repo stringclasses 1
value | github_id int64 1.27B 4.95B | github_node_id stringlengths 18 24 | number int64 8 14.3k | html_url stringlengths 49 53 | api_url stringlengths 59 63 | title stringlengths 1 402 | body stringlengths 1 62.9k ⌀ | state stringclasses 2
values | state_reason stringclasses 4
values | locked bool 2
classes | comments_count int64 0 99 | labels listlengths 0 5 | assignees listlengths 0 5 | created_at stringdate 2022-06-09 16:28:35 2026-07-22 10:32:58 | updated_at stringdate 2022-06-12 22:18:01 2026-07-22 10:32:58 | closed_at stringdate 2022-06-12 22:18:01 2026-07-21 13:17:28 ⌀ | author_association stringclasses 3
values | milestone_title stringclasses 0
values | snapshot_id stringclasses 149
values | extracted_at stringdate 2026-04-07 13:34:13 2026-07-22 10:35:16 | author_login stringlengths 3 28 | author_id int64 1.54k 296M | author_node_id stringlengths 12 20 | author_type stringclasses 3
values | author_site_admin bool 1
class |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
huggingface/diffusers | 1,769,933,945 | I_kwDOHa8MBc5pfwx5 | 3,855 | https://github.com/huggingface/diffusers/issues/3855 | https://api.github.com/repos/huggingface/diffusers/issues/3855 | Would a new "AutoPipeline" API abstraction be helpful? | Many checkpoints can be used for different tasks, e.g. img2img, text2img, but in order to do this at the moment, one has to know what pipelines are linked to each other, e.g. one has to know that `StableDiffusionPipeline` is linked to `StableDiffusionImg2ImgPipeline` when switching from text2img to img2img. Would a new... | closed | completed | false | 4 | [
"stale"
] | [
"yiyixuxu"
] | 2023-06-22T15:29:52Z | 2023-08-18T15:03:30Z | 2023-08-18T15:03:30Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,770,716,266 | I_kwDOHa8MBc5pivxq | 3,858 | https://github.com/huggingface/diffusers/issues/3858 | https://api.github.com/repos/huggingface/diffusers/issues/3858 | Controlnet execution sometime fails while downloading facenet.pth | Controlnet execution sometime fails while downloading facenet.pth. Is there a way to download it once and save it somewhere in filepath ?
```
model = OpenposeDetector.from_pretrained("lllyasviel/ControlNet")
File "/opt/conda/lib/python3.10/site-packages/controlnet_aux/open_pose/__init__.py", line 65, in from_p... | closed | completed | false | 1 | [] | [] | 2023-06-23T03:25:12Z | 2023-06-23T03:28:26Z | 2023-06-23T03:28:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vicky-aipaintr | 135,396,402 | U_kgDOCBH8Mg | User | false |
huggingface/diffusers | 1,772,656,443 | I_kwDOHa8MBc5pqJc7 | 3,861 | https://github.com/huggingface/diffusers/issues/3861 | https://api.github.com/repos/huggingface/diffusers/issues/3861 | Pipeline.ðfrom_pretrained or Pipeline.from_pretrained | ### Describe the bug
In the example, I notice that Pipeline.from_pretrained is called. But I can't find this method, and see ðfrom_pretrained in the source code, and I can only call ðfrom_pretrained. Strange characters, is this caused by some problem?
my environment:
wsl2:ubuntu 22
python:3.11.3
diffusers:0.17.1
... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-06-24T11:59:35Z | 2023-06-24T12:10:09Z | 2023-06-24T12:10:09Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aihao2000 | 51,043,929 | MDQ6VXNlcjUxMDQzOTI5 | User | false |
huggingface/diffusers | 1,772,684,615 | I_kwDOHa8MBc5pqQVH | 3,862 | https://github.com/huggingface/diffusers/issues/3862 | https://api.github.com/repos/huggingface/diffusers/issues/3862 | how to fine-tune the image-to-image model | The documentation only provides instructions for fine-tuning the text-to-image model, but there's no instructions on fine-tuning the image-to-image model.
- Firstly, can the image-to-image model be fine-tuned, or are there still unresolved issues?
- Secondly, are there any sample codes available, similar to the o... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-06-24T12:58:12Z | 2023-08-01T15:03:36Z | 2023-08-01T15:03:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | XiaoyuZhuang | 45,529,460 | MDQ6VXNlcjQ1NTI5NDYw | User | false |
huggingface/diffusers | 1,772,913,705 | I_kwDOHa8MBc5prIQp | 3,864 | https://github.com/huggingface/diffusers/issues/3864 | https://api.github.com/repos/huggingface/diffusers/issues/3864 | How do I get device_map="balance" working to shard? | So I currently have my pipeline loaded in memory. I can make many requests to it and it wonderfully generates images. However, I want to be able to generate images for multiple requests at the same time. I want to be able to generate multiple requests on a single GPU and multiple GPUs. I currently have two 4090s with 2... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-06-24T20:47:36Z | 2023-08-02T15:03:34Z | 2023-08-02T15:03:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JemiloII | 5,192,788 | MDQ6VXNlcjUxOTI3ODg= | User | false |
huggingface/diffusers | 1,773,753,404 | I_kwDOHa8MBc5puVQ8 | 3,871 | https://github.com/huggingface/diffusers/issues/3871 | https://api.github.com/repos/huggingface/diffusers/issues/3871 | text-to-image trainning resume from checkpoints get an error | ### Describe the bug
I try to train a text-to-image model by using the script **diffusers/examples/text_to_image**
it success when I train it from scratch
but when i add the param **--resume_from_checkpoint "checkpoint-10000"**
I get the error
**RuntimeError: Expected all tensors to be on the same device, but fou... | closed | completed | false | 4 | [
"bug",
"stale"
] | [
"sayakpaul"
] | 2023-06-26T03:07:46Z | 2023-08-03T15:03:37Z | 2023-08-03T15:03:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yijinsheng | 4,934,944 | MDQ6VXNlcjQ5MzQ5NDQ= | User | false |
huggingface/diffusers | 1,774,209,788 | I_kwDOHa8MBc5pwEr8 | 3,873 | https://github.com/huggingface/diffusers/issues/3873 | https://api.github.com/repos/huggingface/diffusers/issues/3873 | Conda version is not updated | Hi!
When trying to install diffusers with conda, it installs a pretty old version (0.11):
https://anaconda.org/anaconda/diffusers-torch
Are there plans to update the conda package to the newer version?
| closed | completed | false | 4 | [
"stale"
] | [] | 2023-06-26T08:18:20Z | 2023-08-05T15:04:08Z | 2023-08-05T15:04:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | RafailFridman | 22,198,039 | MDQ6VXNlcjIyMTk4MDM5 | User | false |
huggingface/diffusers | 1,774,280,151 | I_kwDOHa8MBc5pwV3X | 3,874 | https://github.com/huggingface/diffusers/issues/3874 | https://api.github.com/repos/huggingface/diffusers/issues/3874 | stablediffusion tensorrt demo error | I got this error:
UnsupportedOperatorError: Exporting the operator 'aten::scaled_dot_product_attention' to ONNX opset version 17 is not supported.
when runing the trt pipe line demo.
```
pipe = StableDiffusionPipeline.from_pretrained(
model_id,
custom_pipeline="stable_diffusion_tensorrt_txt2img",
... | closed | completed | false | 5 | [] | [] | 2023-06-26T08:59:36Z | 2023-08-07T18:34:31Z | 2023-07-24T02:24:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lucasjinreal | 21,303,438 | MDQ6VXNlcjIxMzAzNDM4 | User | false |
huggingface/diffusers | 1,776,130,542 | I_kwDOHa8MBc5p3Znu | 3,876 | https://github.com/huggingface/diffusers/issues/3876 | https://api.github.com/repos/huggingface/diffusers/issues/3876 | StableDiffusionControlNetInpaintPipeline Demo Error, Trying to Fix yields unexpected results | ### Describe the bug
Following instructions from https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint.py
there is a function for the controlnet image
```
def make_inpaint_condition(image, image_mask):
... image = np.array(image.convert("RG... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-06-27T05:40:40Z | 2023-07-29T04:44:30Z | 2023-07-29T04:44:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | djj0s3 | 1,120,027 | MDQ6VXNlcjExMjAwMjc= | User | false |
huggingface/diffusers | 1,776,373,300 | I_kwDOHa8MBc5p4U40 | 3,878 | https://github.com/huggingface/diffusers/issues/3878 | https://api.github.com/repos/huggingface/diffusers/issues/3878 | I am using train_ Instrument_ Error training model for imagefolder in pix2pix | ### Describe the bug
Traceback (most recent call last):
File "D:\packages\python\Envs\my-diffusers-J71KlA6V\lib\site-packages\datasets\builder.py", line 1629, in _prepare_split_single
for key, record in generator:
File "D:\packages\python\Envs\my-diffusers-J71KlA6V\lib\site-packages\datasets\packaged_module... | closed | completed | false | 11 | [
"bug",
"stale"
] | [] | 2023-06-27T08:27:48Z | 2023-08-04T15:03:18Z | 2023-08-04T15:03:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhenzi0322 | 34,839,719 | MDQ6VXNlcjM0ODM5NzE5 | User | false |
huggingface/diffusers | 1,776,422,938 | I_kwDOHa8MBc5p4hAa | 3,879 | https://github.com/huggingface/diffusers/issues/3879 | https://api.github.com/repos/huggingface/diffusers/issues/3879 | Can Chinese descriptive words be used for training from text to images | null | closed | completed | false | 2 | [] | [] | 2023-06-27T08:57:35Z | 2023-06-29T01:29:40Z | 2023-06-29T01:29:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhenzi0322 | 34,839,719 | MDQ6VXNlcjM0ODM5NzE5 | User | false |
huggingface/diffusers | 1,776,574,720 | I_kwDOHa8MBc5p5GEA | 3,880 | https://github.com/huggingface/diffusers/issues/3880 | https://api.github.com/repos/huggingface/diffusers/issues/3880 | Unmasked areas are degraded during inpainting. | ### Describe the bug
Unmasked areas are degraded during inpainting.
