repo stringclasses 1
value | github_id int64 1.27B 4.67B | github_node_id stringlengths 18 24 | number int64 8 14k | 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 2026-06-15 20:11:23 | updated_at stringdate 2022-06-12 22:18:01 2026-06-16 23:50:48 | closed_at stringdate 2022-06-12 22:18:01 2026-06-16 00:53:57 ⌀ | author_association stringclasses 3
values | milestone_title stringclasses 0
values | snapshot_id stringclasses 80
values | extracted_at stringdate 2026-04-07 13:34:13 2026-06-17 00:35:21 | author_login stringlengths 3 28 | author_id int64 1.54k 290M | author_node_id stringlengths 12 20 | author_type stringclasses 3
values | author_site_admin bool 1
class |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
huggingface/diffusers | 2,364,097,773 | I_kwDOHa8MBc6M6UTt | 8,647 | https://github.com/huggingface/diffusers/issues/8647 | https://api.github.com/repos/huggingface/diffusers/issues/8647 | StableDiffusionControlNetImg2ImgPipeline call report “argument of type 'NoneType' is not iterable” | ### Describe the bug
argument of type 'NoneType' is not iterable
checkpoint = os.path.join(hub_dir, "checkpoints/StableDiffusionXL/model/ud_sdxl-动漫二次元.safetensors")
pipe = StableDiffusionControlNetImg2ImgPipeline.from_single_file(
checkpoint,
controlnet=controlnet,
torch_dtype=torch... | open | null | false | 2 | [
"bug",
"stale"
] | [] | 2024-06-20T10:30:22Z | 2024-09-14T15:08:38Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hjj-lmx | 107,909,464 | U_kgDOBm6RWA | User | false |
huggingface/diffusers | 2,364,207,370 | I_kwDOHa8MBc6M6vEK | 8,649 | https://github.com/huggingface/diffusers/issues/8649 | https://api.github.com/repos/huggingface/diffusers/issues/8649 | SD3 - num_images_per_prompt no longer honoured (throws error) | ### Describe the bug
With models prior to SD3, the parameter num_images_per_prompt is honoured, enabling generation of several images per prompt. With sd3-medium an error is generated.
RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 2 but got size 1 for tensor number 1 in the list.
Not... | closed | completed | false | 4 | [
"bug"
] | [] | 2024-06-20T11:28:22Z | 2024-06-29T13:05:28Z | 2024-06-29T13:05:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zagglez | 61,600,248 | MDQ6VXNlcjYxNjAwMjQ4 | User | false |
huggingface/diffusers | 1,411,082,994 | I_kwDOHa8MBc5UG2ry | 865 | https://github.com/huggingface/diffusers/issues/865 | https://api.github.com/repos/huggingface/diffusers/issues/865 | A Generalist Framework for Panoptic Segmentation of Images and Videos | ### Model/Pipeline/Scheduler description
This work (https://arxiv.org/pdf/2210.06366.pdf) presents how we can apply the advances of diffusion modelling to generate panoptic masks for images and videos conditioned on any image input. In most works related to diffusion modelling, the noise and output space is parametr... | closed | completed | false | 3 | [
"New pipeline/model"
] | [] | 2022-10-17T07:50:41Z | 2026-01-09T18:03:20Z | 2026-01-09T18:03:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | 345ishaan | 7,318,028 | MDQ6VXNlcjczMTgwMjg= | User | false |
huggingface/diffusers | 2,364,607,655 | I_kwDOHa8MBc6M8Qyn | 8,651 | https://github.com/huggingface/diffusers/issues/8651 | https://api.github.com/repos/huggingface/diffusers/issues/8651 | configuration_utils.to_json_string() fails on WindowsPath | ### Describe the bug
Running the diffusion-models-class/unit2/finetune_model.py line 117 (save_pretrained) and 01_finetuning_and_guidance.ipynb crashes trying to convert a WindowsPath to JSON. This worked prior to an conda update (not sure which module update caused the problem.) I have diffusers 0.29.0 installed. Iss... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-06-20T14:31:19Z | 2024-06-27T05:00:56Z | 2024-06-27T05:00:56Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vincedovy | 6,002,260 | MDQ6VXNlcjYwMDIyNjA= | User | false |
huggingface/diffusers | 2,365,653,734 | I_kwDOHa8MBc6NAQLm | 8,655 | https://github.com/huggingface/diffusers/issues/8655 | https://api.github.com/repos/huggingface/diffusers/issues/8655 | i2vgen-xl keep produce black gif | ```py
import torch
from diffusers import I2VGenXLPipeline
from diffusers.utils import export_to_gif, load_image
pipeline = I2VGenXLPipeline.from_pretrained(
"ali-vilab/i2vgen-xl", torch_dtype=torch.float16
)
pipeline.enable_model_cpu_offload()
image_url = (
"https://huggingface.co/datasets/diffus... | closed | completed | false | 7 | [] | [] | 2024-06-21T03:46:18Z | 2024-07-17T08:35:34Z | 2024-06-29T13:04:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xalteropsx | 103,671,642 | U_kgDOBi3nWg | User | false |
huggingface/diffusers | 2,366,421,935 | I_kwDOHa8MBc6NDLuv | 8,659 | https://github.com/huggingface/diffusers/issues/8659 | https://api.github.com/repos/huggingface/diffusers/issues/8659 | train_text_to_image_sdxl.py fail resume from checkpoint and also can not load for infer | ### Describe the bug
I try to finetune a SDXL model. But meet same questions.
I can not resume from a checkpoint model and the error is shown as follows:
[rank0]: load_checkpoint_in_model(
[rank0]: File "/mnt/wangxuekuan/miniconda3/envs/sdxl/lib/python3.8/site-packages/accelerate/utils/modeling.py", line ... | open | null | false | 4 | [
"bug",
"stale"
] | [] | 2024-06-21T12:15:29Z | 2024-12-13T15:05:48Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | XuekuanWang | 20,875,052 | MDQ6VXNlcjIwODc1MDUy | User | false |
huggingface/diffusers | 1,411,085,348 | I_kwDOHa8MBc5UG3Qk | 866 | https://github.com/huggingface/diffusers/issues/866 | https://api.github.com/repos/huggingface/diffusers/issues/866 | The behavior of safety_checker is ambiguous. | I need information about which embeddings the safety_checker's concept_embeds have.
The 3 vectors in concept_embeds feel like they're not enough to work as a filtering.
I think the current safety_checker isn't enough to isolate the brutal or gore images.
How about adding an explicit runtime computed embedding for ... | closed | completed | false | 3 | [
"stale",
"Safety Checker"
] | [] | 2022-10-17T07:52:44Z | 2023-12-04T19:34:36Z | 2022-11-18T15:02:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | klae01 | 101,966,044 | U_kgDOBhPg3A | User | false |
huggingface/diffusers | 2,367,576,605 | I_kwDOHa8MBc6NHlod | 8,663 | https://github.com/huggingface/diffusers/issues/8663 | https://api.github.com/repos/huggingface/diffusers/issues/8663 | AnimateDiff bug not sure if it use adapter or not | s import Transformer2DModelOutput`, instead.
deprecate("Transformer2DModelOutput", "1.0.0", deprecation_message) ... | closed | completed | false | 3 | [] | [] | 2024-06-22T05:21:55Z | 2024-06-29T13:09:23Z | 2024-06-29T13:09:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xalteropsx | 103,671,642 | U_kgDOBi3nWg | User | false |
huggingface/diffusers | 2,367,720,580 | I_kwDOHa8MBc6NIIyE | 8,664 | https://github.com/huggingface/diffusers/issues/8664 | https://api.github.com/repos/huggingface/diffusers/issues/8664 | AnimateDiffSDXL + Multi Controlnets support | **Is your feature request related to a problem? Please describe.**
The current AnimateDiffSDXLPipeline doesn't support neither 1 controlnet nor multi controlnets.
I've been working on this task for several days by combining StableDiffusionXLControlNetAdapterPipeline and AnimateDiffControlNetPipeline in community fold... | open | null | false | 7 | [
"stale",
"contributions-welcome",
"consider-for-modular-diffusers"
] | [] | 2024-06-22T09:50:32Z | 2024-12-13T15:05:44Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Sundragon1993 | 9,301,684 | MDQ6VXNlcjkzMDE2ODQ= | User | false |
huggingface/diffusers | 2,368,300,028 | I_kwDOHa8MBc6NKWP8 | 8,665 | https://github.com/huggingface/diffusers/issues/8665 | https://api.github.com/repos/huggingface/diffusers/issues/8665 | need a download folder button | https://huggingface.co/runwayml/stable-diffusion-v1-5/tree/main
the problem with git clone
it download the file 4time size of original size and disk usage is also spammed
please consider provide an single download button on website itself
need to avoid uses of git cli | closed | completed | false | 7 | [] | [] | 2024-06-23T06:05:34Z | 2024-06-30T11:54:15Z | 2024-06-23T10:05:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xalteropsx | 103,671,642 | U_kgDOBi3nWg | User | false |
huggingface/diffusers | 2,368,377,122 | I_kwDOHa8MBc6NKpEi | 8,666 | https://github.com/huggingface/diffusers/issues/8666 | https://api.github.com/repos/huggingface/diffusers/issues/8666 | Attention api changes no documentation ? | how can i see ur previous changes on attention ?
u have rename`` _slice_size , _sliced_attention and _attention`` attribute from attention
need to know what are alternative using of its ? | closed | completed | false | 4 | [] | [] | 2024-06-23T07:08:58Z | 2024-06-23T11:31:47Z | 2024-06-23T11:31:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xalteropsx | 103,671,642 | U_kgDOBi3nWg | User | false |
huggingface/diffusers | 2,368,848,902 | I_kwDOHa8MBc6NMcQG | 8,667 | https://github.com/huggingface/diffusers/issues/8667 | https://api.github.com/repos/huggingface/diffusers/issues/8667 | Tensor size mismatch for non pow of 2 sized image, SD3ControlNetModel | ### Describe the bug
There seems to be an issue with certain non power of 2 sized control net guidance images when using SD3ControlNetModel
### Reproduction
```
import diffusers
import PIL.Image
import os
import torch
os.environ['HF_TOKEN'] = 'your token'
cn = diffusers.SD3ControlNetModel.from_pretrained... | closed | completed | false | 12 | [
"bug"
] | [] | 2024-06-23T19:45:25Z | 2024-07-04T03:53:25Z | 2024-07-04T03:53:25Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Teriks | 14,919,098 | MDQ6VXNlcjE0OTE5MDk4 | User | false |
huggingface/diffusers | 2,368,867,724 | I_kwDOHa8MBc6NMg2M | 8,668 | https://github.com/huggingface/diffusers/issues/8668 | https://api.github.com/repos/huggingface/diffusers/issues/8668 | StableDiffusion3Img2ImgPipeline fails when input image is not aligned to 16 | ### Describe the bug
The pipeline fails if the input image is not aligned to 16, this is not mentioned in documentation.
