Video Diffusion Model](https://liewfeng.github.io/TeaCache/) as it's quite popular in the community. Code is here:
https:... | open | null | false | 11 | [
"contributions-welcome",
"performance",
"roadmap"
] | [] | 2025-11-04T09:46:26Z | 2026-03-09T12:08:06Z | null | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 3,592,491,160 | I_kwDOHa8MBc7WIRCY | 12,597 | https://github.com/huggingface/diffusers/issues/12597 | https://api.github.com/repos/huggingface/diffusers/issues/12597 | [i8n-pt] Updating, improving and expanding current PT translation | ## Summary
The Portuguese documentation has only **3 translated files** out of 336 (0.9% coverage) and contains outdated
content and few grammar errors. This issue tracks fixing existing translations and adding critical getting-started
documentation.
**Current status:**
- `index.md` - grammar errors (line... | closed | completed | false | 0 | [] | [] | 2025-11-05T19:27:53Z | 2025-12-12T19:04:30Z | 2025-12-12T19:04:30Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | cdutr | 13,859,415 | MDQ6VXNlcjEzODU5NDE1 | User | false |
huggingface/diffusers | 3,593,024,327 | I_kwDOHa8MBc7WKTNH | 12,599 | https://github.com/huggingface/diffusers/issues/12599 | https://api.github.com/repos/huggingface/diffusers/issues/12599 | Moving to CUDA after loading is extremely slow | ### Describe the bug
Moving a freshly loaded module to CUDA is extremely slow. Much slower than it should. This seems to be related to the way `diffusers` loads weights because cloning the weights before moving them is much faster.
### Reproduction
```python
import time
import torch
from diffusers import StableDiff... | closed | completed | false | 10 | [
"enhancement",
"performance"
] | [] | 2025-11-05T22:19:20Z | 2026-01-15T07:26:46Z | 2026-01-15T07:26:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | francois-rozet | 37,352,336 | MDQ6VXNlcjM3MzUyMzM2 | User | false |
huggingface/diffusers | 1,313,684,472 | I_kwDOHa8MBc5OTTv4 | 126 | https://github.com/huggingface/diffusers/issues/126 | https://api.github.com/repos/huggingface/diffusers/issues/126 | Add a `conda` install option for `diffusers` | It will be helpful to have `diffusers` added `conda-forge` channel and install with conda.
:point_right: I have started the work on it already.
- https://github.com/conda-forge/staged-recipes/pull/19751
Once the PR is merged, the library can be installed as:
```sh
conda install -c conda-forge diffusers
``... | closed | completed | false | 2 | [] | [] | 2022-07-21T19:08:41Z | 2022-08-03T14:46:26Z | 2022-08-03T14:46:26Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sugatoray | 10,201,242 | MDQ6VXNlcjEwMjAxMjQy | User | false |
huggingface/diffusers | 1,446,073,661 | I_kwDOHa8MBc5WMVU9 | 1,260 | https://github.com/huggingface/diffusers/issues/1260 | https://api.github.com/repos/huggingface/diffusers/issues/1260 | Cannot import name 'DPMSolverMultistepScheduler' from 'diffusers' | ### Describe the bug
I'm trying to install diffusers on my new Windows 11 computer, and it's failing on `DPMSolverMultistepScheduler`, which I thought was merged?
### Reproduction
```bash
pip install diffusers[torch] transformers accelerate ftfy
```
and then try
```python
from diffusers import DPMSol... | closed | completed | false | 10 | [
"bug",
"stale"
] | [] | 2022-11-11T22:49:13Z | 2023-01-21T15:03:10Z | 2023-01-21T15:03:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | WASasquatch | 1,151,589 | MDQ6VXNlcjExNTE1ODk= | User | false |
huggingface/diffusers | 3,594,607,534 | I_kwDOHa8MBc7WQVuu | 12,600 | https://github.com/huggingface/diffusers/issues/12600 | https://api.github.com/repos/huggingface/diffusers/issues/12600 | Add KV Cache for Autoregressive Inference | Autoregressive Diffusion Techniques such as [Self Forcing](https://github.com/guandeh17/Self-Forcing/tree/main) rely on a rolling KV Cache across video frame chunks to transfer information from past context frames to the current frames being denoised.