Easy to see the difference under th... | closed | completed | false | 16 | [
"bug",
"stale"
] | [] | 2023-06-27T10:16:21Z | 2023-08-20T15:03:39Z | 2023-08-20T15:03:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | oniatsu | 5,919,569 | MDQ6VXNlcjU5MTk1Njk= | User | false |
huggingface/diffusers | 1,776,665,633 | I_kwDOHa8MBc5p5cQh | 3,881 | https://github.com/huggingface/diffusers/issues/3881 | https://api.github.com/repos/huggingface/diffusers/issues/3881 | How many images and how many epochs are required to fine tune LORA for stable diffusion on custom image dataset | I am trying to finetune LORA on a movie dataset , but I am using custom dataset which has 3-4 movie characters , instead of using the actual names of the actor we are using in movie name of the characters , how big the dataset would be required in terms of total number of images, and number of images per character and ... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-06-27T11:05:53Z | 2023-08-04T15:03:17Z | 2023-08-04T15:03:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | atharmzaalo2023 | 134,913,498 | U_kgDOCAqd2g | User | false |
huggingface/diffusers | 1,777,232,945 | I_kwDOHa8MBc5p7mwx | 3,882 | https://github.com/huggingface/diffusers/issues/3882 | https://api.github.com/repos/huggingface/diffusers/issues/3882 | How to use models like chilloutmix to do inpainting task? | I tried as https://huggingface.co/docs/diffusers/api/diffusion_pipeline mentioned:
`text2img = StableDiffusionPipeline.from_pretrained("/data/cx/ysp/aigc-smart-painter/models/chilloutmix_NiPrunedFp32Fix")
inpaint = StableDiffusionInpaintPipeline(**text2img.components)
seger = RawSeger()
REST_API_URL = 'http://local... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-06-27T15:25:31Z | 2023-08-05T15:04:07Z | 2023-08-05T15:04:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AdamMayor2018 | 37,369,009 | MDQ6VXNlcjM3MzY5MDA5 | User | false |
huggingface/diffusers | 1,777,367,618 | I_kwDOHa8MBc5p8HpC | 3,883 | https://github.com/huggingface/diffusers/issues/3883 | https://api.github.com/repos/huggingface/diffusers/issues/3883 | Cannot use realistic vision in dreambooth | Is there any way to use realistic vision model with safetensors extension in diffusers library to train a dreambooth on it? I have tried almost everything but couln't find appropriate solution | closed | completed | false | 4 | [
"stale"
] | [] | 2023-06-27T16:46:58Z | 2023-08-05T15:04:05Z | 2023-08-05T15:04:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | borismartirosyandenovo | 132,906,166 | U_kgDOB-v8tg | User | false |
huggingface/diffusers | 1,777,592,915 | I_kwDOHa8MBc5p8-pT | 3,884 | https://github.com/huggingface/diffusers/issues/3884 | https://api.github.com/repos/huggingface/diffusers/issues/3884 | Error when use train_text_to_image_lora.py | ### Describe the bug
Got following error when try to use train_text_to_image_lora.py.
```
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cpu and cuda:0! (when checking argument for argument mat1 in method wrapper_CUDA_addmm)
```
here is full log:
https://pastebin.co... | closed | completed | false | 10 | [
"bug",
"stale"
] | [] | 2023-06-27T19:08:40Z | 2023-08-05T15:04:04Z | 2023-08-05T15:04:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ShunL12324 | 30,139,978 | MDQ6VXNlcjMwMTM5OTc4 | User | false |
huggingface/diffusers | 1,778,147,621 | I_kwDOHa8MBc5p_GEl | 3,885 | https://github.com/huggingface/diffusers/issues/3885 | https://api.github.com/repos/huggingface/diffusers/issues/3885 | are there any ways to reduce the vram usage of train_text_to_image.py? | ### Describe the bug
at present, the script almost run out of my gpu vram.i got a 4090. i set the bs=1,resolution 512,i enable the xformers.the training procedure used 23g vram with torch2.1, 22g vram with torch1.13.1.
and when i want to use the dynamo of torch2.1 to reduce vram usage, some weird things happened.
... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2023-06-28T04:19:45Z | 2023-08-19T15:03:22Z | 2023-08-19T15:03:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhangvia | 38,352,569 | MDQ6VXNlcjM4MzUyNTY5 | User | false |
huggingface/diffusers | 1,778,389,143 | I_kwDOHa8MBc5qABCX | 3,886 | https://github.com/huggingface/diffusers/issues/3886 | https://api.github.com/repos/huggingface/diffusers/issues/3886 | Watchdog caught collective operation timeout when training ddpm using train_unconditional.py | ### Describe the bug
When using mutil-gpu,and the data set is large or the resolution is high(need more time to inference), problems arise when saving the image.This is due to timeout in NCCL(1800s,30min)，it can be solved by the following code.
kwargs = InitProcessGroupKwargs(timeout=timedelta(seconds=7200))
... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-06-28T07:39:44Z | 2023-07-12T17:24:30Z | 2023-07-12T17:24:30Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dreaming-huang | 68,060,593 | MDQ6VXNlcjY4MDYwNTkz | User | false |
huggingface/diffusers | 1,778,663,240 | I_kwDOHa8MBc5qBD9I | 3,887 | https://github.com/huggingface/diffusers/issues/3887 | https://api.github.com/repos/huggingface/diffusers/issues/3887 | diffusers sd pipeline large difference in effect | ### Describe the bug
shape affects the semantic information of the generated image
### Reproduction
from lpw_stable_diffusion import StableDiffusionLongPromptWeightingPipeline
pipe = StableDiffusionLongPromptWeightingPipeline.from_pretrained(
model_id,
# torch_dtype=torch.f... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-06-28T10:29:01Z | 2023-08-05T15:04:02Z | 2023-08-05T15:04:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dushwe | 19,389,452 | MDQ6VXNlcjE5Mzg5NDUy | User | false |
huggingface/diffusers | 1,364,513,722 | I_kwDOHa8MBc5RVNO6 | 389 | https://github.com/huggingface/diffusers/issues/389 | https://api.github.com/repos/huggingface/diffusers/issues/389 | Code in readme seems to be broken | ### Describe the bug
Several sample code units provided in README.md are broken:
**1. Example of image generation**
This line does not work:
`image = pipe(prompt).images[0]`
Instead I used:
`image = pipe(prompt)["sample"][0] ` after reading the [blog](https://huggingface.co/blog/stable_diffusion)
**2. Exa... | closed | completed | false | 4 | [
"bug"
] | [] | 2022-09-07T11:22:06Z | 2022-09-13T16:08:48Z | 2022-09-13T16:08:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AndreyRub | 10,506,299 | MDQ6VXNlcjEwNTA2Mjk5 | User | false |
huggingface/diffusers | 1,778,713,637 | I_kwDOHa8MBc5qBQQl | 3,890 | https://github.com/huggingface/diffusers/issues/3890 | https://api.github.com/repos/huggingface/diffusers/issues/3890 | How to apply the schedulers in diffusers to original SD | Hi! Thanks for this great work! Diffusers helps me a lot in many aspects!
Because of my recent work, I would like to know wether the schedulers in diffusers can be directly used in original SD? If yes, what should I do?
Any response will be greatly appreciated! Again, thank you all for this convenient framework! | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-28T11:02:41Z | 2023-08-05T15:04:00Z | 2023-08-05T15:04:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | volcverse | 72,564,247 | MDQ6VXNlcjcyNTY0MjQ3 | User | false |
huggingface/diffusers | 1,778,849,720 | I_kwDOHa8MBc5qBxe4 | 3,891 | https://github.com/huggingface/diffusers/issues/3891 | https://api.github.com/repos/huggingface/diffusers/issues/3891 | Add Tortoise TTS as a pipeline | ### Model/Pipeline/Scheduler description
TorToise is a multi-voice text-to-speech system, which describes a way to apply recent advances in the image generative domain to speech synthesis. It would be great to have this model in diffusers.
I would love to contribute this.
### Open source status
- [X] The ... | open | null | false | 16 | [
"New pipeline/model",
"Good second issue"
] | [] | 2023-06-28T12:35:51Z | 2023-07-17T09:52:31Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | susnato | 56,069,179 | MDQ6VXNlcjU2MDY5MTc5 | User | false |
huggingface/diffusers | 1,779,086,324 | I_kwDOHa8MBc5qCrP0 | 3,892 | https://github.com/huggingface/diffusers/issues/3892 | https://api.github.com/repos/huggingface/diffusers/issues/3892 | Run the safety checker as a standalone module | I've been trying for some time to get the safety checker to run outside of the DiffusionPipeline but can´t seem to get it running.
After a lot of messing around with the loading, it seems something happens to the `vision_model` when initialized through `DiffusionPipeline.from_pretrained` which I'm missing.
Here is... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-28T14:32:23Z | 2023-08-05T15:03:59Z | 2023-08-05T15:03:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | thjan | 20,506,105 | MDQ6VXNlcjIwNTA2MTA1 | User | false |
huggingface/diffusers | 1,780,729,463 | I_kwDOHa8MBc5qI8Z3 | 3,898 | https://github.com/huggingface/diffusers/issues/3898 | https://api.github.com/repos/huggingface/diffusers/issues/3898 | convert checkpoints of inP2P to diffuser type | MagicBrush released their dataset and checkpoints of instructPix2Pix rencently,but the checkpoints doesn't match diffuser, can someone provide solution? | closed | completed | false | 5 | [
"stale"
] | [] | 2023-06-29T12:30:34Z | 2023-08-12T15:03:43Z | 2023-08-12T15:03:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hey0chen | 118,654,642 | U_kgDOBxKGsg | User | false |
huggingface/diffusers | 1,780,996,013 | I_kwDOHa8MBc5qJ9et | 3,901 | https://github.com/huggingface/diffusers/issues/3901 | https://api.github.com/repos/huggingface/diffusers/issues/3901 | ERROR: Could not open requirements file: [Errno 2] No such file or directory: 'requirements.txt' | ### Describe the bug
Running
pip install -r requirements.txt in the examples folder gives the following:
ERROR: Could not open requirements file: [Errno 2] No such file or directory: 'requirements.txt'
I find a requirements.txt not directly in the examples folder but in in many of the subfolders of examples... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-06-29T14:52:23Z | 2023-06-29T15:00:12Z | 2023-06-29T15:00:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | johanperjulius1 | 68,548,402 | MDQ6VXNlcjY4NTQ4NDAy | User | false |
huggingface/diffusers | 1,781,077,898 | I_kwDOHa8MBc5qKReK | 3,902 | https://github.com/huggingface/diffusers/issues/3902 | https://api.github.com/repos/huggingface/diffusers/issues/3902 | control-a-video | ### Model/Pipeline/Scheduler description
https://github.com/Weifeng-Chen/control-a-video
### Open source status
- [ ] The model implementation is available
- [ ] The model weights are available (Only relevant if addition is not a scheduler).