### Reproduction
```
import diffusers
import PIL.Image
import os
os.environ['HF_TOKEN'] = 'your token'
pipe = diffusers.StableDiffusion3Img2ImgPipeline.from_pretrained(
'stabilityai/... | closed | completed | false | 1 | [
"bug"
] | [] | 2024-06-23T19:59:13Z | 2024-06-24T08:18:03Z | 2024-06-24T08:18:03Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Teriks | 14,919,098 | MDQ6VXNlcjE0OTE5MDk4 | User | false |
huggingface/diffusers | 1,411,089,073 | I_kwDOHa8MBc5UG4Kx | 867 | https://github.com/huggingface/diffusers/issues/867 | https://api.github.com/repos/huggingface/diffusers/issues/867 | Running Stable Diffusion in FastAPI Does Not Release GPU Memory | ### Describe the bug
I am running Stable Diffusionas as a web service using FastAPI. It runs fine, but after doing multiple inference calls, I noticed the memory of the GPU becomes full and the inference fails. It is as if the memory is not released right after doing the inference.
### Reproduction
**requi... | closed | completed | false | 13 | [
"bug",
"stale"
] | [] | 2022-10-17T07:55:56Z | 2023-08-30T01:08:37Z | 2022-11-25T15:03:09Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexisrolland | 13,064,696 | MDQ6VXNlcjEzMDY0Njk2 | User | false |
huggingface/diffusers | 2,369,110,943 | I_kwDOHa8MBc6NNcOf | 8,671 | https://github.com/huggingface/diffusers/issues/8671 | https://api.github.com/repos/huggingface/diffusers/issues/8671 | Potentially redundant code in `train_dreambooth_lora_sd3.py` | ### Describe the bug
The sd3 pipeline has been initialized twice in the example code for `train_dreambooth_lora_sd3.py`.
This is the code from L1641-L1655. You may also check it via [this link](https://github.com/huggingface/diffusers/blob/963ee05d164192956e1bf3f157ee4da077460f9e/examples/dreambooth/train_dreamb... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-06-24T03:22:41Z | 2024-06-24T12:51:07Z | 2024-06-24T12:30:34Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hmdliu | 30,016,169 | MDQ6VXNlcjMwMDE2MTY5 | User | false |
huggingface/diffusers | 2,369,214,235 | I_kwDOHa8MBc6NN1cb | 8,673 | https://github.com/huggingface/diffusers/issues/8673 | https://api.github.com/repos/huggingface/diffusers/issues/8673 | Attention masks are missing in SD3 to mask out text padding tokens | ### Describe the bug
In the attention implementation of SD3, attention masks currently are not used. This will result in inconsistent outputs for the different values `max_seq_length` where padding exists in text tokens as the attention scores of padding tokens are non-zero. This issue has been discussed in https://gi... | open | null | false | 9 | [
"wip",
"contributions-welcome"
] | [] | 2024-06-24T05:03:16Z | 2026-01-09T04:52:46Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | reminisce | 4,978,794 | MDQ6VXNlcjQ5Nzg3OTQ= | User | false |
huggingface/diffusers | 2,370,425,363 | I_kwDOHa8MBc6NSdIT | 8,683 | https://github.com/huggingface/diffusers/issues/8683 | https://api.github.com/repos/huggingface/diffusers/issues/8683 | Why do Diffusers schedulers produce lower quality outputs compared to ComfyUI? | ### Discussed in https://github.com/huggingface/diffusers/discussions/8682
<sup>Originally posted by **nducthang** June 24, 2024</sup>
Hi,
I'm encountering an issue when comparing the quality of ComfyUI and Diffusers. I've noticed that the output of Diffusers is consistently lower than ComfyUI in many cases, des... | closed | completed | false | 20 | [] | [] | 2024-06-24T14:37:19Z | 2024-06-25T06:06:12Z | 2024-06-25T06:06:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nducthang | 42,518,675 | MDQ6VXNlcjQyNTE4Njc1 | User | false |
huggingface/diffusers | 2,370,615,817 | I_kwDOHa8MBc6NTLoJ | 8,686 | https://github.com/huggingface/diffusers/issues/8686 | https://api.github.com/repos/huggingface/diffusers/issues/8686 | `SDXLCFGCutoffCallback` does not work with `StableDiffusionXLControlNetPipeline` | ### Describe the bug
Running CFGCutoffCallback with ControlNet SDXL will raise following error
````
diffusers/src/diffusers/models/attention.py:372, in BasicTransformerBlock.forward(self, hidden_states, attention_mask, encoder_hidden_states, encoder_attention_mask, timestep, cross_attention_kwargs, class_labels, add... | closed | completed | false | 6 | [
"bug"
] | [
"asomoza"
] | 2024-06-24T16:10:33Z | 2024-10-23T16:21:58Z | 2024-10-23T16:21:58Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rootonchair | 23,548,268 | MDQ6VXNlcjIzNTQ4MjY4 | User | false |
huggingface/diffusers | 2,371,647,744 | I_kwDOHa8MBc6NXHkA | 8,689 | https://github.com/huggingface/diffusers/issues/8689 | https://api.github.com/repos/huggingface/diffusers/issues/8689 | A bug about one-step inference in PixArtAlphaPipeline | ### Describe the bug
When implementing the PixArtAlphaPipeline, one step of inference was bound to the DMD, which is inappropriate. This resulted in errors in other one-step inference codes based on Pixar-alpha.
### Reproduction
`import torch
from diffusers import PixArtAlphaPipeline, LCMScheduler, Transforme... | open | null | false | 6 | [
"bug",
"stale"
] | [] | 2024-06-25T04:52:01Z | 2024-12-13T15:05:40Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luo-Yihong | 65,936,780 | MDQ6VXNlcjY1OTM2Nzgw | User | false |
huggingface/diffusers | 2,371,916,003 | I_kwDOHa8MBc6NYJDj | 8,693 | https://github.com/huggingface/diffusers/issues/8693 | https://api.github.com/repos/huggingface/diffusers/issues/8693 | SD3 + SDXL refine fix lying on grass. How to do in diffusers colab workflow? | this is comfy workflow
how can i do in diffusers colab workflow? | closed | completed | false | 3 | [
"stale"
] | [] | 2024-06-25T07:30:55Z | 2024-09-23T11:37:25Z | 2024-09-23T11:37:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | s9anus98a | 151,509,142 | U_kgDOCQfYlg | User | false |
huggingface/diffusers | 2,373,413,034 | I_kwDOHa8MBc6Nd2iq | 8,700 | https://github.com/huggingface/diffusers/issues/8700 | https://api.github.com/repos/huggingface/diffusers/issues/8700 | [PAG] add `StableDiffusionXLControlNetPAGImg2ImgPipeline` | We recently integrated PAG into diffusers! See the PR here: https://github.com/huggingface/diffusers/pull/7944
Does anyone want to add a `StableDiffusionXLControlNetPAGImg2ImgPipeline`?
1. You should put it under the [pag folder](https://github.com/huggingface/diffusers/tree/main/src/diffusers/pipelines/pag)
2. yo... | closed | completed | false | 6 | [
"good first issue",
"help wanted",
"contributions-welcome"
] | [
"satani99"
] | 2024-06-25T18:52:18Z | 2024-08-21T17:24:23Z | 2024-08-21T17:24:23Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 2,374,335,186 | I_kwDOHa8MBc6NhXrS | 8,704 | https://github.com/huggingface/diffusers/issues/8704 | https://api.github.com/repos/huggingface/diffusers/issues/8704 | PAG is now supported in core 🤗 | Hello folks!
https://github.com/huggingface/diffusers/pull/7944/ introduced support for [Perturbed Attention Guidance (PAG)](https://ku-cvlab.github.io/Perturbed-Attention-Guidance/) which enhances image generation quality training-free.
| Generated Image without PAG | Generated Image with PAG |
|--------------... | closed | completed | false | 7 | [
"stale"
] | [] | 2024-06-26T05:41:29Z | 2024-11-09T15:23:47Z | 2024-11-09T15:23:47Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 2,374,364,322 | I_kwDOHa8MBc6Nheyi | 8,707 | https://github.com/huggingface/diffusers/issues/8707 | https://api.github.com/repos/huggingface/diffusers/issues/8707 | Trying to train sdxl text to image model, but getting the following error. | Map: 49%|████▉ | 977000/1985039 [5:01:52<5:08:01, 54.54 examples/s]
Map: 50%|█████ | 993000/1985039 [5:01:56<4:58:49, 55.33 examples/s]
Map: 50%|█████ | 994000/1985039 [5:01:56<5:05:28, 54.07 examples/s]
Map: 51%|█████ | 1017000/1985039 [5:01:56<4:45:43, 56.47 examples/s]
Map: 49%|████▉ | 9... | closed | completed | false | 5 | [] | [] | 2024-06-26T06:03:14Z | 2024-06-27T05:33:02Z | 2024-06-27T05:33:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | preethamp0197 | 170,499,459 | U_kgDOCimdgw | User | false |
huggingface/diffusers | 2,374,581,901 | I_kwDOHa8MBc6NiT6N | 8,708 | https://github.com/huggingface/diffusers/issues/8708 | https://api.github.com/repos/huggingface/diffusers/issues/8708 | [SD3] vae.config.shift_factor missing in dreambooth training examples | ### Describe the bug
shift_factor missing in traning code:
https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth_lora_sd3.py#L1617,
but used in inference code:
https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusio... | closed | completed | false | 7 | [
"bug"
] | [] | 2024-06-26T07:31:52Z | 2024-07-25T07:24:50Z | 2024-07-25T07:24:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bendanzzc | 26,294,676 | MDQ6VXNlcjI2Mjk0Njc2 | User | false |
huggingface/diffusers | 1,411,646,072 | I_kwDOHa8MBc5UJAJ4 | 871 | https://github.com/huggingface/diffusers/issues/871 | https://api.github.com/repos/huggingface/diffusers/issues/871 | [Community Pipeline] Speech to Image | ## Intro
Community Pipelines are introduced in `diffusers==0.4.0` with the idea of allowing the community to quickly add, integrate, and share their custom pipelines on top of `diffusers`.