This rolling KV Cache design (or variants similar to it) is likely ... | open | null | false | 15 | [
"stale",
"contributions-welcome",
"advanced",
"consider-for-modular-diffusers",
"modular-diffusers"
] | [
"zucchini-nlp"
] | 2025-11-06T08:03:55Z | 2026-05-06T20:36:05Z | null | MEMBER | null | 20260506T213517Z | 2026-05-06T21:35:17Z | DN6 | 7,529,846 | MDQ6VXNlcjc1Mjk4NDY= | User | false |
huggingface/diffusers | 3,598,847,583 | I_kwDOHa8MBc7Wgg5f | 12,603 | https://github.com/huggingface/diffusers/issues/12603 | https://api.github.com/repos/huggingface/diffusers/issues/12603 | [feature] Add context parallel cp_plan support to many mainstream models | I am working in cache-dit, where I use diffusers as the **native** context parallel backend and extensively utilize custom cp_plan to support context parallel for mainstream models (and has also make many models worked with context parallel + attn_mask via native attn backend). These models include: FLUX, Qwen-Image, L... | open | null | false | 4 | [
"stale"
] | [] | 2025-11-07T05:54:49Z | 2026-02-03T15:16:15Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DefTruth | 31,974,251 | MDQ6VXNlcjMxOTc0MjUx | User | false |
huggingface/diffusers | 3,599,335,576 | I_kwDOHa8MBc7WiYCY | 12,605 | https://github.com/huggingface/diffusers/issues/12605 | https://api.github.com/repos/huggingface/diffusers/issues/12605 | dispatch_attention_fn silently ignores attn_mask for certain backends | ### Describe the bug
When using the [attention backends feature](https://huggingface.co/docs/diffusers/main/en/optimization/attention_backends), it's possible to pass an attention mask to the dispatch_attention_fn to a backend which does not support them (ie. attn_mask is not in the method signature registered to that... | closed | completed | false | 4 | [
"bug",
"contributions-welcome"
] | [] | 2025-11-07T08:54:04Z | 2026-01-06T17:22:14Z | 2026-01-06T17:22:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zzlol63 | 241,569,438 | U_kgDODmYOng | User | false |
huggingface/diffusers | 3,599,666,313 | I_kwDOHa8MBc7WjoyJ | 12,606 | https://github.com/huggingface/diffusers/issues/12606 | https://api.github.com/repos/huggingface/diffusers/issues/12606 | Wrong import of scripts/convert_sd3_to_diffusers.py | ### Describe the bug
Method `load_model_dict_into_meta` should be imported from `diffusers.models.model_loading_utils`.
https://github.com/huggingface/diffusers/blob/9f669e7b5d212f74f3fb1183af178435d785c0b4/scripts/convert_sd3_to_diffusers.py#L10
### Reproduction
Running script reports an error
`ImportError: canno... | closed | completed | false | 2 | [
"bug"
] | [] | 2025-11-07T10:23:46Z | 2025-11-10T09:17:19Z | 2025-11-10T09:17:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Chlorinesu | 148,052,444 | U_kgDOCNMZ3A | User | false |
huggingface/diffusers | 3,599,889,071 | I_kwDOHa8MBc7WkfKv | 12,608 | https://github.com/huggingface/diffusers/issues/12608 | https://api.github.com/repos/huggingface/diffusers/issues/12608 | yudeng970 | test_icedit | closed | completed | false | 1 | [] | [] | 2025-11-07T11:23:10Z | 2025-11-07T14:38:14Z | 2025-11-07T14:38:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hskaskkan | 115,149,037 | U_kgDOBt0I7Q | User | false |
huggingface/diffusers | 3,600,071,013 | I_kwDOHa8MBc7WlLll | 12,609 | https://github.com/huggingface/diffusers/issues/12609 | https://api.github.com/repos/huggingface/diffusers/issues/12609 | wan 2.2 cause nan in latents | ### Describe the bug
here is the source code:
import asyncio, aiohttp.web, math, pathlib, torch, builtins, transformers, diffusers, av, io, numpy, uvloop, concurrent, multiprocessing, os, zipfile, sys
from modelscope.hub.snapshot_download import snapshot_download
from modelscope.hub.file_download import model_file_dow... | closed | completed | false | 1 | [
"bug"
] | [] | 2025-11-07T12:12:28Z | 2025-11-08T14:56:32Z | 2025-11-08T14:56:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chaowenguo | 229,142,208 | U_kgDODahuwA | User | false |
huggingface/diffusers | 1,446,293,415 | I_kwDOHa8MBc5WNK-n | 1,261 | https://github.com/huggingface/diffusers/issues/1261 | https://api.github.com/repos/huggingface/diffusers/issues/1261 | Xformers attention backwards pass not working in diffusers | RuntimeError: p.gQ_strideM() == grad_q.stride(1) INTERNAL ASSERT FAILED at "*/mem_eff_attention/attention_backward_generic.cu":180, | ### Describe the bug
I've been trying to use xformers for training in dreambooth, as well as training with waifu diffusion. It crashes before finishing even 1 iteration, with the message in the logs in both instances.