### Provide useful links for the implementation
_No response_ | closed | completed | false | 6 | [
"New scheduler",
"New pipeline/model"
] | [] | 2023-06-29T15:32:40Z | 2024-11-23T00:05:29Z | 2024-11-23T00:05:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,781,677,956 | I_kwDOHa8MBc5qMj-E | 3,906 | https://github.com/huggingface/diffusers/issues/3906 | https://api.github.com/repos/huggingface/diffusers/issues/3906 | invalid pipeline tests `test_float16_inference` | ### Describe the bug
I think this test is invalid
The same `components` objects are attached to both `pipe` and `pipe_fp16` - when we do `pipe_fp16.to(torch.float16)`, we change all the weights in `components` to `fp16` for both `pipe` and `pipe_fp16`
Am I right? I verified this by adding some lines to print... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-06-30T00:46:41Z | 2023-08-01T17:50:28Z | 2023-08-01T17:50:28Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 1,781,805,872 | I_kwDOHa8MBc5qNDMw | 3,907 | https://github.com/huggingface/diffusers/issues/3907 | https://api.github.com/repos/huggingface/diffusers/issues/3907 | StableDiffusionControlNetImg2ImgPipeline cannot load CKPT or Safetensor file directly | How to load a ckpt or safetensor file directly using StableDiffusionControlNetImg2ImgPipeline ? | open | null | false | 10 | [
"enhancement"
] | [] | 2023-06-30T03:18:02Z | 2023-07-19T12:48:03Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | thenol | 17,017,064 | MDQ6VXNlcjE3MDE3MDY0 | User | false |
huggingface/diffusers | 1,781,842,431 | I_kwDOHa8MBc5qNMH_ | 3,908 | https://github.com/huggingface/diffusers/issues/3908 | https://api.github.com/repos/huggingface/diffusers/issues/3908 | Cache model when demo with Streamlit | ### Describe the bug
I use `Streamlit` to start a webdemo of `StableDiffusionPipeline` on gpu. But when I delete model and call `torch.cuda.empty_cache()`, it cache about 2Gb on my GPU. Why this happen, because of `StableDiffusionPipeline` or `Streamlit`?
### Reproduction
```
import streamlit as st
import gc
... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-06-30T04:02:38Z | 2023-08-14T15:03:58Z | 2023-08-14T15:03:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tungdq212 | 104,067,740 | U_kgDOBjPynA | User | false |
huggingface/diffusers | 1,781,938,115 | I_kwDOHa8MBc5qNjfD | 3,910 | https://github.com/huggingface/diffusers/issues/3910 | https://api.github.com/repos/huggingface/diffusers/issues/3910 | Certain LoRA will not load | ### Describe the bug
# Issue
Certain LoRA will not load
**Diffusers version:** v0.17.1
## Examples
**Problem: ** LoRA file that will not load with SD 2.1 768
- [Pixhell LoRA](https://civitai.com/models/21461?modelVersionId=28943) which is supposed to work with [SD 2.1 768](https://huggingface.co/stabili... | closed | completed | false | 3 | [
"bug",
"stale"
] | [
"sayakpaul"
] | 2023-06-30T05:50:11Z | 2023-08-08T15:03:18Z | 2023-08-08T15:03:18Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | w4ffl35 | 25,737,761 | MDQ6VXNlcjI1NzM3NzYx | User | false |
huggingface/diffusers | 1,782,092,443 | I_kwDOHa8MBc5qOJKb | 3,913 | https://github.com/huggingface/diffusers/issues/3913 | https://api.github.com/repos/huggingface/diffusers/issues/3913 | Is there a bug when pre-cpmpute text embeddings? | When I read the examples/dreambooth/train_dreambooth_xx.py code, i find it confusing when pre-cpmpute text embeddings: when args.pre_compute_text_embeddings is set, it uses instance_prompt to compute it, but on the other situation, it uses class_prompt instead.
```python
# where computation happens: line 993
pre_com... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-06-30T08:08:47Z | 2023-08-01T19:00:18Z | 2023-08-01T19:00:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | krNeko9t | 42,951,801 | MDQ6VXNlcjQyOTUxODAx | User | false |
huggingface/diffusers | 1,782,464,095 | I_kwDOHa8MBc5qPj5f | 3,914 | https://github.com/huggingface/diffusers/issues/3914 | https://api.github.com/repos/huggingface/diffusers/issues/3914 | how to fine-tuning the sd model in low resolutions | When fine-tuning the stable diffusion model, there is a parameter called 'resolution' which, if set to a value like 128 or 256 to reduce GPU memory usage, could potentially have negative effects on training performance and results.
Would setting the resolution to a value other than 512, such as 128 or 256, have any ... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-30T12:42:12Z | 2023-08-08T15:03:16Z | 2023-08-08T15:03:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | XiaoyuZhuang | 45,529,460 | MDQ6VXNlcjQ1NTI5NDYw | User | false |
huggingface/diffusers | 1,782,543,631 | I_kwDOHa8MBc5qP3UP | 3,915 | https://github.com/huggingface/diffusers/issues/3915 | https://api.github.com/repos/huggingface/diffusers/issues/3915 | `torch.compile` doesn't seem to be working for text-to-video pipelines | ### Describe the bug
Trying to use `torch.compile` on a text-to-video model doesn't work
If I try to follow the docs and do a `pipe.unet.to(memory_format=torch.channels_last)`
```py
from diffusers import DiffusionPipeline, DPMSolverMultistepScheduler
pipe = DiffusionPipeline.from_pretrained("cerspense/zeroscop... | closed | completed | false | 6 | [
"bug"
] | [
"patrickvonplaten"
] | 2023-06-30T13:34:06Z | 2024-04-27T09:35:06Z | 2023-07-06T09:00:52Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | apolinario | 788,417 | MDQ6VXNlcjc4ODQxNw== | User | false |
huggingface/diffusers | 1,783,200,605 | I_kwDOHa8MBc5qSXtd | 3,917 | https://github.com/huggingface/diffusers/issues/3917 | https://api.github.com/repos/huggingface/diffusers/issues/3917 | Zero convolution? | Hi, I was comparing the diffusers implementation of stable diffusion vs stability's implementation. I noticed that stability uses zero convolution for the transformer out projection whereas diffusers do not. Can anyone help me understand why please?
Diffusers:
https://github.com/huggingface/diffusers/blame/da2ce1a6... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-06-30T21:53:02Z | 2023-08-14T15:03:54Z | 2023-08-14T15:03:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | okaris | 1,448,702 | MDQ6VXNlcjE0NDg3MDI= | User | false |
huggingface/diffusers | 1,783,376,184 | I_kwDOHa8MBc5qTCk4 | 3,918 | https://github.com/huggingface/diffusers/issues/3918 | https://api.github.com/repos/huggingface/diffusers/issues/3918 | How to control the position of an object in an image using text in a txt2img model? | How to control the position of an object in an image using text in a txt2img model? I know this is easy to achieve in an img2img model, but how can it be done in a txt2img model?
Or, how can a model be fine-tuned to achieve this effect? For example, specifying x=0, y=1, which corresponds to the top-left corner.
I... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-07-01T02:44:24Z | 2023-08-08T15:03:15Z | 2023-08-08T15:03:15Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | XiaoyuZhuang | 45,529,460 | MDQ6VXNlcjQ1NTI5NDYw | User | false |
huggingface/diffusers | 1,783,819,355 | I_kwDOHa8MBc5qUuxb | 3,919 | https://github.com/huggingface/diffusers/issues/3919 | https://api.github.com/repos/huggingface/diffusers/issues/3919 | convert_ldm_clip_checkpoint() got an unexpected keyword argument 'local_files_only' | I'm got this error while using [convert_from_ckpt.py]
D:\ai\HCP-Diffusion>python -m hcpdiff.tools.sd2diffusers --checkpoint_path="C:\Users\nmsl\Downloads\v1-5-pruned.ckpt" --original_config_file="C:\Users\nmsl\Downloads\v1-inference.yaml" --dump_path="D:\ai\HCP-Diffusion\model" --to_safetensors
WARNING[XFORMERS]: x... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-07-01T11:17:44Z | 2023-08-08T15:03:14Z | 2023-08-08T15:03:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Thymustw | 84,407,272 | MDQ6VXNlcjg0NDA3Mjcy | User | false |
huggingface/diffusers | 1,783,928,635 | I_kwDOHa8MBc5qVJc7 | 3,920 | https://github.com/huggingface/diffusers/issues/3920 | https://api.github.com/repos/huggingface/diffusers/issues/3920 | Can Schedule method be different for train and inference? | Hi guys,
Suppose I trained a Stable Diffusion with DDPM Schedule sampling. Is it possible to use other method Schedule in the inference phase? I want to know, the best practice is that to be same method in both train and inference? Or can be different? | closed | completed | false | 2 | [
"stale"
] | [] | 2023-07-01T13:46:44Z | 2023-07-31T15:17:04Z | 2023-07-31T15:17:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mzeynali | 95,717,811 | U_kgDOBbSJsw | User | false |
huggingface/diffusers | 1,784,291,625 | I_kwDOHa8MBc5qWiEp | 3,922 | https://github.com/huggingface/diffusers/issues/3922 | https://api.github.com/repos/huggingface/diffusers/issues/3922 | enable_sequential_cpu_offload followed by load_textual_inversion does not handle placement of loaded weights | ### Describe the bug
Using `pipe.enable_sequential_cpu_offload()` and `pipe.load_textual_inversion()` together crashes diffusers.