You can find a guide about Community Pipelines [here](https://github.com/huggingface/diffusers/issues/841). You can also fin... | closed | completed | false | 9 | [
"good first issue",
"community-examples"
] | [
"MikailINTech"
] | 2022-10-17T14:04:29Z | 2022-10-26T03:57:02Z | 2022-10-25T13:50:41Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | osanseviero | 7,246,357 | MDQ6VXNlcjcyNDYzNTc= | User | false |
huggingface/diffusers | 2,374,691,448 | I_kwDOHa8MBc6Niup4 | 8,710 | https://github.com/huggingface/diffusers/issues/8710 | https://api.github.com/repos/huggingface/diffusers/issues/8710 | Add PAG support to SD1.5 | We recently integrated PAG into diffusers! See this PR [here] (https://github.com/huggingface/diffusers/pull/7944) we added PAG to SDXL
we also want to add PAG support to SD1.5 pipelines! we will need:
- [x] StableDiffusionPAGPipeline (assigned to @shauray8, PR https://github.com/huggingface/diffusers/pull/8725)
... | closed | completed | false | 17 | [
"good first issue",
"help wanted",
"contributions-welcome"
] | [] | 2024-06-26T08:23:17Z | 2024-10-09T20:40:59Z | 2024-10-09T20:40:59Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 2,374,904,681 | I_kwDOHa8MBc6Njitp | 8,711 | https://github.com/huggingface/diffusers/issues/8711 | https://api.github.com/repos/huggingface/diffusers/issues/8711 | Error thrown in `scheduling_dpmsolver_multistep` when re-used between validation steps | I am using `scheduling_dpmsolver_multistep` with 3D point cloud data with a second order update. Because the point clouds contain different numbers of points, I concatenate all the point clouds in a batch into a single cloud. This results in two batches having different numbers of points and thus different sizes.
Wh... | closed | completed | false | 5 | [
"scheduler"
] | [
"yiyixuxu"
] | 2024-06-26T10:00:19Z | 2024-07-02T08:21:52Z | 2024-07-02T08:21:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | EllingtonKirby | 8,492,979 | MDQ6VXNlcjg0OTI5Nzk= | User | false |
huggingface/diffusers | 1,411,665,765 | I_kwDOHa8MBc5UJE9l | 872 | https://github.com/huggingface/diffusers/issues/872 | https://api.github.com/repos/huggingface/diffusers/issues/872 | [Community Pipeline] Wildcards | ## Intro
Community Pipelines are introduced in `diffusers==0.4.0` with the idea of allowing the community to quickly add, integrate, and share their custom pipelines on top of `diffusers`.
You can find a guide about Community Pipelines [here](https://github.com/huggingface/diffusers/issues/841). You can also fin... | closed | completed | false | 10 | [
"community-examples",
"stale"
] | [] | 2022-10-17T14:17:08Z | 2023-04-21T15:03:54Z | 2023-04-21T15:03:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | osanseviero | 7,246,357 | MDQ6VXNlcjcyNDYzNTc= | User | false |
huggingface/diffusers | 2,377,724,982 | I_kwDOHa8MBc6NuTQ2 | 8,721 | https://github.com/huggingface/diffusers/issues/8721 | https://api.github.com/repos/huggingface/diffusers/issues/8721 | how to unload a pipeline | how to unload a pipeline and release the gpu memory | closed | completed | false | 3 | [] | [] | 2024-06-27T10:04:39Z | 2024-07-02T14:40:39Z | 2024-07-02T14:40:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nono909090 | 112,999,782 | U_kgDOBrw9Zg | User | false |
huggingface/diffusers | 2,378,169,404 | I_kwDOHa8MBc6Nv_w8 | 8,723 | https://github.com/huggingface/diffusers/issues/8723 | https://api.github.com/repos/huggingface/diffusers/issues/8723 | Error(s) in initializing SD3ControlNetModel by from_transformer | ### Describe the bug
### Reproduction
```python
from diffusers.models.controlnet_sd3 import SD3ControlNetModel
from diffusers.models.transformers import SD3Transformer2DModel
transformer = SD3Trans... | closed | completed | false | 15 | [
"bug"
] | [] | 2024-06-27T13:30:12Z | 2024-07-30T05:28:24Z | 2024-07-08T20:38:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ChenhLiwnl | 49,861,565 | MDQ6VXNlcjQ5ODYxNTY1 | User | false |
huggingface/diffusers | 2,379,308,748 | I_kwDOHa8MBc6N0V7M | 8,726 | https://github.com/huggingface/diffusers/issues/8726 | https://api.github.com/repos/huggingface/diffusers/issues/8726 | Text encoder errors in both SD3 dreambooth training scripts | ### Describe the bug
Multiple mismatched argument errors in train_dreambooth_sd3.py when training with train_text_encoder set to True.
Ordering/assignment errors in train_dreambooth_lora_sd3.py when training with train_text_encoder set to True.
### Reproduction
set --train_text_encoder and all other necessary... | open | null | false | 8 | [
"bug",
"stale"
] | [] | 2024-06-28T00:34:39Z | 2024-09-14T15:08:07Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | neuron-party | 96,799,331 | U_kgDOBcUKYw | User | false |
huggingface/diffusers | 2,379,472,076 | I_kwDOHa8MBc6N09zM | 8,727 | https://github.com/huggingface/diffusers/issues/8727 | https://api.github.com/repos/huggingface/diffusers/issues/8727 | Rethinking the `encode_prompt()` method in pipelines | This thread is for discussing the possibility of making the most widely used `encode_prompt()` methods of our pipelines `classmethod`s.
For historical context, I have made such attempts in the past but for different reasons, we decided to not move that forward. We are revisiting that now.
I have given it a goo... | closed | completed | false | 4 | [] | [] | 2024-06-28T03:09:49Z | 2024-11-19T01:57:51Z | 2024-11-19T01:57:51Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 2,379,550,902 | I_kwDOHa8MBc6N1RC2 | 8,728 | https://github.com/huggingface/diffusers/issues/8728 | https://api.github.com/repos/huggingface/diffusers/issues/8728 | Using `torchsde.BrownianInterval` instead of `torchsde.BrownianTree` in class `BatchedBrownianTree` | **Is your feature request related to a problem? Please describe.**
When I was doing some optimization for my pipeline, i found that the BrownianTree somehow took a bit more time.
**Describe the solution you'd like.**
I further dig into torchsde document, and found that they encouraged to use `BrownianInterval` to ... | closed | completed | false | 5 | [] | [] | 2024-06-28T04:33:55Z | 2024-09-12T08:46:54Z | 2024-09-12T08:46:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dianyo | 11,516,370 | MDQ6VXNlcjExNTE2Mzcw | User | false |
huggingface/diffusers | 2,379,717,006 | I_kwDOHa8MBc6N15mO | 8,729 | https://github.com/huggingface/diffusers/issues/8729 | https://api.github.com/repos/huggingface/diffusers/issues/8729 | Unexpected validation image on webdataset SDXL controlnet training script | ### Describe the bug
I am using the [webdataset sdxl controlnet training script](https://github.com/huggingface/diffusers/blob/main/examples/research_projects/controlnet/train_controlnet_webdataset.py).
I ran the script two times with 2 different learning rates in which both runs gave unexpected generated validat... | closed | completed | true | 2 | [
"bug"
] | [] | 2024-06-28T06:47:08Z | 2024-06-29T03:26:48Z | 2024-06-29T03:26:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | brycegoh | 16,372,429 | MDQ6VXNlcjE2MzcyNDI5 | User | false |
huggingface/diffusers | 1,411,672,877 | I_kwDOHa8MBc5UJGst | 873 | https://github.com/huggingface/diffusers/issues/873 | https://api.github.com/repos/huggingface/diffusers/issues/873 | [Community Pipelines] SD v1-v4 compare | ## Intro
Community Pipelines are introduced in `diffusers==0.4.0` with the idea of allowing the community to quickly add, integrate, and share their custom pipelines on top of `diffusers`.
You can find a guide about Community Pipelines [here](https://github.com/huggingface/diffusers/issues/841). You can also fin... | closed | completed | false | 2 | [
"good first issue",
"community-examples"
] | [] | 2022-10-17T14:21:40Z | 2022-12-13T10:31:35Z | 2022-12-13T10:31:35Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | osanseviero | 7,246,357 | MDQ6VXNlcjcyNDYzNTc= | User | false |
huggingface/diffusers | 2,380,981,479 | I_kwDOHa8MBc6N6uTn | 8,730 | https://github.com/huggingface/diffusers/issues/8730 | https://api.github.com/repos/huggingface/diffusers/issues/8730 | Implementation of DDIM, why taking Xt and (t-1) as input? | ### Describe the bug
I have tried to infer a diffusion model with DDIM with the number of timesteps = 10 and maximize timesteps as 1000.