Xformers appears to be working perfectly fine for inference,
`100%|█████████████████████████████... | closed | completed | false | 1 | [
"bug"
] | [] | 2022-11-12T06:43:53Z | 2022-11-12T21:38:59Z | 2022-11-12T21:38:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | IdiotSandwichTheThird | 84,629,823 | MDQ6VXNlcjg0NjI5ODIz | User | false |
huggingface/diffusers | 3,600,268,407 | I_kwDOHa8MBc7Wl7x3 | 12,610 | https://github.com/huggingface/diffusers/issues/12610 | https://api.github.com/repos/huggingface/diffusers/issues/12610 | Quanto + Group Offload causes device mismatch error (weights on cpu, mat1 on gpu) | ### Describe the bug
When using quanto int8 and Diffusers group offload, we're seeing device mismatch on forward pass.
### Reproduction
```py
"""
This demonstrates the core issue: when .to(device) is called on WeightQBytesTensor,
the wrapper reports the new device but internal components may not be moved properly
wh... | open | null | false | 5 | [
"bug",
"stale"
] | [] | 2025-11-07T13:07:13Z | 2026-02-03T15:16:12Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 3,603,003,422 | I_kwDOHa8MBc7WwXge | 12,612 | https://github.com/huggingface/diffusers/issues/12612 | https://api.github.com/repos/huggingface/diffusers/issues/12612 | diffusers 0.35.2 wan 2.1 cause nan in latents | ### Describe the bug
import asyncio, aiohttp.web, math, pathlib, torch, builtins, transformers, diffusers, av, io, numpy, uvloop, concurrent, multiprocessing, os, zipfile, sys
from modelscope.hub.snapshot_download import snapshot_download
from modelscope.hub.file_download import model_file_download
#import sd_embed.em... | closed | completed | false | 2 | [
"bug"
] | [] | 2025-11-08T07:05:25Z | 2025-11-14T04:39:15Z | 2025-11-08T14:52:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chaowenguo | 229,142,208 | U_kgDODahuwA | User | false |
huggingface/diffusers | 3,603,474,351 | I_kwDOHa8MBc7WyKev | 12,613 | https://github.com/huggingface/diffusers/issues/12613 | https://api.github.com/repos/huggingface/diffusers/issues/12613 | wan 2.2 cause nan in latent in i2v | ### Describe the bug
import asyncio, aiohttp.web, math, pathlib, torch, builtins, transformers, diffusers, PIL, av, io, numpy, uvloop, concurrent, multiprocessing, os, zipfile, sys
from modelscope.hub.snapshot_download import snapshot_download
from modelscope.hub.file_download import model_file_download
#import sd_emb... | open | null | false | 28 | [
"bug",
"stale"
] | [] | 2025-11-08T11:57:35Z | 2026-02-03T15:16:09Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chaowenguo | 229,142,208 | U_kgDODahuwA | User | false |
huggingface/diffusers | 1,446,332,148 | I_kwDOHa8MBc5WNUb0 | 1,262 | https://github.com/huggingface/diffusers/issues/1262 | https://api.github.com/repos/huggingface/diffusers/issues/1262 | train a diffusion model on rectangle images | ### Describe the bug
How to train the pictures with Rectangle ,I tried 128*128, and it is fine,but 128*256 input then there are error of tuple size, does anyone know how to solve it?
I just use the diffusers_training_example on colab
![image](https://user-images.githubusercontent.com/56218499/201465424-9bbbbec7-12be... | closed | completed | false | 1 | [
"bug"
] | [] | 2022-11-12T08:26:34Z | 2022-11-23T10:11:41Z | 2022-11-23T10:11:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Fairywang9 | 56,218,499 | MDQ6VXNlcjU2MjE4NDk5 | User | false |
huggingface/diffusers | 3,606,675,566 | I_kwDOHa8MBc7W-YBu | 12,620 | https://github.com/huggingface/diffusers/issues/12620 | https://api.github.com/repos/huggingface/diffusers/issues/12620 | Missing files in release assets | ### Describe the bug
I installed this library of version 0.35.2 with pip.
And I found that multiple files are missing compared to github.