### Reproduction
```py
import torch
from diffusers import StableDiffusionPipeline
pipe = StableDiffusionPipeline.from_pretrained(
"runwayml/stable-diffusion-v1-5",
... | closed | completed | false | 21 | [
"bug",
"stale"
] | [] | 2023-07-01T22:19:20Z | 2023-10-30T15:09:40Z | 2023-10-30T15:09:40Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JeLuF | 5,852,422 | MDQ6VXNlcjU4NTI0MjI= | User | false |
huggingface/diffusers | 1,785,389,262 | I_kwDOHa8MBc5qauDO | 3,926 | https://github.com/huggingface/diffusers/issues/3926 | https://api.github.com/repos/huggingface/diffusers/issues/3926 | Docs: Add example for image-to-image-video from a file | **Is your feature request related to a problem? Please describe.**
I'm so happy that you have added image-to-image-video. Thank you!
One small thing. In the current example, it is upscaling a generated video, so maybe an example for styling of an existing video could maybe also be added?
I mean, show in an exam... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-07-03T05:52:43Z | 2023-08-10T15:04:06Z | 2023-08-10T15:04:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,785,729,365 | I_kwDOHa8MBc5qcBFV | 3,927 | https://github.com/huggingface/diffusers/issues/3927 | https://api.github.com/repos/huggingface/diffusers/issues/3927 | [ControlNet] Usage example fails with the default scheduler | The `StableDiffusionControlNetPipeline` errors out with the default scheduler (PNDM).
I took the code example from here: https://huggingface.co/docs/diffusers/main/en/api/pipelines/controlnet#usage-example.
To get reproducible results, check out the Colab Notebook: https://colab.research.google.com/gist/sayakpa... | closed | completed | false | 0 | [] | [] | 2023-07-03T09:17:49Z | 2023-07-03T13:10:09Z | 2023-07-03T13:10:09Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,786,557,901 | I_kwDOHa8MBc5qfLXN | 3,933 | https://github.com/huggingface/diffusers/issues/3933 | https://api.github.com/repos/huggingface/diffusers/issues/3933 | Graph compile breaks due to logging.info | ### Describe the bug
When building a unet via `torch.compile(pipeline.unet)`, I receive an error.
### Reproduction
1. Load SD 2.1 `ptx0/pseudo-flex-base`
2. Compile the unet
3. See that the error prevents continuance
4. Remove the logging.info line from the module
5. Compile works.
### Logs
```shell
in skipfil... | closed | completed | false | 15 | [
"bug"
] | [] | 2023-07-03T17:18:23Z | 2023-09-15T10:52:48Z | 2023-09-15T10:52:48Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 1,787,015,674 | I_kwDOHa8MBc5qg7H6 | 3,935 | https://github.com/huggingface/diffusers/issues/3935 | https://api.github.com/repos/huggingface/diffusers/issues/3935 | [Docs] Add docstrings about feed-forward chunking | @patrickvonplaten added support for feed-forward chunking to optimize memory usage for video pipelines. PRs: https://github.com/huggingface/diffusers/pull/3930, https://github.com/huggingface/diffusers/pull/3934.
Even though `enable_forward_chunking` is included in the docstring, I believe we should also add a litt... | closed | completed | false | 6 | [
"stale"
] | [
"patrickvonplaten"
] | 2023-07-04T01:56:02Z | 2024-02-11T15:06:13Z | 2024-02-11T15:06:13Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,787,204,512 | I_kwDOHa8MBc5qhpOg | 3,937 | https://github.com/huggingface/diffusers/issues/3937 | https://api.github.com/repos/huggingface/diffusers/issues/3937 | ZeroscopeXL will not work as normal model | ### Describe the bug
When using zeroscopeXL as a normal model I get this error:
`TypeError: VideoToVideoSDPipeline.__call__() got an unexpected keyword argument 'height'`
(Also, Animov, which also comes in more trained sizes, can't be used for upscaling. Is vid2vid limited to use the zeroscopeXL model, currently?... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-07-04T05:58:18Z | 2023-07-05T07:13:26Z | 2023-07-05T07:13:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,787,435,062 | I_kwDOHa8MBc5qihg2 | 3,938 | https://github.com/huggingface/diffusers/issues/3938 | https://api.github.com/repos/huggingface/diffusers/issues/3938 | Intermediate image at each time step | Is there an interface to retrieve the image generated at each time step during the diffusion/denoising process? | closed | completed | false | 3 | [
"stale"
] | [] | 2023-07-04T08:30:36Z | 2023-08-11T15:04:34Z | 2023-08-11T15:04:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zwx8981 | 14,050,646 | MDQ6VXNlcjE0MDUwNjQ2 | User | false |
huggingface/diffusers | 1,787,445,032 | I_kwDOHa8MBc5qij8o | 3,939 | https://github.com/huggingface/diffusers/issues/3939 | https://api.github.com/repos/huggingface/diffusers/issues/3939 | Regional prompts for Diffusers | While building images on complex prompts, it's missing out on certain details of prompts.
Automatic1111 has come up with an option for regional prompting now which can fix that to a good end
Is there a version in the diffusers pipeline as well? I am not able to locate
Eg. Currently I am doing controlNet Img2Img In... | closed | completed | false | 13 | [
"stale"
] | [] | 2023-07-04T08:36:39Z | 2023-08-21T15:03:50Z | 2023-08-21T15:03:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | 0sparsh2 | 50,074,241 | MDQ6VXNlcjUwMDc0MjQx | User | false |
huggingface/diffusers | 1,787,453,878 | I_kwDOHa8MBc5qimG2 | 3,940 | https://github.com/huggingface/diffusers/issues/3940 | https://api.github.com/repos/huggingface/diffusers/issues/3940 | huggingface folder in .cache folder in ubuntu and its exact purposes | hi. I want to create a system where the models are in external disks and I suddenly saw that the folder in the cache folder exists and it's something that clearly exists in the disk. What would be the most efficient approach for me? | closed | completed | false | 11 | [
"stale"
] | [] | 2023-07-04T08:42:13Z | 2023-08-06T14:10:45Z | 2023-08-06T14:10:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,788,174,525 | I_kwDOHa8MBc5qlWC9 | 3,944 | https://github.com/huggingface/diffusers/issues/3944 | https://api.github.com/repos/huggingface/diffusers/issues/3944 | Ability to customize network rank LoRA dreambooth training | **Is your feature request related to a problem? Please describe.**
For LoRA training, custom UIs like A1111 or kohya SD allow customization of the LoRA network rank which can make the results look much nicer but with the side effect of making LoRAs larger.
**Describe the solution you'd like**
Allowing configuratio... | closed | completed | false | 5 | [] | [] | 2023-07-04T15:57:48Z | 2023-07-17T14:48:19Z | 2023-07-17T13:59:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | isidentical | 47,358,913 | MDQ6VXNlcjQ3MzU4OTEz | User | false |
huggingface/diffusers | 1,788,264,338 | I_kwDOHa8MBc5qlr-S | 3,946 | https://github.com/huggingface/diffusers/issues/3946 | https://api.github.com/repos/huggingface/diffusers/issues/3946 | `from_ckpt` is a bad name | ### Describe the bug
We've gotten a couple of issues like: https://github.com/huggingface/diffusers/issues/3390 that show that `from_ckpt` is a bad name because people think they cannot load `safetensors` checkpoints with it. To me `ckpt` just stood for "checkpoint" and I didn't realize that people strongly associated... | closed | completed | false | 12 | [
"bug"
] | [] | 2023-07-04T17:28:02Z | 2023-07-12T19:22:31Z | 2023-07-06T16:18:59Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,788,636,442 | I_kwDOHa8MBc5qnG0a | 3,948 | https://github.com/huggingface/diffusers/issues/3948 | https://api.github.com/repos/huggingface/diffusers/issues/3948 | Generated images are pale | ### Describe the bug
I'm experimenting with prompted image editing. I use this [implementation](https://github.com/huggingface/diffusers/tree/main/examples/community#edict-image-editing-pipeline) using diffusers lib EDICT pipeline. Everything works pretty fine but I receive "pale" images. I do it using google colab ... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-07-05T01:54:25Z | 2023-08-12T15:03:36Z | 2023-08-12T15:03:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bohdanWsib | 133,372,767 | U_kgDOB_MbXw | User | false |
huggingface/diffusers | 1,788,783,312 | I_kwDOHa8MBc5qnqrQ | 3,950 | https://github.com/huggingface/diffusers/issues/3950 | https://api.github.com/repos/huggingface/diffusers/issues/3950 | Found 3 more it/s for Euler_a to hit 50.73 it/s on a 4090. | ### Describe the bug
I was helping vladmandic test his diffusers integration into sdnext and found small changes to hit 47.73 it/s in his dev branch.