I have printed the t in the for-loop, and the result is 901, 801, 801, 701, 601, 501, 401, 301, 201, 101, 1. It's really weird to me why 801 appears two times, and why we start f... | closed | completed | true | 1 | [
"bug"
] | [] | 2024-06-28T18:45:55Z | 2024-07-01T17:24:49Z | 2024-07-01T17:24:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | EPIC-Lab-sjtu | 172,622,062 | O_kgDOCkoA7g | Organization | false |
huggingface/diffusers | 2,381,289,635 | I_kwDOHa8MBc6N75ij | 8,731 | https://github.com/huggingface/diffusers/issues/8731 | https://api.github.com/repos/huggingface/diffusers/issues/8731 | Batched Text to image generation really slow with pipeline | ### Describe the bug
batched diffusers pipeline inference is really slow
### Reproduction
```
class prompt_data(Dataset):
def __init__(self,prompt_file):
self.txt_prompts = []
with open(prompt_file,'r') as file:self.txt_prompts = file.readlines()
def __len__(self... | closed | completed | false | 9 | [
"bug"
] | [] | 2024-06-28T22:48:53Z | 2024-11-20T02:39:10Z | 2024-11-20T02:39:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nighting0le01 | 81,643,693 | MDQ6VXNlcjgxNjQzNjkz | User | false |
huggingface/diffusers | 2,381,388,638 | I_kwDOHa8MBc6N8Rte | 8,732 | https://github.com/huggingface/diffusers/issues/8732 | https://api.github.com/repos/huggingface/diffusers/issues/8732 | Advanced training SD1.5 has an issue when saving checkpoints | ### Describe the bug
Today I trained using examples/dreambooth/train_dreambooth_lora.py in google colab, everything was OK
I wanted to try examples/advanced_diffusion_training/train_dreambooth_lora_sd15_advanced.py I use the stable diffusion 1.5 model original (which I cloned on my HF), but when I try to save to ... | closed | completed | false | 7 | [
"bug"
] | [] | 2024-06-29T01:43:25Z | 2024-07-05T18:46:10Z | 2024-07-05T18:46:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | josemerinom | 20,940,429 | MDQ6VXNlcjIwOTQwNDI5 | User | false |
huggingface/diffusers | 2,381,961,171 | I_kwDOHa8MBc6N-dfT | 8,738 | https://github.com/huggingface/diffusers/issues/8738 | https://api.github.com/repos/huggingface/diffusers/issues/8738 | `SD3Transformer2DModel` doesn't get configured | ### Describe the bug
`SD3Transformer2DModel` doesn't get correctly configured
### Reproduction
```python
from diffusers import SD3Transformer2DModel
config = {"caption_projection_dim": 1024, "pooled_projection_dim": 1024}
transformer = SD3Transformer2DModel(**config)
print(transformer)
```
### Logs
... | closed | completed | false | 7 | [
"bug"
] | [] | 2024-06-29T21:09:17Z | 2024-06-30T08:29:45Z | 2024-06-30T08:29:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MohamedAliRashad | 26,205,298 | MDQ6VXNlcjI2MjA1Mjk4 | User | false |
huggingface/diffusers | 2,382,191,345 | I_kwDOHa8MBc6N_Vrx | 8,742 | https://github.com/huggingface/diffusers/issues/8742 | https://api.github.com/repos/huggingface/diffusers/issues/8742 | Broken diffusers from version 0.29.0 LoRAAttnProcessor2_0 is removed project that might depend on this class will error | ```
from diffusers.models.attention_processor import (
Attention,
AttnProcessor2_0,
FusedAttnProcessor2_0,
LoRAAttnProcessor2_0,
LoRAXFormersAttnProcessor,
XFormersAttnProcessor,
)
```
ImportError: cannot import name 'LoRAAttnProcessor2_0' from 'diffusers.models.attention_process... | closed | completed | false | 2 | [] | [] | 2024-06-30T10:29:23Z | 2024-06-30T10:48:14Z | 2024-06-30T10:48:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | s9anus98a | 151,509,142 | U_kgDOCQfYlg | User | false |
huggingface/diffusers | 2,382,311,834 | I_kwDOHa8MBc6N_zGa | 8,744 | https://github.com/huggingface/diffusers/issues/8744 | https://api.github.com/repos/huggingface/diffusers/issues/8744 | use load from single file for advance model training | few days ago i was planing to train Dora , So i converted my model to Diffusers using load from single file and use original sdxl base config as mentiond in tutorials , the first issue when i used the original sdxl base for training it download more than 12 G all files !
also when i used my diffusers model(ABDALLAL... | closed | completed | false | 4 | [] | [] | 2024-06-30T15:45:33Z | 2024-06-30T19:05:26Z | 2024-06-30T18:01:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | al-swaiti | 95,714,669 | U_kgDOBbR9bQ | User | false |
huggingface/diffusers | 2,383,052,820 | I_kwDOHa8MBc6OCoAU | 8,748 | https://github.com/huggingface/diffusers/issues/8748 | https://api.github.com/repos/huggingface/diffusers/issues/8748 | SD3 cannot finetunes a better model (hand and face deformation)? | ### Describe the bug
I want to finetune sd3 to improve its human generation quality with 3million high-quality human datasets (which has been proven useful on sdxl and other models). But hand and face deformation doesn't improve much after two days of training.
I am using [train](https://github.com/huggingface/di... | closed | completed | true | 4 | [
"bug"
] | [] | 2024-07-01T07:21:19Z | 2024-07-17T06:01:31Z | 2024-07-17T06:01:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | KaiWU5 | 38,367,713 | MDQ6VXNlcjM4MzY3NzEz | User | false |
huggingface/diffusers | 1,411,679,324 | I_kwDOHa8MBc5UJIRc | 875 | https://github.com/huggingface/diffusers/issues/875 | https://api.github.com/repos/huggingface/diffusers/issues/875 | [Community Pipeline] Seed resize | ## Intro
Community Pipelines are introduced in `diffusers==0.4.0` with the idea of allowing the community to quickly add, integrate, and share their custom pipelines on top of `diffusers`.
You can find a guide about Community Pipelines [here](https://github.com/huggingface/diffusers/issues/841). You can also fin... | closed | completed | false | 9 | [
"community-examples",
"stale"
] | [
"daspartho"
] | 2022-10-17T14:25:26Z | 2022-11-21T04:44:10Z | 2022-11-21T04:44:10Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | osanseviero | 7,246,357 | MDQ6VXNlcjcyNDYzNTc= | User | false |
huggingface/diffusers | 2,383,355,054 | I_kwDOHa8MBc6ODxyu | 8,752 | https://github.com/huggingface/diffusers/issues/8752 | https://api.github.com/repos/huggingface/diffusers/issues/8752 | StableDiffusionXLPipeline.from_single_file can't use in offline server | ### Describe the bug
error
```
`local_files_only` is True but no local configs were found for this checkpoint.
Attempting to download the necessary config files for this pipeline.
Traceback (most recent call last):
File "/opt/conda/envs/tensorrt/lib/python3.11/site-packages/urllib3/connection.py", line 196, i... | closed | completed | false | 6 | [
"bug"
] | [] | 2024-07-01T09:42:19Z | 2024-07-29T05:32:12Z | 2024-07-02T10:09:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | 631068264 | 8,144,089 | MDQ6VXNlcjgxNDQwODk= | User | false |
huggingface/diffusers | 2,384,009,945 | I_kwDOHa8MBc6OGRrZ | 8,757 | https://github.com/huggingface/diffusers/issues/8757 | https://api.github.com/repos/huggingface/diffusers/issues/8757 | AutoEncoderKL only works for 2D shaped inputs | I wanted to use right off the bat the AutoEncoderKL class but to my surprise the Up and Down blocks are only for 2D data.
I want to use this class for 3D data.
A nice solution would be to adapt this code to make it work for 1D, 2D and 3D as well. (Higher might be overkill)
I wanted to write my own code for this ... | closed | completed | false | 5 | [] | [] | 2024-07-01T14:28:45Z | 2025-03-24T08:00:20Z | 2024-07-03T12:38:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DemyCode | 18,736,443 | MDQ6VXNlcjE4NzM2NDQz | User | false |
huggingface/diffusers | 2,384,626,889 | I_kwDOHa8MBc6OIoTJ | 8,759 | https://github.com/huggingface/diffusers/issues/8759 | https://api.github.com/repos/huggingface/diffusers/issues/8759 | SD3 - image_processor.py:111: RuntimeWarning: invalid value encountered in cast images = (images * 255).round().astype("uint8") | ### Describe the bug
Fails at late stage - after initial SD3 - possibly during VAE operation?
Warning message is:
/usr/local/lib/python3.10/dist-packages/diffusers/image_processor.py:111: RuntimeWarning: invalid value encountered in cast
images = (images * 255).round().astype("uint8")
Black image is created.... | closed | completed | false | 9 | [
"bug"
] | [] | 2024-07-01T19:55:38Z | 2024-11-20T02:32:20Z | 2024-11-20T02:32:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zagglez | 61,600,248 | MDQ6VXNlcjYxNjAwMjQ4 | User | false |
huggingface/diffusers | 1,411,684,862 | I_kwDOHa8MBc5UJJn- | 876 | https://github.com/huggingface/diffusers/issues/876 | https://api.github.com/repos/huggingface/diffusers/issues/876 | [Community Pipeline] Composable Diffusion | ## Intro
Community Pipelines are introduced in `diffusers==0.4.0` with the idea of allowing the community to quickly add, integrate, and share their custom pipelines on top of `diffusers`.
You can find a guide about Community Pipelines [here](https://github.com/huggingface/diffusers/issues/841). You can also fin... | closed | completed | false | 4 | [
"good first issue",
"community-examples"
] | [] | 2022-10-17T14:28:43Z | 2023-03-18T18:44:42Z | 2023-03-18T18:44:42Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | osanseviero | 7,246,357 | MDQ6VXNlcjcyNDYzNTc= | User | false |
huggingface/diffusers | 2,384,652,694 | I_kwDOHa8MBc6OIumW | 8,760 | https://github.com/huggingface/diffusers/issues/8760 | https://api.github.com/repos/huggingface/diffusers/issues/8760 | Clarification Needed on Hardcoded Value in Conditional Statement in LeditPP | Hello @manuelbrack,
I was reviewing the source code and came across a line that seems to have a hardcoded value in a conditional statement. The line in question is:
https://github.com/huggingface/diffusers/blob/0bae6e447cba0459456c4f7e7e87d7db141d3235/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_dif... | open | null | false | 3 | [
"stale"
] | [] | 2024-07-01T20:12:20Z | 2024-12-13T15:05:35Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ardofski | 34,197,573 | MDQ6VXNlcjM0MTk3NTcz | User | false |
huggingface/diffusers | 1,411,689,676 | I_kwDOHa8MBc5UJKzM | 877 | https://github.com/huggingface/diffusers/issues/877 | https://api.github.com/repos/huggingface/diffusers/issues/877 | [Community Pipeline] Checkpoint merging | ## Intro
Community Pipelines are introduced in `diffusers==0.4.0` with the idea of allowing the community to quickly add, integrate, and share their custom pipelines on top of `diffusers`.
You can find a guide about Community Pipelines [here](https://github.com/huggingface/diffusers/issues/841). You can also fin... | closed | completed | false | 20 | [
"good first issue",
"community-examples"
] | [] | 2022-10-17T14:31:37Z | 2023-03-18T18:44:50Z | 2023-03-18T18:44:50Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | osanseviero | 7,246,357 | MDQ6VXNlcjcyNDYzNTc= | User | false |
huggingface/diffusers | 2,386,219,494 | I_kwDOHa8MBc6OOtHm | 8,771 | https://github.com/huggingface/diffusers/issues/8771 | https://api.github.com/repos/huggingface/diffusers/issues/8771 | Removing LoRAAttnProcessor causes many dependencies to fail | ### Describe the bug
https://github.com/huggingface/diffusers/pull/8623 removed obsolete `LoRAAttnProcessor` which in principle is a good thing, but it was done without considerations where is that feature currently in-use so it breaks many (and i mean many) community pipelines
it also breaks some core libraries s... | closed | completed | false | 1 | [
"bug"
] | [] | 2024-07-02T13:11:33Z | 2024-07-03T16:37:08Z | 2024-07-03T16:37:08Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 1,411,698,013 | I_kwDOHa8MBc5UJM1d | 878 | https://github.com/huggingface/diffusers/issues/878 | https://api.github.com/repos/huggingface/diffusers/issues/878 | [Community Pipeline] Modified Cross-Attention - Structured Diffusion Guidance for Compositional T2I synthesis | ## Intro
Community Pipelines are introduced in `diffusers==0.4.0` with the idea of allowing the community to quickly add, integrate, and share their custom pipelines on top of `diffusers`.