Missing files I found:
src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet.py
src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet_inpaint.py
src/diffusers/p... | closed | completed | false | 2 | [
"bug"
] | [] | 2025-11-10T08:06:14Z | 2025-11-17T05:20:08Z | 2025-11-17T05:20:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kwonmha | 8,953,934 | MDQ6VXNlcjg5NTM5MzQ= | User | false |
huggingface/diffusers | 3,608,858,695 | I_kwDOHa8MBc7XGtBH | 12,624 | https://github.com/huggingface/diffusers/issues/12624 | https://api.github.com/repos/huggingface/diffusers/issues/12624 | please add strength to wanimagetovideopipeline | **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 [...].
i use wanimagetovideopipeline and works fine. now i switch to lightx2v lora and use 4steps. the video just horner the original image and seems ignore the promp... | open | null | false | 1 | [
"stale"
] | [] | 2025-11-10T16:47:16Z | 2026-01-09T15:06:26Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chaowenguo | 229,142,208 | U_kgDODahuwA | User | false |
huggingface/diffusers | 1,446,435,367 | I_kwDOHa8MBc5WNton | 1,263 | https://github.com/huggingface/diffusers/issues/1263 | https://api.github.com/repos/huggingface/diffusers/issues/1263 | cannot import name 'StableDiffusionSafetyChecker' from 'diffusers.pipelines.stable_diffusion | ### Describe the bug
Whenever you run the convert_original_stable_diffusion_to_diffusers.py in google colab it results in the error
cannot import name 'StableDiffusionSafetyChecker' from 'diffusers.pipelines.stable_diffusion
### Reproduction
run the code on google colab
### Logs
```shell
------------------------... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2022-11-12T12:38:22Z | 2022-12-24T15:03:20Z | 2022-12-24T15:03:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | The-Ramosian | 118,053,996 | U_kgDOBwlcbA | User | false |
huggingface/diffusers | 3,611,065,812 | I_kwDOHa8MBc7XPH3U | 12,630 | https://github.com/huggingface/diffusers/issues/12630 | https://api.github.com/repos/huggingface/diffusers/issues/12630 | AttributeError: 'GPT2Model' object has no attribute '_get_initial_cache_position' in AudioLDM2Pipeline | ### Describe the bug
```python3
! python3 -m pip install -U phonemizer
! sudo apt install -y --no-install-recommends espeak-ng
import scipy, transformers
import torch
from diffusers import AudioLDM2Pipeline
repo_id = "anhnct/audioldm2_gigaspeech"
pipe = AudioLDM2Pipeline.from_pretrained(repo_id, torch_dtype=torch.flo... | open | null | false | 3 | [
"bug"
] | [] | 2025-11-11T07:31:16Z | 2026-01-21T19:29:13Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chaowenguo | 229,142,208 | U_kgDODahuwA | User | false |
huggingface/diffusers | 3,611,147,196 | I_kwDOHa8MBc7XPbu8 | 12,632 | https://github.com/huggingface/diffusers/issues/12632 | https://api.github.com/repos/huggingface/diffusers/issues/12632 | stable_diffusion_xl_img2img failed due to dtype missmatch | ### Describe the bug
The official usage of [stabilityai/stable-diffusion-xl-refiner-1.0](https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0) :
```python
import torch
from diffusers import StableDiffusionXLImg2ImgPipeline
from diffusers.utils import load_image
pipe = StableDiffusionXLImg2ImgPipeline.fr... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2025-11-11T07:57:17Z | 2026-01-10T01:39:42Z | 2026-01-10T01:39:42Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiqing-feng | 107,918,818 | U_kgDOBm614g | User | false |
huggingface/diffusers | 3,611,179,156 | I_kwDOHa8MBc7XPjiU | 12,633 | https://github.com/huggingface/diffusers/issues/12633 | https://api.github.com/repos/huggingface/diffusers/issues/12633 | DDIMScheduler.step() logic mismatches set_timesteps() list when using timestep_spacing='linspace' | ### Describe the bug
The step() function calculates prev_timestep using a hard-coded formula:
`prev_timestep = timestep - self.config.num_train_timesteps // self.num_inference_steps.`
https://github.com/huggingface/diffusers/blob/main/src/diffusers/schedulers/scheduling_ddim.py#L402
However, if the scheduler is init... | open | null | false | 1 | [
"bug",
"stale"
] | [] | 2025-11-11T08:08:45Z | 2026-01-09T15:06:09Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | cbprfCsun | 145,126,978 | U_kgDOCKZ2Qg | User | false |
huggingface/diffusers | 3,611,548,955 | I_kwDOHa8MBc7XQ90b | 12,635 | https://github.com/huggingface/diffusers/issues/12635 | https://api.github.com/repos/huggingface/diffusers/issues/12635 | The Diffusers MVP 🚀 | Hello folks 👋
## What❓
We’re excited to bring a contributor-focused program to you! In this program, we want to work and collaborate with serious contributors to Diffusers and reward them for their time and energy spent with us. Keep reading this thread if that sounds interesting to you.