I decided to do a py profile, found the nonzero() function standing out a lot, and found it being called in scheduling_euler_ancestral_discrete.py in two places where... | closed | completed | false | 17 | [
"bug",
"Good Example PR"
] | [
"yiyixuxu"
] | 2023-07-05T05:17:18Z | 2023-08-28T15:08:10Z | 2023-08-28T15:08:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aifartist | 116,415,616 | U_kgDOBvBcgA | User | false |
huggingface/diffusers | 1,788,922,768 | I_kwDOHa8MBc5qoMuQ | 3,951 | https://github.com/huggingface/diffusers/issues/3951 | https://api.github.com/repos/huggingface/diffusers/issues/3951 | Demo code for UniDiffuser returns output of NaNs | ### Describe the bug
Model is able to go through all steps of diffusion but fails on the self.numpy_to_pil() step where it says it encounters illegal values. upon printing them i see the array is full of NaNs
### Reproduction
<img width="913" alt="Screen Shot 2023-07-05 at 3 07 37 AM" src="https://github.com/hugging... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-07-05T07:10:21Z | 2023-08-12T15:03:35Z | 2023-08-12T15:03:35Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ethansmith2000 | 98,723,285 | U_kgDOBeJl1Q | User | false |
huggingface/diffusers | 1,789,139,965 | I_kwDOHa8MBc5qpBv9 | 3,953 | https://github.com/huggingface/diffusers/issues/3953 | https://api.github.com/repos/huggingface/diffusers/issues/3953 | which scheduler is "DPM++ 2s a"? | I see the DPMSolverMultistepScheduler == "DPM++ 2M", but is there any pipe equal to the sampler named "DPM++ 2s a" in webui? | closed | completed | false | 3 | [
"stale"
] | [] | 2023-07-05T09:21:57Z | 2023-08-14T15:03:51Z | 2023-08-14T15:03:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CrazyBoyM | 35,400,185 | MDQ6VXNlcjM1NDAwMTg1 | User | false |
huggingface/diffusers | 1,789,186,114 | I_kwDOHa8MBc5qpNBC | 3,954 | https://github.com/huggingface/diffusers/issues/3954 | https://api.github.com/repos/huggingface/diffusers/issues/3954 | Wrong learning rate scheduler training step count for examples with multi-gpu. | ### Describe the bug
For cosine learning rate schedules, the learning rate should approach zero when the training ends. However, due to accelerator prepare logic a learning rate scheduler will be stepped N times for each optimization step when there are N GPUs. Thus, we need to multiply the steps by `num_processes` wh... | closed | completed | false | 12 | [
"bug"
] | [
"eliphatfs"
] | 2023-07-05T09:46:50Z | 2023-07-13T12:20:27Z | 2023-07-13T12:20:27Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eliphatfs | 23,738,781 | MDQ6VXNlcjIzNzM4Nzgx | User | false |
huggingface/diffusers | 1,790,297,996 | I_kwDOHa8MBc5qtceM | 3,958 | https://github.com/huggingface/diffusers/issues/3958 | https://api.github.com/repos/huggingface/diffusers/issues/3958 | LoRA (civitai format) with enable_model_cpu_offload | ### Describe the bug
LoRA (civitai format) with enable_model_cpu_offload option and ControlNet (have not tested with basic Stable Diffusion) does not work correctly. See the code and logs. There is a notebook to reproduce the problem.
### Reproduction
Colab notebook available here: https://colab.research.google.com... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-07-05T20:54:09Z | 2023-09-12T15:04:25Z | 2023-09-12T15:04:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jpmerc | 5,333,924 | MDQ6VXNlcjUzMzM5MjQ= | User | false |
huggingface/diffusers | 1,790,662,262 | I_kwDOHa8MBc5qu1Z2 | 3,959 | https://github.com/huggingface/diffusers/issues/3959 | https://api.github.com/repos/huggingface/diffusers/issues/3959 | "StableDiffusionControlNetInpaintPipeline" inpainting in edge areas always does not fully fit the mask | My goal is to use StableDiffusionControlNetInpaintPipeline to realize automatic dressing of models. I used controlnet with openpose and mask image of dress.But I found in edge areas, inpainting will more of less tamper with my mask image.
An example:
original image:
![dress1](https://github.com/huggingface/diffu... | closed | completed | false | 25 | [
"stale"
] | [
"yiyixuxu",
"patrickvonplaten"
] | 2023-07-06T02:08:16Z | 2024-01-09T15:09:00Z | 2024-01-09T15:09:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AdamMayor2018 | 37,369,009 | MDQ6VXNlcjM3MzY5MDA5 | User | false |
huggingface/diffusers | 1,790,772,917 | I_kwDOHa8MBc5qvQa1 | 3,960 | https://github.com/huggingface/diffusers/issues/3960 | https://api.github.com/repos/huggingface/diffusers/issues/3960 | HeaderTooLarge : Lora | ### Describe the bug
it show error while loading lora model which is a safetensors model
here is the lora model
https://civitai.com/api/download/models/31475
### Reproduction
import torch
import requests
from PIL import Image
from diffusers import StableDiffusionImg2ImgPipeline
from diffusers import DP... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-07-06T04:21:01Z | 2023-07-07T09:56:58Z | 2023-07-07T09:56:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | satou23 | 99,410,124 | U_kgDOBezgzA | User | false |
huggingface/diffusers | 1,791,341,272 | I_kwDOHa8MBc5qxbLY | 3,963 | https://github.com/huggingface/diffusers/issues/3963 | https://api.github.com/repos/huggingface/diffusers/issues/3963 | Gradio is not auto downloading Hugging Face model weights - auto downloading on Google Colab but not on Windows | ### Describe the bug
The below code automatically downloading necessary model files and weights on Google Colab but not on Windows
Any ideas? How to fix?
```
import gradio as gr
from diffusers import DiffusionPipeline
import torch
import base64
from io import BytesIO
import os
import gc
from shar... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-07-06T11:13:22Z | 2023-08-15T15:03:16Z | 2023-08-15T15:03:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FurkanGozukara | 19,240,467 | MDQ6VXNlcjE5MjQwNDY3 | User | false |
huggingface/diffusers | 1,791,385,473 | I_kwDOHa8MBc5qxl-B | 3,964 | https://github.com/huggingface/diffusers/issues/3964 | https://api.github.com/repos/huggingface/diffusers/issues/3964 | What's the difference between diffusers[training] and diffusers[torch] | In [readme doc](https://github.com/huggingface/diffusers), `pip install diffusers[torch]`.
In [training doc](https://huggingface.co/docs/diffusers/tutorials/basic_training), `pip install diffusers[training]`.
Why different package is designed? Can these two package be designed to only one. | closed | completed | false | 1 | [] | [] | 2023-07-06T11:39:25Z | 2023-07-07T07:49:56Z | 2023-07-07T07:49:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wikiwen | 38,649,098 | MDQ6VXNlcjM4NjQ5MDk4 | User | false |
huggingface/diffusers | 1,791,439,428 | I_kwDOHa8MBc5qxzJE | 3,965 | https://github.com/huggingface/diffusers/issues/3965 | https://api.github.com/repos/huggingface/diffusers/issues/3965 | without using pretrained models | How long will it take to train without using pretrained models | closed | completed | false | 3 | [] | [] | 2023-07-06T12:12:56Z | 2023-07-07T07:48:52Z | 2023-07-07T07:48:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bistuxlf | 63,131,915 | MDQ6VXNlcjYzMTMxOTE1 | User | false |
huggingface/diffusers | 1,791,447,359 | I_kwDOHa8MBc5qx1E_ | 3,966 | https://github.com/huggingface/diffusers/issues/3966 | https://api.github.com/repos/huggingface/diffusers/issues/3966 | freeze some parameters | How to freeze some parameters of the pretrained model when training | closed | completed | false | 1 | [] | [] | 2023-07-06T12:17:52Z | 2023-07-07T07:46:36Z | 2023-07-07T07:46:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bistuxlf | 63,131,915 | MDQ6VXNlcjYzMTMxOTE1 | User | false |
huggingface/diffusers | 1,791,464,035 | I_kwDOHa8MBc5qx5Jj | 3,967 | https://github.com/huggingface/diffusers/issues/3967 | https://api.github.com/repos/huggingface/diffusers/issues/3967 | SD-XL not working anymore after you deleted the branch | ### Describe the bug
No errors but not loading anything into vram
installed requirements are like this
```
--extra-index-url https://download.pytorch.org/whl/cu113
torch==2.0.1
python-dotenv
accelerate
transformers
invisible-watermark
numpy>=1.17
PyWavelets>=1.1.1
opencv-python>=4.1.0.25
safetensors
... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-07-06T12:28:22Z | 2023-07-07T16:00:32Z | 2023-07-07T07:42:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FurkanGozukara | 19,240,467 | MDQ6VXNlcjE5MjQwNDY3 | User | false |
huggingface/diffusers | 1,791,859,312 | I_kwDOHa8MBc5qzZpw | 3,970 | https://github.com/huggingface/diffusers/issues/3970 | https://api.github.com/repos/huggingface/diffusers/issues/3970 | SDXL prompt embeddings not working | ### Describe the bug
```
pipe = StableDiffusionXLPipeline.from_pretrained(
model_id, torch_dtype=torch.float16, use_safetensors=True, variant="fp16", use_auth_token=hf_token).to("cuda")
cond, uncond = text_embeddings(
self.pipe, args_dict["prompt"], args_dict["negative_prompt"], clip_skip=2... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-07-06T16:17:52Z | 2023-07-07T15:31:58Z | 2023-07-07T14:50:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | adhikjoshi | 11,740,719 | MDQ6VXNlcjExNzQwNzE5 | User | false |
huggingface/diffusers | 1,791,978,704 | I_kwDOHa8MBc5qz2zQ | 3,973 | https://github.com/huggingface/diffusers/issues/3973 | https://api.github.com/repos/huggingface/diffusers/issues/3973 | TypeError: download_from_original_stable_diffusion_ckpt() got an unexpected keyword argument 'vae_path' | ### Describe the bug
There is a TypeError when using the script convert_original_stable_diffusion_to_diffusers.py
### Reproduction
!python convert_original_stable_diffusion_to_diffusers.py --from_safetensors --checkpoint_path="/content/Counterfeit-V2.5_pruned.safetensors" --dump_path='/content/model' --device='cuda... | closed | completed | false | 10 | [
"bug",
"stale"
] | [] | 2023-07-06T17:46:15Z | 2023-11-13T13:38:27Z | 2023-11-09T15:07:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hello202320 | 131,837,895 | U_kgDOB9uvxw | User | false |