You can find a guide about Community Pipelines [here](https://github.com/huggingface/diffusers/issues/841). You can also fin... | closed | completed | false | 8 | [
"community-examples",
"stale"
] | [
"patil-suraj"
] | 2022-10-17T14:36:34Z | 2023-04-21T15:03:53Z | 2023-04-21T15:03:53Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | apolinario | 788,417 | MDQ6VXNlcjc4ODQxNw== | User | false |
huggingface/diffusers | 2,387,832,613 | I_kwDOHa8MBc6OU28l | 8,780 | https://github.com/huggingface/diffusers/issues/8780 | https://api.github.com/repos/huggingface/diffusers/issues/8780 | Model and input data type is not same | **Is your feature request related to a problem? Please describe.**
Hi, when I trained sdv1.5 model with fp16 mode by using the `examples/text_to_image/train_text_to_image.py` file, I found there is a mismatch between unet model and input data. Specificaly, In this [line](https://github.com/huggingface/diffusers/blob/... | open | null | false | 1 | [
"stale"
] | [] | 2024-07-03T06:57:44Z | 2024-09-14T15:07:36Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andyjiang1116 | 87,074,272 | MDQ6VXNlcjg3MDc0Mjcy | User | false |
huggingface/diffusers | 2,388,233,203 | I_kwDOHa8MBc6OWYvz | 8,782 | https://github.com/huggingface/diffusers/issues/8782 | https://api.github.com/repos/huggingface/diffusers/issues/8782 | Integrate HiDiffusion | **Is your feature request related to a problem? Please describe.**
Object duplication and low efficiency happens when generating high-resolution image (2048x2048, 4096x4096) for stable diffusion models.
HiDiffusion can solve these problems, increasing the resolution and speed of pretrained diffusion models.
[paper ... | closed | completed | false | 10 | [] | [] | 2024-07-03T10:10:51Z | 2024-11-17T07:53:10Z | 2024-11-17T07:53:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ShenZhang-Shin | 58,697,142 | MDQ6VXNlcjU4Njk3MTQy | User | false |
huggingface/diffusers | 2,389,198,825 | I_kwDOHa8MBc6OaEfp | 8,785 | https://github.com/huggingface/diffusers/issues/8785 | https://api.github.com/repos/huggingface/diffusers/issues/8785 | adding PAG Support for Hunyuan-DIT and Pixart-Sigma | we recently added PAG support for SDXL. Is Anyone interested in extending PAG support to Hunyuan-DIT and Pixart-Sigma?
There is no implementation available, so it is a bit of a research-oriented project (= fun!!). and you can get directly feedbacks from the authors @sunovivid @HyoungwonCho
to add PAG support to n... | closed | completed | false | 4 | [
"help wanted",
"contributions-welcome",
"advanced"
] | [] | 2024-07-03T18:17:32Z | 2024-08-30T11:09:04Z | 2024-08-30T11:09:04Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 2,389,259,695 | I_kwDOHa8MBc6OaTWv | 8,786 | https://github.com/huggingface/diffusers/issues/8786 | https://api.github.com/repos/huggingface/diffusers/issues/8786 | max_sequence_length=512 doesn't have any effect on SD3 # of tokens | ### Describe the bug
max_sequence_length=512 doesn't have any effect on SD3 # of tokens
### Reproduction
```
import torch
from diffusers import StableDiffusion3Pipeline
pipe = StableDiffusion3Pipeline.from_single_file(
"https://huggingface.co/stabilityai/stable-diffusion-3-medium/blob/main/sd3_medium_incl_... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2024-07-03T19:01:53Z | 2024-10-15T18:21:14Z | 2024-10-15T18:21:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 2,389,842,731 | I_kwDOHa8MBc6Ochsr | 8,787 | https://github.com/huggingface/diffusers/issues/8787 | https://api.github.com/repos/huggingface/diffusers/issues/8787 | The model loading has suddenly become slow. | ### Describe the bug
The model loading has suddenly become slow.
<img width="2064" alt="image" src="https://github.com/huggingface/diffusers/assets/1147704/f464d7ee-cacc-4860-97e9-ae64857077bc">
### Reproduction
```
text2imgPipe = StableDiffusionControlNetPipeline.from_single_file(
'./base.safetensors',
... | closed | completed | false | 5 | [
"bug"
] | [] | 2024-07-04T03:12:33Z | 2024-08-12T05:57:21Z | 2024-08-12T05:57:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | crapthings | 1,147,704 | MDQ6VXNlcjExNDc3MDQ= | User | false |
huggingface/diffusers | 2,389,880,522 | I_kwDOHa8MBc6Ocq7K | 8,788 | https://github.com/huggingface/diffusers/issues/8788 | https://api.github.com/repos/huggingface/diffusers/issues/8788 | VAE Tiling not supported with SD3 for non power of 2 images? | ### Describe the bug
VAE tiling works for SD3 with power of 2 images, but for no other alignments.
The mentioned issues with VAE tiling are due to: [vae/config.json](https://huggingface.co/stabilityai/stable-diffusion-3-medium-diffusers/blob/main/vae/config.json)
Having:
```
"use_post_quant_conv": false,
"... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-07-04T03:52:54Z | 2024-07-11T20:41:37Z | 2024-07-11T20:41:37Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Teriks | 14,919,098 | MDQ6VXNlcjE0OTE5MDk4 | User | false |
huggingface/diffusers | 2,390,154,635 | I_kwDOHa8MBc6Odt2L | 8,790 | https://github.com/huggingface/diffusers/issues/8790 | https://api.github.com/repos/huggingface/diffusers/issues/8790 | SDXL TensorRT + IP Adapter | ### Model/Pipeline/Scheduler description
Dear authors,
I was wondering if there will be a pipeline that support SDXL + IP Adapter + TensorRT in near future? In the original implementation of TensorRT, they've implemented the SDXL + TensorRT but lacking the IP Adapter/ T2I Adapter...
### Open source status
- [X] T... | open | null | false | 1 | [
"stale"
] | [] | 2024-07-04T07:28:47Z | 2024-09-14T15:07:24Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Sundragon1993 | 9,301,684 | MDQ6VXNlcjkzMDE2ODQ= | User | false |
huggingface/diffusers | 2,390,676,780 | I_kwDOHa8MBc6OftUs | 8,792 | https://github.com/huggingface/diffusers/issues/8792 | https://api.github.com/repos/huggingface/diffusers/issues/8792 | Image quality degradation when migrating from automatic1111 to diffusers | When migrating from automatic1111 to diffusers, I'm experiencing a significant degradation in image quality despite using the same parameters. The images generated with diffusers are of noticeably lower quality compared to those produced by automatic1111. I also noticed I'm not the only one facing this issue [here](htt... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2024-07-04T11:46:27Z | 2024-10-15T18:23:26Z | 2024-10-15T18:23:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Swarzox | 16,492,753 | MDQ6VXNlcjE2NDkyNzUz | User | false |
huggingface/diffusers | 2,392,003,930 | I_kwDOHa8MBc6OkxVa | 8,794 | https://github.com/huggingface/diffusers/issues/8794 | https://api.github.com/repos/huggingface/diffusers/issues/8794 | ComfyUI example for train_dreambooth_lora_sdxl_advanced.py | Can you provide a ComfyUI example for the model in the train_dreambooth_lora_sdxl_advanced.py script?
I used the train_dreambooth_lora_sdxl_advanced.py script to train and perform inference, and achieved good results. I want to apply the trained lora and embedding on comfyUI, but the results are very poor?
Refer to... | closed | completed | false | 9 | [] | [] | 2024-07-05T07:20:32Z | 2024-07-18T07:45:44Z | 2024-07-18T07:45:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Heimdall-Nss | 58,160,555 | MDQ6VXNlcjU4MTYwNTU1 | User | false |
huggingface/diffusers | 2,393,361,393 | I_kwDOHa8MBc6Op8vx | 8,801 | https://github.com/huggingface/diffusers/issues/8801 | https://api.github.com/repos/huggingface/diffusers/issues/8801 | Support for Kolors | ### Model/Pipeline/Scheduler description
Yesterday Kwai-Kolors published their new model named **Kolors**, which uses unet as backbone and ChatGLM3 as text encoder.
Kolors is a large-scale text-to-image generation model based on latent diffusion, developed by the Kuaishou Kolors team. Trained on billions of text-... | closed | completed | false | 7 | [] | [] | 2024-07-06T03:16:31Z | 2024-07-11T16:09:19Z | 2024-07-11T16:09:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JincanDeng | 26,006,156 | MDQ6VXNlcjI2MDA2MTU2 | User | false |
huggingface/diffusers | 2,393,452,310 | I_kwDOHa8MBc6OqS8W | 8,803 | https://github.com/huggingface/diffusers/issues/8803 | https://api.github.com/repos/huggingface/diffusers/issues/8803 | duplicate assignment for hidden_states at models/attention_processor.py:1142 | ### Describe the bug
duplicate assignment for hidden_states at models/attention_processor.py:1142
### Reproduction
N/A
### Logs
_No response_
### System Info
N/A
### Who can help?
_No response_ | closed | completed | false | 2 | [
"bug"
] | [] | 2024-07-06T07:24:59Z | 2024-07-08T21:20:47Z | 2024-07-08T21:20:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | 131131yhx | 22,545,591 | MDQ6VXNlcjIyNTQ1NTkx | User | false |
huggingface/diffusers | 2,394,115,134 | I_kwDOHa8MBc6Os0w- | 8,805 | https://github.com/huggingface/diffusers/issues/8805 | https://api.github.com/repos/huggingface/diffusers/issues/8805 | convert_diffusers_to_original_sdxl.py doesn't work anymore | ### Describe the bug
Nowadays models are saved as `diffusion_pytorch_model-00001-of-00002.safetensors` and `diffusion_pytorch_model-00002-of-00002.safetensors` instead of a single `diffusion_pytorch_model.safetensors`
This causes convert_diffusers_to_original_sdxl.py to generate file not found error
```
FileNotFoun... | closed | completed | false | 10 | [
"bug",
"stale"
] | [] | 2024-07-07T16:08:09Z | 2025-01-29T04:21:33Z | 2024-11-18T20:09:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zyddnys | 17,797,982 | MDQ6VXNlcjE3Nzk3OTgy | User | false |
huggingface/diffusers | 2,395,703,112 | I_kwDOHa8MBc6Oy4dI | 8,809 | https://github.com/huggingface/diffusers/issues/8809 | https://api.github.com/repos/huggingface/diffusers/issues/8809 | Feedforward Output Inconsistency with Varying Sequence Lengths | ### Describe the bug
When using the Feedforward in `diffusers.models.attention`.