To ease the process, we hav... | open | null | false | 19 | [
"diffusers-mvp"
] | [
"DN6",
"yiyixuxu",
"sayakpaul"
] | 2025-11-11T09:54:56Z | 2026-06-13T04:55:36Z | null | MEMBER | null | 20260613T053522Z | 2026-06-13T05:35:22Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 3,614,601,441 | I_kwDOHa8MBc7XcnDh | 12,638 | https://github.com/huggingface/diffusers/issues/12638 | https://api.github.com/repos/huggingface/diffusers/issues/12638 | How to design network with DiT blocks that are friendly to Tensorrt fp16 conversion? | We had a network that structed as `a convnet pre-encoder -> DiT blocks -> final block for last sampling`, it worked well with torch format and onnx format, but when we tried to convert it into tensorrt fp16 format, the inference will get value overflow. we had seen the data differene [between onnx and trt fp16, wit... | closed | completed | true | 1 | [
"stale"
] | [] | 2025-11-12T02:23:37Z | 2026-01-10T01:40:48Z | 2026-01-10T01:40:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JohnHerry | 8,011,802 | MDQ6VXNlcjgwMTE4MDI= | User | false |
huggingface/diffusers | 1,446,527,898 | I_kwDOHa8MBc5WOEOa | 1,264 | https://github.com/huggingface/diffusers/issues/1264 | https://api.github.com/repos/huggingface/diffusers/issues/1264 | Why is noise scheduler in train_dreambooth.py hard coded to "CompVis/stable-diffusion-v1-4" ? | I noticed that the noise scheduler in the train_dreambooth.py is hardcoded to "CompVis/stable-diffusion-v1-4" here
https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth.py#L475
The script also takes a command line parameter for pretrained_model https://github.com/huggingface/diffus... | closed | completed | false | 1 | [] | [] | 2022-11-12T17:06:35Z | 2022-11-17T23:31:19Z | 2022-11-17T23:31:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sausax | 481,150 | MDQ6VXNlcjQ4MTE1MA== | User | false |
huggingface/diffusers | 3,615,376,593 | I_kwDOHa8MBc7XfkTR | 12,641 | https://github.com/huggingface/diffusers/issues/12641 | https://api.github.com/repos/huggingface/diffusers/issues/12641 | Potential incorrect indentation for logging in train_dreambooth.py | **Description:**
In the `train_dreambooth.py` script, the logging and progress bar updates appear to be executed on every training step, even when using gradient accumulation. This might lead to incorrect or redundant logging.
The relevant code is located around lines 1393-1395:
https://github.com/huggingface/diffus... | open | null | false | 1 | [
"stale"
] | [] | 2025-11-12T07:45:28Z | 2026-01-09T15:06:01Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Men1scus | 115,412,181 | U_kgDOBuEM1Q | User | false |
huggingface/diffusers | 3,619,154,514 | I_kwDOHa8MBc7Xt-pS | 12,645 | https://github.com/huggingface/diffusers/issues/12645 | https://api.github.com/repos/huggingface/diffusers/issues/12645 | Why Qwen-Image inference speed may slower than comfyui | ### Describe the bug
I have tested diffusers and Comfyui with the same parameters and check the input shapes
The parameters is
<img width="788" height="462" alt="Image" src="https://github.com/user-attachments/assets/3cb6630e-2e2a-4bc4-a74f-34a2b8bff1c5" />
<img width="1352" height="716" alt="Image" src="https://git... | open | null | false | 17 | [
"bug",
"stale"
] | [] | 2025-11-13T03:02:38Z | 2026-01-09T15:05:54Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hanggun | 39,485,441 | MDQ6VXNlcjM5NDg1NDQx | User | false |
huggingface/diffusers | 3,619,391,782 | I_kwDOHa8MBc7Xu4km | 12,646 | https://github.com/huggingface/diffusers/issues/12646 | https://api.github.com/repos/huggingface/diffusers/issues/12646 | 【 EasyAnimate 】 pipine_ easyanimate_inpaint.Py, line 1191: image_1atents used without assigning a value, causing the program to crash | ### Describe the bug