huggingface/diffusers | 1,792,036,729 | I_kwDOHa8MBc5q0E95 | 3,974 | https://github.com/huggingface/diffusers/issues/3974 | https://api.github.com/repos/huggingface/diffusers/issues/3974 | StableDiffusionControlNetInpaintPipeline does not seem to support .safetensors? | ### Model/Pipeline/Scheduler description
I'd like to use Dreamshaper's inpaint model for StableDiffusionControlNetInpaintPipeline, I used the code of
pipe = StableDiffusionControlNetInpaintPipeline.from_single_file(
"/home/heran/DreamShaper/DreamShaper_6.31_INPAINTING.inpainting.safetensors", controln... | closed | completed | false | 9 | [
"stale",
"Good Example PR"
] | [] | 2023-07-06T18:29:33Z | 2023-08-20T15:03:33Z | 2023-08-20T15:03:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hzhou17 | 44,785,490 | MDQ6VXNlcjQ0Nzg1NDkw | User | false |
huggingface/diffusers | 1,792,151,798 | I_kwDOHa8MBc5q0hD2 | 3,975 | https://github.com/huggingface/diffusers/issues/3975 | https://api.github.com/repos/huggingface/diffusers/issues/3975 | typo in error message instructing to install safetensors | ### Describe the bug
typo in line
https://github.com/huggingface/diffusers/blob/51593da25aba44ed27d8680dcf0cfca4459f1e85/src/diffusers/pipelines/pipeline_utils.py#L1171
### Reproduction
load a `DiffusionPipeline.from_pretrained(..., use_safetensors=True, variant="fp16")` with `use_safetensors=True` then try to pip ... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-07-06T20:00:53Z | 2023-07-07T07:34:06Z | 2023-07-07T07:34:05Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | YoraiLevi | 50,873,841 | MDQ6VXNlcjUwODczODQx | User | false |
huggingface/diffusers | 1,792,655,677 | I_kwDOHa8MBc5q2cE9 | 3,978 | https://github.com/huggingface/diffusers/issues/3978 | https://api.github.com/repos/huggingface/diffusers/issues/3978 | StableDiffusionImg2Img not working on MPS | ### Describe the bug
Img2Img pipeline doesn't work. It works perfectly fine on text2img (using the same settings). I get this error:
```
RuntimeError: cannot reshape tensor of 0 elements into shape [0, -1, 1, 512] because the unspecified dimension size -1 can be any value and is ambiguous
```
### Reproduction
`... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-07-07T03:09:20Z | 2023-07-07T22:19:30Z | 2023-07-07T22:19:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ballenvironment | 111,252,136 | U_kgDOBqGSqA | User | false |
huggingface/diffusers | 1,364,679,797 | I_kwDOHa8MBc5RV1x1 | 398 | https://github.com/huggingface/diffusers/issues/398 | https://api.github.com/repos/huggingface/diffusers/issues/398 | Sampler change | I'm trying to control the sampler, but can't find a way to do it. How can I change the sampler? | closed | completed | false | 5 | [
"wip"
] | [
"patil-suraj"
] | 2022-09-07T13:37:35Z | 2026-01-09T17:55:46Z | 2026-01-09T17:55:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | BoazKG93 | 21,085,347 | MDQ6VXNlcjIxMDg1MzQ3 | User | false |
huggingface/diffusers | 1,792,720,983 | I_kwDOHa8MBc5q2sBX | 3,980 | https://github.com/huggingface/diffusers/issues/3980 | https://api.github.com/repos/huggingface/diffusers/issues/3980 | different results with A1111 when prompt is longer than 75 | The code is below
```
import torch
from diffusers import StableDiffusionPipeline, EulerDiscreteScheduler
pipeline = StableDiffusionPipeline.from_pretrained(
"runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16, safety_checker=None
).to("cuda")
pipeline.scheduler = EulerDiscreteScheduler.from_confi... | closed | completed | false | 17 | [
"stale"
] | [] | 2023-07-07T04:03:12Z | 2023-09-27T16:40:57Z | 2023-08-16T15:04:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | icech | 16,433,164 | MDQ6VXNlcjE2NDMzMTY0 | User | false |
huggingface/diffusers | 1,792,817,127 | I_kwDOHa8MBc5q3Dfn | 3,981 | https://github.com/huggingface/diffusers/issues/3981 | https://api.github.com/repos/huggingface/diffusers/issues/3981 | Timeembedding sin and cos is Error???? | I get code from transformers or ddpm, the time embedding is :
``` python
emb = torch.stack([torch.sin(emb), torch.cos(emb)], dim=-1)
emb = emb.view(T, d_model)#[sin,cos,sin,cos,.....]
```
but diffusers/src/diffusers/models/embeddings.py get_timestep_embedding is :
``` python
emb = torch.cat([torch.sin(emb), to... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-07-07T05:36:27Z | 2023-08-14T15:03:47Z | 2023-08-14T15:03:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kache1995 | 43,060,760 | MDQ6VXNlcjQzMDYwNzYw | User | false |
huggingface/diffusers | 1,792,907,198 | I_kwDOHa8MBc5q3Ze- | 3,984 | https://github.com/huggingface/diffusers/issues/3984 | https://api.github.com/repos/huggingface/diffusers/issues/3984 | ValueError: class_labels should be provided when num_class_embeds > 0 | ### Describe the bug
class_labels should be provided when num_class_embeds > 0
I found the error when i use unet in forward
### Reproduction
def forward(self,batch):
images = batch[:]
... | closed | completed | false | 9 | [
"bug",
"stale"
] | [
"yiyixuxu"
] | 2023-07-07T06:42:27Z | 2023-08-14T15:03:45Z | 2023-08-14T15:03:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | misi0202 | 101,965,629 | U_kgDOBhPfPQ | User | false |
huggingface/diffusers | 1,792,921,535 | I_kwDOHa8MBc5q3c-_ | 3,986 | https://github.com/huggingface/diffusers/issues/3986 | https://api.github.com/repos/huggingface/diffusers/issues/3986 | Support mesh output in `ShapEPipeline` | **Is your feature request related to a problem? Please describe.**
`ShapEPipeline` was added in `diffusers==0.18.0`, which is nice, but currently it doesn't support outputting the generated result in mesh formats and only supports outputting rendered video in a specific way like this:
![](https://github.com/hugging... | closed | completed | false | 1 | [] | [
"yiyixuxu"
] | 2023-07-07T06:54:12Z | 2023-07-24T00:54:09Z | 2023-07-24T00:54:09Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hysts | 25,161,192 | MDQ6VXNlcjI1MTYxMTky | User | false |
huggingface/diffusers | 1,793,137,786 | I_kwDOHa8MBc5q4Rx6 | 3,988 | https://github.com/huggingface/diffusers/issues/3988 | https://api.github.com/repos/huggingface/diffusers/issues/3988 | how to use part of the controlnet models with a "StableDiffusionControlNetInpaintPipeline" object? | I created a "StableDiffusionControlNetInpaintPipeline" object with a list of controlnet models such as "canny","openpose", but sometimes I want to use canny only or openpose only.Is there's a way to reuse part of the controlnet models with a already inited "StableDiffusionControlNetInpaintPipeline" object? | closed | completed | false | 4 | [] | [] | 2023-07-07T09:18:18Z | 2023-08-01T04:51:41Z | 2023-08-01T04:51:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AdamMayor2018 | 37,369,009 | MDQ6VXNlcjM3MzY5MDA5 | User | false |
huggingface/diffusers | 1,793,459,960 | I_kwDOHa8MBc5q5gb4 | 3,993 | https://github.com/huggingface/diffusers/issues/3993 | https://api.github.com/repos/huggingface/diffusers/issues/3993 | Different prompting strategy lead to quite different results. | The topic was discussed in #2431. And it mentioned that we may realize similar results by Compel.
But I tried the library today, and get results that are far apart.
- model id: https://huggingface.co/prompthero/openjourney-v4
- common parameter: CFG scale:7.5, schedular: DDM++ 2M, steps: 20, seed: 835450454
- A1111... | closed | completed | false | 12 | [
"stale"
] | [] | 2023-07-07T12:24:19Z | 2023-08-15T15:03:13Z | 2023-08-15T15:03:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | krNeko9t | 42,951,801 | MDQ6VXNlcjQyOTUxODAx | User | false |
huggingface/diffusers | 1,794,713,561 | I_kwDOHa8MBc5q-SfZ | 3,998 | https://github.com/huggingface/diffusers/issues/3998 | https://api.github.com/repos/huggingface/diffusers/issues/3998 | Inferencing with Lora | > Currently, LoRA is only supported for the attention layers of the UNet2DConditionalModel.
Does the above statement only apply to the training process, or does it also imply that only UNet is supported during the inference process?
I found that using the same LoRA model, the image quality generated by diffusers... | closed | completed | false | 5 | [] | [] | 2023-07-08T04:15:33Z | 2023-07-22T23:51:39Z | 2023-07-11T11:43:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | oops-oom | 12,108,174 | MDQ6VXNlcjEyMTA4MTc0 | User | false |
huggingface/diffusers | 1,794,882,024 | I_kwDOHa8MBc5q-7no | 4,000 | https://github.com/huggingface/diffusers/issues/4000 | https://api.github.com/repos/huggingface/diffusers/issues/4000 | Converting stable diffusion 2.0's ckpt file to diffusers' format | ### Describe the bug
Cannot convert stable diffusion 2.0's ckpt file into diffusers format using `convert_original_stable_diffusion_to_diffusers.py`
### Reproduction
Assuming you have used the stable diffusion 2.0 pipeline before, you can just run the following commands to reproduce the error:
```
python convert... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-07-08T10:27:19Z | 2023-08-15T15:03:11Z | 2023-08-15T15:03:10Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | larme | 66,847 | MDQ6VXNlcjY2ODQ3 | User | false |
huggingface/diffusers | 1,795,111,029 | I_kwDOHa8MBc5q_zh1 | 4,001 | https://github.com/huggingface/diffusers/issues/4001 | https://api.github.com/repos/huggingface/diffusers/issues/4001 | [Doc] Add demo(space) to Readme file | @patrickvonplaten
Shall we add space to this table? I can send a pull request if you want.
https://github.com/huggingface/diffusers#popular-tasks--pipelines | closed | completed | false | 4 | [] | [] | 2023-07-08T21:02:50Z | 2023-07-11T17:28:16Z | 2023-07-11T11:37:18Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kadirnar | 36,204,372 | MDQ6VXNlcjM2MjA0Mzcy | User | false |
huggingface/diffusers | 1,795,157,647 | I_kwDOHa8MBc5q_-6P | 4,002 | https://github.com/huggingface/diffusers/issues/4002 | https://api.github.com/repos/huggingface/diffusers/issues/4002 | StableDiffusionInpaintPipeline missing from_single_file | **Is your feature request related to a problem? Please describe.**
The inpaint pipeline is missing `from_single_file` method, which is inconsistent with the other pipelines.