I've observed discrepancies in the results when processing subsets of the original input that vary in sequence length. However, I think the Feedforward mechanism should operate independently of sequence length?
### Reproduction
```p... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-07-08T13:49:07Z | 2024-07-08T16:36:06Z | 2024-07-08T16:36:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | soryxie | 78,358,442 | MDQ6VXNlcjc4MzU4NDQy | User | false |
huggingface/diffusers | 2,397,003,410 | I_kwDOHa8MBc6O316S | 8,813 | https://github.com/huggingface/diffusers/issues/8813 | https://api.github.com/repos/huggingface/diffusers/issues/8813 | Differential Diffusion with Kolors | **Is your feature request related to a problem? Please describe.**
One might wish to soft inpaint or edit pictures with differential diffusion with Kolors .
**Describe the solution you'd like.**
I believe we can port diff-diff to Kolors , by adding a new pipeline for it.
**Describe alternatives you've con... | closed | completed | false | 3 | [
"stale"
] | [] | 2024-07-09T03:14:35Z | 2024-10-15T18:16:27Z | 2024-10-15T18:16:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chuck-ma | 74,402,255 | MDQ6VXNlcjc0NDAyMjU1 | User | false |
huggingface/diffusers | 2,397,634,811 | I_kwDOHa8MBc6O6QD7 | 8,817 | https://github.com/huggingface/diffusers/issues/8817 | https://api.github.com/repos/huggingface/diffusers/issues/8817 | Can't use ip adapter faceid for some reason | ### Describe the bug
dimensions are wrong
### Reproduction
```
from insightface.app import FaceAnalysis
from insightface.utils import face_align
import numpy as np
import cv2
import PIL
from PIL import Image
from diffusers import ControlNetModel,T2IAdapter, StableDiffusionPipeline,EulerDiscreteScheduler,Aut... | closed | completed | false | 1 | [
"bug"
] | [] | 2024-07-09T09:19:55Z | 2024-07-09T12:01:21Z | 2024-07-09T12:01:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 2,399,793,216 | I_kwDOHa8MBc6PCfBA | 8,824 | https://github.com/huggingface/diffusers/issues/8824 | https://api.github.com/repos/huggingface/diffusers/issues/8824 | [Solved] How to make custom datasets for instruct-pix2pix? | ### Describe the bug
```
[rank0]: Traceback (most recent call last):
[rank0]: File "/opt/venv/lib/python3.10/site-packages/datasets/builder.py", line 1750, in _prepare_split_single
[rank0]: for key, record in generator:
[rank0]: File "/opt/venv/lib/python3.10/site-packages/datasets/packaged_modules/folde... | closed | completed | false | 1 | [
"bug"
] | [] | 2024-07-10T05:35:38Z | 2024-07-11T02:18:40Z | 2024-07-11T02:06:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jeonga0303 | 52,391,624 | MDQ6VXNlcjUyMzkxNjI0 | User | false |
huggingface/diffusers | 2,400,228,016 | I_kwDOHa8MBc6PEJKw | 8,826 | https://github.com/huggingface/diffusers/issues/8826 | https://api.github.com/repos/huggingface/diffusers/issues/8826 | RuntimeError: torch.cat(): expected a non-empty list of Tensors | ### Describe the bug
**Problem:**
When the "officially provided" example (see https://github.com/huggingface/diffusers/blob/a785992c1d6fcb1ff66f8a0d68d09a0a81b909e8/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion.py#L34) is executed, the error occurs.
**Expected behaviour:**
The example shoul... | open | null | false | 6 | [
"bug",
"stale"
] | [] | 2024-07-10T09:27:41Z | 2025-01-20T15:06:13Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | byrkbrk | 59,096,859 | MDQ6VXNlcjU5MDk2ODU5 | User | false |
huggingface/diffusers | 2,400,551,168 | I_kwDOHa8MBc6PFYEA | 8,827 | https://github.com/huggingface/diffusers/issues/8827 | https://api.github.com/repos/huggingface/diffusers/issues/8827 | error in using ip-adapter-faceid_sdxl | Hi
I want to use ip-adapter-faceid_sdxl as below code but I get error.
code :
```
import torch
import cv2
import numpy as np
from diffusers.utils import load_image
from diffusers import AutoPipelineForText2Image, AutoencoderKL, DDIMScheduler
from insightface.app import FaceAnalysis
image1 = load_image(... | closed | completed | false | 3 | [] | [] | 2024-07-10T11:49:04Z | 2024-10-20T23:02:06Z | 2024-10-20T23:02:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | saeedkhanehgir | 65,589,645 | MDQ6VXNlcjY1NTg5NjQ1 | User | false |
huggingface/diffusers | 2,400,811,251 | I_kwDOHa8MBc6PGXjz | 8,828 | https://github.com/huggingface/diffusers/issues/8828 | https://api.github.com/repos/huggingface/diffusers/issues/8828 | SD3 16GB training error. | ### Describe the bug
When I try to run the script "train_dreambooth_lora_sd3_miniature.py" with the argument "resume_from_checkpoint" it returns the following error:
`Traceback (most recent call last):
File "/kaggle/working/./diffusers/examples/research_projects/sd3_lora_colab/train_dreambooth_lora_sd3_miniature... | closed | completed | false | 4 | [
"bug"
] | [] | 2024-07-10T13:44:29Z | 2024-07-11T06:05:46Z | 2024-07-11T06:05:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | paoloski97 | 124,374,738 | U_kgDOB2nO0g | User | false |
huggingface/diffusers | 2,401,908,738 | I_kwDOHa8MBc6PKjgC | 8,831 | https://github.com/huggingface/diffusers/issues/8831 | https://api.github.com/repos/huggingface/diffusers/issues/8831 | LORA Unsupported Layers continuation from #6368 | ### Describe the bug
Based on #6368, I attempted the code provided for pruning the unsupported layers, but it actually prunes the entire dictionary.
### Reproduction
```
from requests import get
from diffusers import StableDiffusionXLPipeline, LCMScheduler
import safetensors
sdxl_pipeline_id = "stabilityai/sta... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2024-07-10T23:28:16Z | 2025-01-27T01:48:01Z | 2025-01-27T01:47:59Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AbhinavGopal | 43,016,805 | MDQ6VXNlcjQzMDE2ODA1 | User | false |
huggingface/diffusers | 2,402,136,768 | I_kwDOHa8MBc6PLbLA | 8,832 | https://github.com/huggingface/diffusers/issues/8832 | https://api.github.com/repos/huggingface/diffusers/issues/8832 | Crash when I run image = pipe(prompt).images[0] | ### Describe the bug
It crashes when I call `image = pipe(prompt).images[0]`. The code is:
```Python
from diffusers import DiffusionPipeline
pipe = DiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5")
pipe = pipe.to("mps")
# Recommended if your computer has < 64 GB of RAM
pipe.enable_attenti... | closed | completed | false | 7 | [
"bug"
] | [] | 2024-07-11T03:07:38Z | 2024-11-17T07:50:10Z | 2024-11-17T07:50:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhouhao27 | 8,099,731 | MDQ6VXNlcjgwOTk3MzE= | User | false |
huggingface/diffusers | 2,402,137,513 | I_kwDOHa8MBc6PLbWp | 8,833 | https://github.com/huggingface/diffusers/issues/8833 | https://api.github.com/repos/huggingface/diffusers/issues/8833 | stable-diffusion-3-medium-diffusers does not appear to have a file named config.json | ### Describe the bug
stable-diffusion-3-medium-diffusers does not appear to have a file named config.json
### Reproduction
stable-diffusion-3-medium-diffusers does not appear to have a file named config.json
### Logs
```shell
pipe = StableDiffusion3Pipeline.from_pretrained("stable-diffusion-3-medium-diffusers", to... | closed | completed | false | 0 | [
"bug"
] | [] | 2024-07-11T03:08:13Z | 2024-07-11T07:10:24Z | 2024-07-11T07:10:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | 404289680 | 15,352,958 | MDQ6VXNlcjE1MzUyOTU4 | User | false |
huggingface/diffusers | 2,402,167,994 | I_kwDOHa8MBc6PLiy6 | 8,834 | https://github.com/huggingface/diffusers/issues/8834 | https://api.github.com/repos/huggingface/diffusers/issues/8834 | Will the training code of SD3 Controlnet be released? | **Is your feature request related to a problem? Please describe.**
Training code of SD3 ControlNet
**Describe the solution you'd like.**
Could you please release training code of SD3 controlnet? I tried to train it but failed so I want to check whats the reason
| closed | completed | false | 3 | [] | [] | 2024-07-11T03:32:55Z | 2024-09-11T01:34:38Z | 2024-09-11T01:34:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ChenhLiwnl | 49,861,565 | MDQ6VXNlcjQ5ODYxNTY1 | User | false |
huggingface/diffusers | 2,402,228,024 | I_kwDOHa8MBc6PLxc4 | 8,835 | https://github.com/huggingface/diffusers/issues/8835 | https://api.github.com/repos/huggingface/diffusers/issues/8835 | Concerns about Transformer2DModelOutput and future censorship of models. | This may be a little pre-mature but want to get ahead of it. I've been following the recent changes in the diffusers library here, especially decisions to move to standardize output classes like `Transformer2DModelOutput`. You have all probably seen the deprecation warnings about this. While I see the benefits of consi... | closed | completed | false | 2 | [] | [] | 2024-07-11T04:35:44Z | 2024-11-18T20:19:12Z | 2024-11-18T20:19:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | 311-code | 23,625,562 | MDQ6VXNlcjIzNjI1NTYy | User | false |
huggingface/diffusers | 2,402,288,763 | I_kwDOHa8MBc6PMAR7 | 8,837 | https://github.com/huggingface/diffusers/issues/8837 | https://api.github.com/repos/huggingface/diffusers/issues/8837 | bug of SlicedAttnProcessor | ### Describe the bug
<img width="1075" alt="image" src="https://github.com/huggingface/diffusers/assets/18564791/e98ed9ef-b874-4502-9cd4-f775e279ad9a">
"batch_size_attention // self.slice_size" will forget to compute the last batch of attention。
should be this:
"(batch_size_attention - 1) // self.slice_siz... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-07-11T05:34:55Z | 2024-08-30T11:05:34Z | 2024-08-30T11:05:34Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shinetzh | 18,564,791 | MDQ6VXNlcjE4NTY0Nzkx | User | false |
huggingface/diffusers | 2,402,670,907 | I_kwDOHa8MBc6PNdk7 | 8,839 | https://github.com/huggingface/diffusers/issues/8839 | https://api.github.com/repos/huggingface/diffusers/issues/8839 | sd3 lora weight error | ### Describe the bug
` pipe.fuse_lora(lora_scale=lora_w)`
sd3 pipeline use this way to generate image, the image is terrible.