When using EasyAnimate's image restoration function, in some cases it may crash directly with an error message: UnboundLocalError: local variable 'image_latents' referenced before assignment
Problem cause:
The variable image_1atents is only assigned when "using precomputed latents (use_recomputed... | open | null | false | 3 | [
"bug",
"stale"
] | [] | 2025-11-13T04:37:36Z | 2026-06-13T11:16:06Z | null | NONE | null | 20260613T113521Z | 2026-06-13T11:35:21Z | KittenHaerin | 177,937,460 | U_kgDOCpscNA | User | false |
huggingface/diffusers | 1,446,610,520 | I_kwDOHa8MBc5WOYZY | 1,265 | https://github.com/huggingface/diffusers/issues/1265 | https://api.github.com/repos/huggingface/diffusers/issues/1265 | How to convert Flax model to .ckpt ? | null | closed | completed | false | 7 | [
"stale"
] | [] | 2022-11-12T21:24:13Z | 2023-01-23T15:04:10Z | 2023-01-23T15:04:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sneccc | 61,990,796 | MDQ6VXNlcjYxOTkwNzk2 | User | false |
huggingface/diffusers | 3,620,894,801 | I_kwDOHa8MBc7X0nhR | 12,650 | https://github.com/huggingface/diffusers/issues/12650 | https://api.github.com/repos/huggingface/diffusers/issues/12650 | Question about the `# Copied from` system | Hi team! 👋
While working on improving docstrings and type hints across scheduler files (issue #9567), I've noticed the `# Copied from` pattern used extensively throughout the codebase.
Examples:
- Functions like `betas_for_alpha_bar` are duplicated across multiple schedulers
- Output classes like `DDPMSchedulerOutpu... | open | null | false | 11 | [
"stale"
] | [] | 2025-11-13T11:53:22Z | 2026-02-12T15:17:45Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | delmalih | 26,625,111 | MDQ6VXNlcjI2NjI1MTEx | User | false |
huggingface/diffusers | 3,622,340,694 | I_kwDOHa8MBc7X6IhW | 12,653 | https://github.com/huggingface/diffusers/issues/12653 | https://api.github.com/repos/huggingface/diffusers/issues/12653 | PRX transformers version of rope has a float64 reference, breaks MPS (and npu ?) | ### Describe the bug
Once again there's a ROPE implementation with torch.float64 references which breaks when run against MPS and I assume nps as the last time I reported this it was fixed for both. This time it is in the PRX transformers code,
```
File "/Volumes/SSD2TB/AI/Diffusers/lib/python3.11/site-packages/dif... | closed | completed | false | 2 | [
"bug"
] | [] | 2025-11-13T18:34:24Z | 2025-11-15T15:14:36Z | 2025-11-15T15:14:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Vargol | 62,868 | MDQ6VXNlcjYyODY4 | User | false |
huggingface/diffusers | 3,623,795,991 | I_kwDOHa8MBc7X_r0X | 12,656 | https://github.com/huggingface/diffusers/issues/12656 | https://api.github.com/repos/huggingface/diffusers/issues/12656 | [Bug] Positional Embedding Index Out-of-Bounds in AuraFlow | ### Describe the bug
When training the AuraFlow (Pony V7) lora model (via SimpleTuner), the positional embedding selection function
`pe_selection_index_based_on_dim()` can generate **negative** or **out-of-range** indices for the positional embedding table.
This causes:
```
CUDA error: device-side assert triggered
v... | open | null | false | 8 | [
"bug",
"stale"
] | [] | 2025-11-14T04:00:41Z | 2026-02-03T15:15:50Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ikuto32 | 47,182,163 | MDQ6VXNlcjQ3MTgyMTYz | User | false |
huggingface/diffusers | 3,623,979,510 | I_kwDOHa8MBc7YAYn2 | 12,657 | https://github.com/huggingface/diffusers/issues/12657 | https://api.github.com/repos/huggingface/diffusers/issues/12657 | Qwen Image Edit Quantized Stopped working | ### Describe the bug
i have been using the qwen edit for a little while now an all of a sudden its crashing when i spin up my same docker conatiner on a new machine.
all with main branch of Diffusers from GitHub pulling fresh on spin up.