**Describe the solution you'd like**
```py
generator = StableDiffusionInpaintPipeline.from_single_file(model)
```
**Describe alternative... | closed | completed | false | 3 | [] | [] | 2023-07-08T22:59:47Z | 2023-07-11T22:03:21Z | 2023-07-11T22:03:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ballenvironment | 111,252,136 | U_kgDOBqGSqA | User | false |
huggingface/diffusers | 1,795,176,432 | I_kwDOHa8MBc5rADfw | 4,003 | https://github.com/huggingface/diffusers/issues/4003 | https://api.github.com/repos/huggingface/diffusers/issues/4003 | [SD-XL] Enhanced inference control | ## Feature Request: Enhanced Control of the Inference Process in SDXL Pipeline
We are seeking to enhance the control over the denoising process within the SDXL pipeline. The objective is to better handle the generation and refinement of images through fine-tuning of noise application and control of the inference proce... | closed | completed | false | 17 | [] | [] | 2023-07-09T00:20:30Z | 2023-10-17T22:12:19Z | 2023-07-12T11:54:29Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 1,795,182,487 | I_kwDOHa8MBc5rAE-X | 4,004 | https://github.com/huggingface/diffusers/issues/4004 | https://api.github.com/repos/huggingface/diffusers/issues/4004 | [SD-XL] Ability to easily split prompt over the two text encoders | **Is your feature request related to a problem? Please describe.**
SDXL 0.9 comes with a new dual text encoder pipeline.
OpenCLIP ViT-bigG/14 and CLIP-L are both paired up in this pipeline. When running through ComfyUI, the CLIP nodes allow for inputting different pieces of the prompt, to different encoders. The de... | closed | completed | false | 10 | [] | [] | 2023-07-09T00:44:17Z | 2024-01-29T18:04:45Z | 2023-07-21T12:50:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 1,795,451,514 | I_kwDOHa8MBc5rBGp6 | 4,005 | https://github.com/huggingface/diffusers/issues/4005 | https://api.github.com/repos/huggingface/diffusers/issues/4005 | StableDiffusionInpaintPipeline doesn't preserve aspect ratio | **Is your feature request related to a problem? Please describe.**
StableDiffusionImg2ImgPipeline works without providing `width` and `height` (it automatically uses the original width/height and preserves the aspect ratio). On the other hand, StableDiffusionInpaintPipeline doesn't use the original width/height when
... | closed | completed | false | 3 | [
"stale"
] | [
"yiyixuxu"
] | 2023-07-09T15:36:15Z | 2023-08-18T15:03:19Z | 2023-08-18T15:03:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ballenvironment | 111,252,136 | U_kgDOBqGSqA | User | false |
huggingface/diffusers | 1,795,461,536 | I_kwDOHa8MBc5rBJGg | 4,006 | https://github.com/huggingface/diffusers/issues/4006 | https://api.github.com/repos/huggingface/diffusers/issues/4006 | Does `Gradient Checkpointing` is incompatible with `DeepSpeed Stage3` | ### Describe the bug
I tried to train a ControlNet, with both `DeepSpeed Stage-3`and `gradient checkpointing`, but unexpected errors will occur. There is no problem using either of these alone, the errors seems to happen in the loss backforward:
```
RuntimeError: The size of tensor a (0) must match the size of ten... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-07-09T16:02:04Z | 2023-08-18T15:03:18Z | 2023-08-18T15:03:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | liming-ai | 73,068,772 | MDQ6VXNlcjczMDY4Nzcy | User | false |
huggingface/diffusers | 1,795,463,503 | I_kwDOHa8MBc5rBJlP | 4,007 | https://github.com/huggingface/diffusers/issues/4007 | https://api.github.com/repos/huggingface/diffusers/issues/4007 | [SD-XL] Flax implementation | Is there a plan for supporting the SD-XL pipeline in flax? If so, I would be happy to help bringing it to diffusers 🤗 | closed | completed | false | 7 | [
"stale"
] | [] | 2023-07-09T16:07:04Z | 2023-08-17T15:03:33Z | 2023-08-17T15:03:33Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mar-muel | 19,345,805 | MDQ6VXNlcjE5MzQ1ODA1 | User | false |
huggingface/diffusers | 1,795,677,285 | I_kwDOHa8MBc5rB9xl | 4,013 | https://github.com/huggingface/diffusers/issues/4013 | https://api.github.com/repos/huggingface/diffusers/issues/4013 | [SDXL 0.9] Quick question regarding SDXL and example training scripts | Hi team, thank you for the hard work in porting SDXL 0.9 to diffusers. If I may, can I know what are the status of the compatibility of example training scripts (text-to-image, dreambooth, controlnet, etc) for SDXL 0.9? I'm using the version 0.18.1 of diffusers right now. | closed | completed | false | 1 | [] | [] | 2023-07-10T00:51:40Z | 2023-07-11T11:22:41Z | 2023-07-11T11:22:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xiankgx | 4,113,258 | MDQ6VXNlcjQxMTMyNTg= | User | false |
huggingface/diffusers | 1,795,921,997 | I_kwDOHa8MBc5rC5hN | 4,014 | https://github.com/huggingface/diffusers/issues/4014 | https://api.github.com/repos/huggingface/diffusers/issues/4014 | StableDiffusionXLPipeline's watermark introduces undesirable pixel artifacts in generated images. | ### Describe the bug
The introduction of [watermark in the StableDiffusionXLPipeline](https://github.com/huggingface/diffusers/blob/e3d71ad89abfee3817340b2245a49eec894a1705/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py#L810) has introduced undesirable pixel artifacts in the generated imag... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-07-10T04:32:21Z | 2023-08-17T15:03:32Z | 2023-08-17T15:03:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | newtonapple | 1,544 | MDQ6VXNlcjE1NDQ= | User | false |
huggingface/diffusers | 1,796,144,786 | I_kwDOHa8MBc5rDv6S | 4,017 | https://github.com/huggingface/diffusers/issues/4017 | https://api.github.com/repos/huggingface/diffusers/issues/4017 | [Lora] control weight block by bloak | https://github.com/hako-mikan/sd-webui-lora-block-weight
this project provides the function of controlling different loras weight block by block.
I think that is a good way to fuse different loras and control the effect of lora.
Is there any way we can control the weight in diffuers?
| closed | completed | false | 2 | [] | [] | 2023-07-10T07:18:09Z | 2023-07-13T08:51:04Z | 2023-07-13T08:51:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ZhouXiner | 29,979,258 | MDQ6VXNlcjI5OTc5MjU4 | User | false |
huggingface/diffusers | 1,796,218,376 | I_kwDOHa8MBc5rEB4I | 4,018 | https://github.com/huggingface/diffusers/issues/4018 | https://api.github.com/repos/huggingface/diffusers/issues/4018 | Outpaint Help | Hello, I am trying to implement outpaint, which doesn't have a pipeline in diffusers, but I thought it would be possible with `StableDiffusionInpaintPipeline`. However, I can't seem to get it to work.
My initial idea was to extend the borders of the image and fill it with white, then create a mask with white on the ... | closed | completed | false | 8 | [] | [] | 2023-07-10T08:06:01Z | 2023-07-14T03:30:43Z | 2023-07-14T02:32:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ballenvironment | 111,252,136 | U_kgDOBqGSqA | User | false |
huggingface/diffusers | 1,796,597,314 | I_kwDOHa8MBc5rFeZC | 4,021 | https://github.com/huggingface/diffusers/issues/4021 | https://api.github.com/repos/huggingface/diffusers/issues/4021 | [SD-XL] Running SD XL via "from_single_file" unnecessarily downloads a 10 GB CLIP model | ### Describe the bug
Running `StableDiffusionXLPipeline` downloads `laion/CLIP-ViT-bigG-14-laion2B-39B-b160k`, which is about 10 GB in size.
Is it possible to avoid this big download?
ComfyUI, for example, seems to be able to run SDXL without this huge CLIP model.
### Reproduction
Run this code example:
... | closed | completed | false | 18 | [
"bug",
"stale"
] | [] | 2023-07-10T11:48:03Z | 2023-11-02T08:34:38Z | 2023-08-17T15:03:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | n00mkrad | 61,149,547 | MDQ6VXNlcjYxMTQ5NTQ3 | User | false |
huggingface/diffusers | 1,797,022,677 | I_kwDOHa8MBc5rHGPV | 4,023 | https://github.com/huggingface/diffusers/issues/4023 | https://api.github.com/repos/huggingface/diffusers/issues/4023 | About CtrlSDXL | Hi, @sayakpaul,
I noticed that you have developed the control net branch of SDXL. Thank you very much for your work, which has been very helpful to me.
I tried to train this branch locally, but I have a problem.
The final release version of SDxl was trained on 1024 x 1024 images. Do we have to train the control ... | closed | completed | false | 3 | [] | [] | 2023-07-10T15:29:56Z | 2023-07-11T08:18:12Z | 2023-07-11T08:18:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pixeli99 | 46,072,190 | MDQ6VXNlcjQ2MDcyMTkw | User | false |
huggingface/diffusers | 1,797,132,755 | I_kwDOHa8MBc5rHhHT | 4,024 | https://github.com/huggingface/diffusers/issues/4024 | https://api.github.com/repos/huggingface/diffusers/issues/4024 | "Circle Consistency" Option in StableDiffusionPanoramaPipeline for Artifact-Free Panorama Generation | I think that it would be great to have a new feature in the StableDiffusionPanoramaPipeline that would enable the generation of panorama images without any stitching artifacts. Currently, the generated images appear like panoramas, but lack consistency when viewed in 360-degree viewers.