### Reproduction
```python
from diffusers import AutoPipelineForText2Image
pip... | closed | completed | false | 1 | [
"bug"
] | [] | 2024-07-11T09:11:27Z | 2024-07-22T11:25:14Z | 2024-07-22T11:25:14Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | efwfe | 31,640,191 | MDQ6VXNlcjMxNjQwMTkx | User | false |
huggingface/diffusers | 2,403,633,547 | I_kwDOHa8MBc6PRImL | 8,843 | https://github.com/huggingface/diffusers/issues/8843 | https://api.github.com/repos/huggingface/diffusers/issues/8843 | variable (per frame) IP Adapter weights in video | is there a (planned or existing) way to have variable IP Adapter weights for videos (e.g. with AnimateDiff)?
that means setting different values for different frames, as both scaling and masking currently seem to work with the whole generation at once (be it video or still image). | open | null | false | 6 | [
"stale",
"low-priority",
"consider-for-modular-diffusers"
] | [] | 2024-07-11T16:49:43Z | 2024-12-13T15:05:24Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eps696 | 4,808,408 | MDQ6VXNlcjQ4MDg0MDg= | User | false |
huggingface/diffusers | 2,404,528,117 | I_kwDOHa8MBc6PUi_1 | 8,847 | https://github.com/huggingface/diffusers/issues/8847 | https://api.github.com/repos/huggingface/diffusers/issues/8847 | No tqdm download progress bar for ControlNetModel | ### Describe the bug
Downloading a ControlNet model to the cache results in no TQDM progress bar being displayed on ``stderr``.
Seems to occur with any control net class.
Just make sure these models are not cached before testing.
I am guessing this also occurs with ``FlaxControlNetModel`` but it is not easy... | closed | completed | false | 5 | [
"bug"
] | [] | 2024-07-12T02:37:26Z | 2024-07-12T21:08:13Z | 2024-07-12T21:08:13Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Teriks | 14,919,098 | MDQ6VXNlcjE0OTE5MDk4 | User | false |
huggingface/diffusers | 2,404,739,491 | I_kwDOHa8MBc6PVWmj | 8,850 | https://github.com/huggingface/diffusers/issues/8850 | https://api.github.com/repos/huggingface/diffusers/issues/8850 | Community pipelines mirror broken | ### Describe the bug
When using a custom pipline with `StableDiffusionXLPipeline.from_pretrained`, the function fails due to a 404.
Upon further investigation, `https://huggingface.co/datasets/diffusers/community-pipelines-mirror/resolve/main/v0.29.2/lpw_stable_diffusion_xl.py` does not exist, but `https://huggingf... | closed | completed | false | 1 | [
"bug"
] | [] | 2024-07-12T04:44:19Z | 2024-07-16T12:00:29Z | 2024-07-16T12:00:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fubuki4649 | 48,634,762 | MDQ6VXNlcjQ4NjM0NzYy | User | false |
huggingface/diffusers | 2,405,081,776 | I_kwDOHa8MBc6PWqKw | 8,852 | https://github.com/huggingface/diffusers/issues/8852 | https://api.github.com/repos/huggingface/diffusers/issues/8852 | ASCEND TORCH_NPU INFERENCE performance improvement optimization point | FOR ASCEND TORCH_NPU BACKEND:
By the following configuration, private conv format is not allowed, which reduces format conversion and optimizes the speed of the conv operator. It can also avoid the possible functional issues caused by the propagation of private formats and the potential increase in memory usage. The ... | closed | completed | false | 12 | [
"stale"
] | [] | 2024-07-12T08:52:32Z | 2024-12-01T19:56:17Z | 2024-12-01T19:56:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lightnateriver | 48,556,430 | MDQ6VXNlcjQ4NTU2NDMw | User | false |
huggingface/diffusers | 2,405,190,650 | I_kwDOHa8MBc6PXEv6 | 8,854 | https://github.com/huggingface/diffusers/issues/8854 | https://api.github.com/repos/huggingface/diffusers/issues/8854 | Can't load textual embedding (SDXL) | ### Describe the bug
I downloaded a textual embedding but there is an error if I try to load it with load_textual_inversion.
Embedding unaestheticXL_Jug6.safetensors: https://civitai.com/models/119032?modelVersionId=207934
### Reproduction
```
import torch
from diffusers import StableDiffusionXLPipeline
pipe... | closed | completed | false | 7 | [
"bug"
] | [] | 2024-07-12T09:52:11Z | 2024-07-12T11:56:05Z | 2024-07-12T11:56:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | qdacsvx | 14,081,880 | MDQ6VXNlcjE0MDgxODgw | User | false |
huggingface/diffusers | 2,405,202,894 | I_kwDOHa8MBc6PXHvO | 8,855 | https://github.com/huggingface/diffusers/issues/8855 | https://api.github.com/repos/huggingface/diffusers/issues/8855 | "joint_attention_kwargs" don't pass the parameters to AttentionProcessor | I want to train my ipadapter in sd3. I noticed that joint_attention_kwargs is used in passing the parameters to AttentionProcessor (https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/transformers/transformer_sd3.py#L309). But it seems joint_attention_kwargs is not passed to transformer_blocks. Is m... | closed | completed | false | 4 | [] | [] | 2024-07-12T09:58:44Z | 2024-07-27T16:36:29Z | 2024-07-17T07:44:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | cyysc1998 | 35,159,700 | MDQ6VXNlcjM1MTU5NzAw | User | false |
huggingface/diffusers | 2,406,806,302 | I_kwDOHa8MBc6PdPMe | 8,858 | https://github.com/huggingface/diffusers/issues/8858 | https://api.github.com/repos/huggingface/diffusers/issues/8858 | how to know variants=fp16 beforehand | ### Describe the bug
In some diffusion checkponts, some are fp16 and some are not.
```
pipe = DiffusionPipeline.from_pretrained(
'model_id_1',
torch_dtype=torch.float16,
variant='fp16'
)
```
```
pipe = DiffusionPipeline.from_pretrained(
'model_id_2',
torch_dtype=torch.float16,
)
```
How to k... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2024-07-13T08:52:13Z | 2025-01-27T01:45:50Z | 2025-01-27T01:45:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pure-rgb | 45,315,076 | MDQ6VXNlcjQ1MzE1MDc2 | User | false |
huggingface/diffusers | 2,407,313,117 | I_kwDOHa8MBc6PfK7d | 8,863 | https://github.com/huggingface/diffusers/issues/8863 | https://api.github.com/repos/huggingface/diffusers/issues/8863 | AttributeError: 'tuple' object has no attribute 'shape' | ### Describe the bug
I've built the following pipeline with inpainting and Ip-adapter. And to reduce the memory footprint I also enabled CPU offload and efficient transformer methods, which causes the reported error.
### Reproduction
```python
from diffusers import AutoPipelineForInpainting
pipe = AutoPip... | closed | completed | false | 7 | [
"bug"
] | [] | 2024-07-14T07:50:02Z | 2024-11-09T01:40:52Z | 2024-11-09T01:40:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | innat-asj | 81,627,860 | MDQ6VXNlcjgxNjI3ODYw | User | false |
huggingface/diffusers | 2,410,582,417 | I_kwDOHa8MBc6PrpGR | 8,871 | https://github.com/huggingface/diffusers/issues/8871 | https://api.github.com/repos/huggingface/diffusers/issues/8871 | Bug still Found during Lora Training: ValueError: Attempting to unscale FP16 gradients. | ### Describe the bug
I was using the example code for lora training on SDXL model. However, as I launch the training, I get error `ValueError: Attempting to unscale FP16 gradients.`, which seems like training something fp16.
The error message is shown as follows:
```
The following values were not passed to `acce... | closed | completed | false | 5 | [
"bug"
] | [] | 2024-07-16T08:48:33Z | 2024-08-09T08:09:35Z | 2024-08-09T08:09:34Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dibbla | 72,083,122 | MDQ6VXNlcjcyMDgzMTIy | User | false |
huggingface/diffusers | 2,410,791,164 | I_kwDOHa8MBc6PscD8 | 8,872 | https://github.com/huggingface/diffusers/issues/8872 | https://api.github.com/repos/huggingface/diffusers/issues/8872 | Add warning if Attention Processor is changed after loading IP Adapter | Using a pipeline method that changes attention processors after loading the IP Adapter can lead to weird errors when running the pipeline. e.g: https://github.com/huggingface/diffusers/issues/8863
e.g.
```python
pipe.load_ip_adapter(
"h94/IP-Adapter",
subfolder="models",
weight_name="ip-adapter-plu... | closed | completed | false | 12 | [] | [] | 2024-07-16T10:27:21Z | 2024-11-09T01:41:27Z | 2024-11-09T01:41:27Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DN6 | 7,529,846 | MDQ6VXNlcjc1Mjk4NDY= | User | false |
huggingface/diffusers | 2,410,917,500 | I_kwDOHa8MBc6Ps658 | 8,873 | https://github.com/huggingface/diffusers/issues/8873 | https://api.github.com/repos/huggingface/diffusers/issues/8873 | The running using xformers is slower than when not using on A100 | ### Describe the bug
1. My diffusers version is 2.3.1+cu121 and xformers is 0.0.27
2. enable xformers with `pipe.enable_xformers_memory_efficient_attention()`
3. without xformers 4.27it/s. with xformers 4.00it/s on SDXL base model
I don't know wether the upscale attention will have an influence
https://github.co... | closed | completed | false | 7 | [
"bug"
] | [] | 2024-07-16T11:36:44Z | 2024-07-17T02:01:52Z | 2024-07-17T02:01:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bigmover | 26,063,296 | MDQ6VXNlcjI2MDYzMjk2 | User | false |
huggingface/diffusers | 2,411,553,667 | I_kwDOHa8MBc6PvWOD | 8,874 | https://github.com/huggingface/diffusers/issues/8874 | https://api.github.com/repos/huggingface/diffusers/issues/8874 | StableDiffusionPipeline.from_single_file() with CKPT fails: "UnpicklingError: Weights only load failed" | ### Describe the bug
Trying to load local CKPT file using the "from_single_file()" method fails. Works fine with .safetensors file from same repo (Runway ML SD).