ive done hunreds of images with the same docker container that installs the lib... | closed | completed | false | 6 | [
"bug"
] | [] | 2025-11-14T05:42:47Z | 2025-11-16T23:09:36Z | 2025-11-16T23:09:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AlpineVibrations | 10,145,679 | MDQ6VXNlcjEwMTQ1Njc5 | User | false |
huggingface/diffusers | 1,446,644,009 | I_kwDOHa8MBc5WOgkp | 1,266 | https://github.com/huggingface/diffusers/issues/1266 | https://api.github.com/repos/huggingface/diffusers/issues/1266 | Unable to run prompt | ### Describe the bug
Error message when running prompt
NameError Traceback (most recent call last)
[<ipython-input-45-c30b32863e43>](https://localhost:8080/#) in <module>
18 prompt_ids = pipeline.prepare_inputs(prompt)
19 prompt_ids = shard(prompt_ids)
---> 20 images = ... | closed | completed | false | 1 | [
"bug"
] | [] | 2022-11-12T23:20:27Z | 2022-11-13T12:25:29Z | 2022-11-13T12:25:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | G-force78 | 114,336,644 | U_kgDOBtCjhA | User | false |
huggingface/diffusers | 3,624,568,826 | I_kwDOHa8MBc7YCof6 | 12,661 | https://github.com/huggingface/diffusers/issues/12661 | https://api.github.com/repos/huggingface/diffusers/issues/12661 | Crash while using ChronoEdit with context parallelism | ### Describe the bug
Crash while using ChronoEdit with context parallelism
### Reproduction
- test script
```python
import os
import time
import torch
import numpy as np
from PIL import Image
import torch.distributed as dist
from diffusers import (
AutoencoderKLWan,
ChronoEditTransformer3DModel,
Chrono... | closed | completed | false | 0 | [
"bug"
] | [] | 2025-11-14T08:32:56Z | 2025-11-24T11:06:54Z | 2025-11-24T11:06:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DefTruth | 31,974,251 | MDQ6VXNlcjMxOTc0MjUx | User | false |
huggingface/diffusers | 3,624,758,333 | I_kwDOHa8MBc7YDWw9 | 12,662 | https://github.com/huggingface/diffusers/issues/12662 | https://api.github.com/repos/huggingface/diffusers/issues/12662 | question on stable_audio_transformer.py | Execuse me, I am leaning the code of `class StableAudioDiTModel` , I do not know what is the argument ` global_states_input_dim` used to? It seems that it is a must component that should be packed before the hidden_states sequence. and its default dim seems larger then the transformer inner_dim. What is that... | closed | completed | true | 2 | [
"stale"
] | [] | 2025-11-14T09:26:01Z | 2026-01-10T01:44:04Z | 2026-01-10T01:44:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JohnHerry | 8,011,802 | MDQ6VXNlcjgwMTE4MDI= | User | false |
huggingface/diffusers | 3,624,810,337 | I_kwDOHa8MBc7YDjdh | 12,663 | https://github.com/huggingface/diffusers/issues/12663 | https://api.github.com/repos/huggingface/diffusers/issues/12663 | huhu | hhh | closed | completed | false | 0 | [] | [] | 2025-11-14T09:40:37Z | 2025-11-17T07:20:14Z | 2025-11-17T07:20:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JQueenKing | 139,109,850 | U_kgDOCEql2g | User | false |
huggingface/diffusers | 3,626,430,687 | I_kwDOHa8MBc7YJvDf | 12,666 | https://github.com/huggingface/diffusers/issues/12666 | https://api.github.com/repos/huggingface/diffusers/issues/12666 | Kandinsky 5.0 Video Pipeline not in Diffusers 0.35.2 | ### Describe the bug
Kandinsky 5.0 Video Pipeline not present in Diffusers 0.35.2, although on GitHub
### Reproduction
!pip install diffusers==0.35.2
import torch
from diffusers import Kandinsky5T2VPipeline
@yiyixuxu @sayakpaul
### Logs
```shell
```
### System Info
.
### Who can help?
_No response_ | closed | completed | false | 3 | [
"bug"
] | [] | 2025-11-14T17:01:39Z | 2025-11-17T10:44:40Z | 2025-11-17T10:44:39Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | leffff | 57,654,885 | MDQ6VXNlcjU3NjU0ODg1 | User | false |
huggingface/diffusers | 3,628,800,676 | I_kwDOHa8MBc7YSxqk | 12,667 | https://github.com/huggingface/diffusers/issues/12667 | https://api.github.com/repos/huggingface/diffusers/issues/12667 | Any Wan version crashes | ### Describe the bug
Running any diffusers Wan example code crashes for me, when running from inside Blender python (everything else in Diffusers has run fine from Blender Python over the years).