The proposed feature, "Circle... | closed | completed | false | 3 | [] | [] | 2023-07-10T16:35:55Z | 2023-07-12T15:36:15Z | 2023-07-12T15:36:15Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | EvgenyKashin | 21,174,773 | MDQ6VXNlcjIxMTc0Nzcz | User | false |
huggingface/diffusers | 1,797,283,724 | I_kwDOHa8MBc5rIF-M | 4,027 | https://github.com/huggingface/diffusers/issues/4027 | https://api.github.com/repos/huggingface/diffusers/issues/4027 | Unload lora weights after they have been loaded | ### Describe the bug
There's no method to unload LoRA weights after they have been loaded. An issue was opened already (https://github.com/huggingface/diffusers/issues/3689) but that doesn't solve the problem
### Reproduction
1. Generate without LoRA, a fresh loaded pipeline
![image](https://github.com/huggingfac... | closed | completed | false | 10 | [
"bug"
] | [] | 2023-07-10T18:21:49Z | 2023-07-14T11:04:55Z | 2023-07-14T11:04:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pietrobolcato | 3,061,306 | MDQ6VXNlcjMwNjEzMDY= | User | false |
huggingface/diffusers | 1,797,637,684 | I_kwDOHa8MBc5rJcY0 | 4,029 | https://github.com/huggingface/diffusers/issues/4029 | https://api.github.com/repos/huggingface/diffusers/issues/4029 | How can I make diffuser pipeline to use .safetensors file for SDXL? | Cloning entire repo is taking 100 GB
How can I make below code to use .safetensors file instead of diffusers?
Lets say I have downloaded my safetensors file into path.safetensors
How to provide it?
The below code working but we are cloning 100 GB instead of just single 14 GB safetensors. Waste of bandwidth... | closed | completed | false | 6 | [] | [] | 2023-07-10T21:52:22Z | 2023-12-11T18:45:18Z | 2023-07-11T11:55:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FurkanGozukara | 19,240,467 | MDQ6VXNlcjE5MjQwNDY3 | User | false |
huggingface/diffusers | 1,797,762,618 | I_kwDOHa8MBc5rJ646 | 4,030 | https://github.com/huggingface/diffusers/issues/4030 | https://api.github.com/repos/huggingface/diffusers/issues/4030 | Runtime Error in Stable Diffusion v2.1 Embedding | ### Describe the bug
I encountered a runtime error when trying to embed EasyNegative into Stable Diffusion v2.1. However, there was no error when using Stable Diffusion v1.5.
This is a link to the notebook:
https://colab.research.google.com/github/suzukimain/diffusers_in_Colab/blob/test1/Test.ipynb
### Reproducti... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-07-10T23:23:51Z | 2023-07-11T12:01:07Z | 2023-07-11T12:01:07Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | suzukimain | 131,413,573 | U_kgDOB9U2RQ | User | false |
huggingface/diffusers | 1,797,862,421 | I_kwDOHa8MBc5rKTQV | 4,031 | https://github.com/huggingface/diffusers/issues/4031 | https://api.github.com/repos/huggingface/diffusers/issues/4031 | Loading stand alone safety checker from local file fails | ### Describe the bug
I was using stand alone safety checker from local file and it worked like charm on previous versions(0.16.1 I guess ...) like simillar to https://github.com/huggingface/diffusers/issues/3892#issuecomment-1618740051
But on current version, loading fails...
Strange thing is that loading directly... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-07-11T00:34:53Z | 2023-08-02T07:55:49Z | 2023-07-11T09:07:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | moono | 6,761,398 | MDQ6VXNlcjY3NjEzOTg= | User | false |
huggingface/diffusers | 1,797,917,704 | I_kwDOHa8MBc5rKgwI | 4,033 | https://github.com/huggingface/diffusers/issues/4033 | https://api.github.com/repos/huggingface/diffusers/issues/4033 | SDXL 0.9 took 2x times longer to inference on Windows than on Ubuntu | ### Describe the bug
I'm working on a UI to do inference for SDXL 0.9.
https://github.com/lifeisboringsoprogramming/sd-webui-xldemo-txt2img
The inference time on Windows are 2x times longer than on Ubuntu
I found that the models downloaded on Windows are in .bin format
However the models downloaded on Ubuntu... | closed | completed | false | 14 | [
"bug",
"stale"
] | [] | 2023-07-11T01:40:03Z | 2023-08-19T15:03:16Z | 2023-08-19T15:03:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lifeisboringsoprogramming | 119,839,123 | U_kgDOBySZkw | User | false |
huggingface/diffusers | 1,798,245,350 | I_kwDOHa8MBc5rLwvm | 4,035 | https://github.com/huggingface/diffusers/issues/4035 | https://api.github.com/repos/huggingface/diffusers/issues/4035 | [SDXL0.9] Strange pixel in generated img. | Hi, thanks for your work on SDXL. When I use diffuser to inference SDXL, some pixels of the generated image appear strange colors, like the following image.
<img width="908" alt="image" src="https://github.com/huggingface/diffusers/assets/32625100/d05781df-5138-4f58-a941-bc5319ed4f58">
Here is the code I used. I dow... | closed | completed | false | 9 | [] | [] | 2023-07-11T06:57:56Z | 2023-07-17T12:39:30Z | 2023-07-12T03:51:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TinyTigerPan | 32,625,100 | MDQ6VXNlcjMyNjI1MTAw | User | false |
huggingface/diffusers | 1,798,451,940 | I_kwDOHa8MBc5rMjLk | 4,037 | https://github.com/huggingface/diffusers/issues/4037 | https://api.github.com/repos/huggingface/diffusers/issues/4037 | Validation Errors When Training ControlNet with FSDP | ### Describe the bug
Adding FSDP into training [the given ControlNet example training code](https://github.com/huggingface/diffusers/blob/main/examples/controlnet/train_controlnet.py) leads to an unexpected bug, with the following config:
```
compute_environment: LOCAL_MACHINE
distributed_type: FSDP
downcast_bf1... | closed | completed | false | 10 | [
"bug",
"stale"
] | [] | 2023-07-11T08:55:34Z | 2023-10-08T18:32:31Z | 2023-08-22T15:04:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | liming-ai | 73,068,772 | MDQ6VXNlcjczMDY4Nzcy | User | false |
huggingface/diffusers | 1,798,671,683 | I_kwDOHa8MBc5rNY1D | 4,039 | https://github.com/huggingface/diffusers/issues/4039 | https://api.github.com/repos/huggingface/diffusers/issues/4039 | diffusers/optimum interaction, neuron export broken for diffusion models, json serialization error | ### Describe the bug
diffusers.configuration_utils.register_to_config decorator side effect: object becomes non JSON serializable ?
This side effect was encountered while using optimum neuron with the latest diffusers library, more specifically while converting a stable diffusion model to the neuron format
Ru... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-07-11T10:57:20Z | 2023-07-11T16:21:10Z | 2023-07-11T16:21:10Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | oOraph | 13,552,058 | MDQ6VXNlcjEzNTUyMDU4 | User | false |
huggingface/diffusers | 1,798,809,504 | I_kwDOHa8MBc5rN6eg | 4,042 | https://github.com/huggingface/diffusers/issues/4042 | https://api.github.com/repos/huggingface/diffusers/issues/4042 | How to combine the reference-only with inpainting and depth control? | ### Model/Pipeline/Scheduler description
Hi, I recently want to combine the reference-only with image inpaint , with depth control to replace background for portrait images. However, I have no idea to build this pipeline as for there is no reference with inpaint pipeline example. Could you please help me to figure it... | closed | completed | false | 2 | [] | [] | 2023-07-11T12:17:24Z | 2023-07-14T06:12:29Z | 2023-07-14T06:12:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AmberCheng | 18,375,990 | MDQ6VXNlcjE4Mzc1OTkw | User | false |
huggingface/diffusers | 1,799,126,071 | I_kwDOHa8MBc5rPHw3 | 4,043 | https://github.com/huggingface/diffusers/issues/4043 | https://api.github.com/repos/huggingface/diffusers/issues/4043 | [SD-XL] Passing prompt_embeds/negative_prompt_embeds requires also passing pooled_prompt_embeds/negative_pooled_prompt_embeds | ### Describe the bug
When calling the `StableDiffusionXLPipeline` and passing the prompts as embeddings using the prompt_embeds/negative_prompt_embeds parameters, an error is generated requiring the pooled_prompt_embeds/negative_pooled_prompt_embeds parameters also to be passed.
There seems to be no documentation as... | closed | completed | false | 13 | [
"bug"
] | [] | 2023-07-11T14:52:16Z | 2025-12-05T12:00:47Z | 2023-08-04T09:18:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | m-a-r-v-i-n | 118,911,967 | U_kgDOBxZz3w | User | false |
huggingface/diffusers | 1,799,331,473 | I_kwDOHa8MBc5rP56R | 4,046 | https://github.com/huggingface/diffusers/issues/4046 | https://api.github.com/repos/huggingface/diffusers/issues/4046 | [BUG] train_text_to_image_lora.py not support Multi-nodes or Multi-gpus training. | In [`train_text_to_image_lora.py`](https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image_lora.py), I notice that the LORA parameters are extracted into an `AttnProcsLayers` class:
```
518 lora_layers = AttnProcsLayers(unet.attn_processors)
```
And it is only the `lora_laye... | closed | completed | false | 33 | [
"stale"
] | [] | 2023-07-11T16:39:32Z | 2024-11-23T19:39:50Z | 2023-11-09T15:07:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | WindVChen | 56,162,412 | MDQ6VXNlcjU2MTYyNDEy | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.