### Reproduction
```import torch
from diffusers import StableDiffusionPipeline
device = 'cpu'
pipe = StableDiffusionPipeline.from_single_file("mod... | closed | completed | false | 16 | [
"bug",
"stale"
] | [] | 2024-07-16T16:07:29Z | 2025-01-27T01:53:52Z | 2025-01-27T01:53:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | whydna | 177,505 | MDQ6VXNlcjE3NzUwNQ== | User | false |
huggingface/diffusers | 2,412,532,163 | I_kwDOHa8MBc6PzFHD | 8,876 | https://github.com/huggingface/diffusers/issues/8876 | https://api.github.com/repos/huggingface/diffusers/issues/8876 | Pixart Sigma pipeline forgets to specify the resolution microconditioning | ### Describe the bug
Pixart Alpha pipeline:
```py
# 6.1 Prepare micro-conditions.
added_cond_kwargs = {"resolution": None, "aspect_ratio": None}
if self.transformer.config.sample_size == 128:
resolution = torch.tensor([height, width]).repeat(batch_size * num_images_per_prompt, ... | closed | completed | false | 6 | [
"bug"
] | [] | 2024-07-17T03:06:36Z | 2024-07-29T02:05:55Z | 2024-07-28T16:47:29Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AmericanPresidentJimmyCarter | 110,263,573 | U_kgDOBpJ9FQ | User | false |
huggingface/diffusers | 2,412,779,685 | I_kwDOHa8MBc6P0Bil | 8,879 | https://github.com/huggingface/diffusers/issues/8879 | https://api.github.com/repos/huggingface/diffusers/issues/8879 | torch.compile + cudagraph failed with StableDiffusionXL | ### Describe the bug
An error occurs when I run StatbleDiffusionXl using torch.compile (pipe.vae.decode, mode="reduce-overhead", fullgraph=True)
The necessary condition for trigger this error is torch.compile(pipe.vae.decode) + mode="reduce-overhead" + dtype=torch.float16
I suspect 'upcast_vae' caused it :
https:... | closed | completed | false | 18 | [
"bug",
"stale",
"torch-compile"
] | [] | 2024-07-17T06:36:47Z | 2024-10-19T16:43:02Z | 2024-10-19T16:43:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Danielmic | 30,855,238 | MDQ6VXNlcjMwODU1MjM4 | User | false |
huggingface/diffusers | 2,412,788,127 | I_kwDOHa8MBc6P0Dmf | 8,880 | https://github.com/huggingface/diffusers/issues/8880 | https://api.github.com/repos/huggingface/diffusers/issues/8880 | Inference speed half on version 0.26.0 upwards than 0.25.1 | ### Describe the bug
Hi,
I test the controlnet img2img inference on same machine and same input(image and prompt), the difference is the diffusers version, 0.25.1 vs 0.29.2. The speed of 0.25.1 in logs is 16it/s on average and 8it/s on 0.29.2. I also test inpaint and txt2img, also have performance loss.
### Reproduc... | closed | completed | false | 10 | [
"bug",
"stale"
] | [] | 2024-07-17T06:41:34Z | 2024-09-27T08:48:22Z | 2024-09-27T08:48:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | junnikokuki | 17,866,964 | MDQ6VXNlcjE3ODY2OTY0 | User | false |
huggingface/diffusers | 2,412,911,320 | I_kwDOHa8MBc6P0hrY | 8,881 | https://github.com/huggingface/diffusers/issues/8881 | https://api.github.com/repos/huggingface/diffusers/issues/8881 | How to Generate Multiple Image Inference in Instruct Pix2Pix | Hello, I am currently working on how to utilize Instruct Pix2Pix for augmentation.
For this purpose, I want to generate images by putting a Tensor of shape [64,3,84,84] (batch,channel,width,height)shape into the Instruct Pix2Pix pipeline, but the Instruct Pix2Pix provided by diffusers can only edit for one image.
Is ... | closed | completed | false | 0 | [] | [] | 2024-07-17T07:47:09Z | 2024-09-02T00:45:15Z | 2024-09-02T00:45:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | E-SJ | 70,588,265 | MDQ6VXNlcjcwNTg4MjY1 | User | false |
huggingface/diffusers | 2,414,216,150 | I_kwDOHa8MBc6P5gPW | 8,885 | https://github.com/huggingface/diffusers/issues/8885 | https://api.github.com/repos/huggingface/diffusers/issues/8885 | Unable to install clip-retrieval on colab and load modifier token when running Custom Diffusion example | ### Describe the bug
Hello,
When I tried to run the Custom Diffusion example on my colab, I met two issues:
1) I cannot install clip-retrieval on my colab and keep getting the following bug
```python
ERROR: Cannot uninstall 'blinker'. It is a distutils installed project and thus we cannot accurately determine wh... | closed | completed | false | 4 | [
"bug"
] | [] | 2024-07-17T17:58:05Z | 2024-07-17T21:50:23Z | 2024-07-17T21:50:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yj373 | 35,936,254 | MDQ6VXNlcjM1OTM2MjU0 | User | false |
huggingface/diffusers | 2,414,541,308 | I_kwDOHa8MBc6P6vn8 | 8,886 | https://github.com/huggingface/diffusers/issues/8886 | https://api.github.com/repos/huggingface/diffusers/issues/8886 | auraflow: attention mask is not passed into the sdpa function | ### Describe the bug
```py
hidden_states = F.scaled_dot_product_attention(
query, key, value, dropout_p=0.0, scale=attn.scale, is_causal=False
)
```
### Reproduction
when we calculate the hidden states inside the auraflow attention class, we are not passing the attention mask into the... | closed | completed | false | 4 | [
"bug"
] | [] | 2024-07-17T20:43:43Z | 2024-07-20T01:49:43Z | 2024-07-20T01:49:42Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 2,414,693,348 | I_kwDOHa8MBc6P7Uvk | 8,887 | https://github.com/huggingface/diffusers/issues/8887 | https://api.github.com/repos/huggingface/diffusers/issues/8887 | SD3 wrong VAE encode in training script | https://github.com/huggingface/diffusers/blob/c1dc2ae61968cccc3924784e4ed65ccfcd5c86eb/examples/dreambooth/train_dreambooth_lora_sd3.py#L1639
Should it be:
`model_input = (model_input - vae.config.shift_factor) * vae.config.scaling_factor` | closed | completed | false | 3 | [] | [
"sayakpaul"
] | 2024-07-17T22:05:53Z | 2024-07-21T10:54:05Z | 2024-07-21T10:54:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JinxuXiang | 60,202,283 | MDQ6VXNlcjYwMjAyMjgz | User | false |
huggingface/diffusers | 2,416,120,755 | I_kwDOHa8MBc6QAxOz | 8,894 | https://github.com/huggingface/diffusers/issues/8894 | https://api.github.com/repos/huggingface/diffusers/issues/8894 | Error in train_cm_ct_unconditional.py (Even after new commit) | Replication @dg845
``` bash
!accelerate launch --multi_gpu train_cm_ct_unconditional.py \
--dataset_name="cifar10" \
--dataset_image_column_name="img" \
--output_dir="/path/to/output/dir" \
--mixed_precision=fp16 \
--resolution=32 \
--max_train_steps=1000 \
--max_train_samples=10000... | open | null | false | 5 | [
"stale"
] | [] | 2024-07-18T11:30:44Z | 2025-01-17T15:05:32Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | KetanMann | 117,455,191 | U_kgDOBwA5Vw | User | false |
huggingface/diffusers | 2,416,129,921 | I_kwDOHa8MBc6QAzeB | 8,895 | https://github.com/huggingface/diffusers/issues/8895 | https://api.github.com/repos/huggingface/diffusers/issues/8895 | StableDiffusionLatentUpscalePipeline - positive/negative prompt embeds support | I'm trying to deploy the smallest possible SD inpainting model. My production deployment only needs unet+vae+ipadapter weights with prompt and ip adapter image embeds pre-generated. Works well!
Now I wanted to try latent upscaler from diffusers and realized it currently doesn't support pre-generated embeds. Would pr... | closed | completed | false | 5 | [
"good first issue",
"contributions-welcome"
] | [] | 2024-07-18T11:35:56Z | 2024-08-21T04:00:56Z | 2024-08-21T04:00:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DeTeam | 72,801 | MDQ6VXNlcjcyODAx | User | false |
huggingface/diffusers | 2,417,368,938 | I_kwDOHa8MBc6QFh9q | 8,898 | https://github.com/huggingface/diffusers/issues/8898 | https://api.github.com/repos/huggingface/diffusers/issues/8898 | Loading from_pretrained from subfolder with a path to directory raises ValueError: is not a folder containing a `.index.json` file or a pytorch_model.bin or a model.safetensors file | ### Describe the bug
Loading pretrained pipeline component such as UNET from subfolder with path does not work.
Most of your examples use the subfolder formulation, for example [here](https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image.py#L616) or [here](https://github.co... | closed | completed | false | 20 | [
"bug"
] | [] | 2024-07-18T20:47:58Z | 2024-09-04T10:54:38Z | 2024-07-19T03:55:27Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | christopher5106 | 6,875,375 | MDQ6VXNlcjY4NzUzNzU= | User | false |
huggingface/diffusers | 2,417,684,190 | I_kwDOHa8MBc6QGu7e | 8,900 | https://github.com/huggingface/diffusers/issues/8900 | https://api.github.com/repos/huggingface/diffusers/issues/8900 | How to load sd_xl_refiner_1.0.safetensors use from_single_file | ### Describe the bug
```
Traceback (most recent call last):
File "/workspace/work/private/TensorRT/demo/Diffusion/st_base.py", line 300, in <module>
A1111(local_dir, 'sd_xl_base_1.0.safetensors', steps=50, cfs_scale=8)
File "/workspace/work/private/TensorRT/demo/Diffusion/st_base.py", line 235, in A1111
... | closed | completed | false | 5 | [
"bug"
] | [] | 2024-07-19T01:58:05Z | 2024-07-26T10:39:07Z | 2024-07-26T10:39:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | 631068264 | 8,144,089 | MDQ6VXNlcjgxNDQwODk= | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.