Ex. code:
```python
# pip install ftfy
import torch
import numpy as np
from diffusers import AutoModel, WanPipeline
from... | open | null | false | 14 | [
"bug",
"stale"
] | [] | 2025-11-15T14:47:50Z | 2026-01-09T15:05:40Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 3,628,964,089 | I_kwDOHa8MBc7YTZj5 | 12,668 | https://github.com/huggingface/diffusers/issues/12668 | https://api.github.com/repos/huggingface/diffusers/issues/12668 | Complex numbers in Qwen/Z-Image Image pipeline incompatible with torch.compile | ### Describe the bug
Note: This might be something for the MVP program https://github.com/huggingface/diffusers/issues/12635 if there's anyone who already has a deep understanding of rotary embeddings and complex numbers. I don't.
The Qwen image pipeline calls https://github.com/huggingface/diffusers/blob/01a56927f16... | open | null | false | 11 | [
"bug",
"stale"
] | [] | 2025-11-15T17:08:58Z | 2026-01-09T15:05:37Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dxqb | 183,307,934 | U_kgDOCu0Ong | User | false |
huggingface/diffusers | 3,630,447,311 | I_kwDOHa8MBc7YZDrP | 12,669 | https://github.com/huggingface/diffusers/issues/12669 | https://api.github.com/repos/huggingface/diffusers/issues/12669 | Flux1-Dev inference with single file ComfyUI/SD-Forge Safetensors | Is it possible to run inference with diffusers using a single-file safetensors created for ComfyUI/SD-Forge?
It looks like FluxPipeline.from_single_file() might be intended for this purpose, but I'm getting the following errors:
```
import torch
from diffusers import FluxPipeline
pipe = FluxPipeline.from_single_fil... | open | null | false | 14 | [
"stale",
"wan"
] | [] | 2025-11-16T11:57:48Z | 2026-02-03T15:15:45Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ddpasa | 112,642,920 | U_kgDOBrbLaA | User | false |
huggingface/diffusers | 1,446,793,669 | I_kwDOHa8MBc5WPFHF | 1,267 | https://github.com/huggingface/diffusers/issues/1267 | https://api.github.com/repos/huggingface/diffusers/issues/1267 | Schedulers do not work well for text-guided image-to-image generation | For https://lablab.ai/t/stable-diffusion-img2img , it seems that only [LMSDiscreteScheduler](https://huggingface.co/docs/diffusers/api/schedulers#diffusers.LMSDiscreteScheduler) works well in [generating different variations of the pictures sketches](https://colab.research.google.com/drive/1LMDWu5TUUGjjTnis7wkYP1TY77Dx... | closed | completed | false | 4 | [
"stale"
] | [] | 2022-11-13T08:01:42Z | 2022-12-24T15:03:18Z | 2022-12-24T15:03:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | buttercutter | 3,324,659 | MDQ6VXNlcjMzMjQ2NTk= | User | false |
huggingface/diffusers | 3,631,602,135 | I_kwDOHa8MBc7YddnX | 12,672 | https://github.com/huggingface/diffusers/issues/12672 | https://api.github.com/repos/huggingface/diffusers/issues/12672 | How to set pipe "requires_grad=true"？ | I have set the variable and the model "requires_grad=true" with the following:
` pipe.transformer.requires_grad = True
pipe.vae.requires_grad = True`
`prev_sample = prev_sample.detach().requires_grad_(True)`
but the "requires_grad" of result by the pipe is still not true:
`image_tar = pipe.vae.decode(prev_sampl... | closed | completed | false | 2 | [] | [] | 2025-11-17T03:36:43Z | 2025-11-20T12:19:20Z | 2025-11-20T12:19:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | micklexqg | 13,776,012 | MDQ6VXNlcjEzNzc2MDEy | User | false |
huggingface/diffusers | 3,631,909,963 | I_kwDOHa8MBc7YeoxL | 12,673 | https://github.com/huggingface/diffusers/issues/12673 | https://api.github.com/repos/huggingface/diffusers/issues/12673 | QwenImagepipeline imcompatable with group offloading with record_stream=True | ### Describe the bug
I use group offloading with QwenImagepipeline, it shows a lot of blocks are not used in forward. And the final image is a mess. I try enable_sequential_cpu_offload is working fine. I believe it is the error from group offloading.
### Reproduction
```python3
from diffusers import DiffusionPipelin... | closed | completed | false | 7 | [
"bug",
"group-offloading"
] | [] | 2025-11-17T06:18:08Z | 2025-12-03T06:07:12Z | 2025-12-03T06:07:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chaowenguo | 229,142,208 | U_kgDODahuwA | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.