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 | 1,742,188,219 | I_kwDOHa8MBc5n1667 | 3,677 | https://github.com/huggingface/diffusers/issues/3677 | https://api.github.com/repos/huggingface/diffusers/issues/3677 | Textual Inversion / Dreambooth / ControlNet / LoRA for Kandinsky | Thanks to @yiyixuxu Kandinsky 2.1: https://github.com/huggingface/diffusers/pull/3308 is merged to diffusers.
Anybody interesting in giving dreamboothing/textual inversioning the model? | closed | not_planned | false | 5 | [
"enhancement",
"community-examples",
"Good second issue"
] | [] | 2023-06-05T17:06:00Z | 2024-11-24T03:28:54Z | 2024-11-24T03:28:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,742,486,726 | I_kwDOHa8MBc5n3DzG | 3,679 | https://github.com/huggingface/diffusers/issues/3679 | https://api.github.com/repos/huggingface/diffusers/issues/3679 | New watermark-free models for text2video (Modelscope): Potat1, Zeroscope Animov 0.1.1 and Animov 512x | ### Model/Pipeline/Scheduler description
Potat:
https://huggingface.co/camenduru/potat1
First Open-Source 1024x576 Text To Video Model - no watermark.
Zeroscope:
https://huggingface.co/cerspense/zeroscope_v1_320s
Modelscope without the watermark, trained in 320x320, with no skipped frames for less flicker. S... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-06-05T19:45:38Z | 2023-07-14T15:03:38Z | 2023-07-14T15:03:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,742,755,039 | I_kwDOHa8MBc5n4FTf | 3,682 | https://github.com/huggingface/diffusers/issues/3682 | https://api.github.com/repos/huggingface/diffusers/issues/3682 | Weights for loading loras | **Is your feature request related to a problem? Please describe.**
LoRAs when loaded/applied to a pipeline typically include a user provided weight that scales the weights in the LoRA. E.g. see Compel's LoraWeight class.
**Describe the solution you'd like**
Add a parameter to load_lora_weights that scales the w... | closed | completed | false | 2 | [] | [] | 2023-06-05T23:01:04Z | 2023-06-07T15:11:28Z | 2023-06-07T15:11:26Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pdoane | 1,621,950 | MDQ6VXNlcjE2MjE5NTA= | User | false |
huggingface/diffusers | 1,743,060,258 | I_kwDOHa8MBc5n5P0i | 3,683 | https://github.com/huggingface/diffusers/issues/3683 | https://api.github.com/repos/huggingface/diffusers/issues/3683 | Implement `save_pretrained()` for Multi-ControlNet | ### Describe the bug
If we load multiple controlnet onto `StableDiffusionControlNetPipeline`, it refuses to be saved using `pipe.save_pretrained` because it is not implemented.
I'd like to implement it myself if that's okay, but I am fairly new to the development of `diffusers`, so I would like to get some pointe... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-06-06T04:27:16Z | 2023-07-06T17:51:55Z | 2023-07-06T17:51:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DJBen | 2,078,653 | MDQ6VXNlcjIwNzg2NTM= | User | false |
huggingface/diffusers | 1,743,068,439 | I_kwDOHa8MBc5n5R0X | 3,684 | https://github.com/huggingface/diffusers/issues/3684 | https://api.github.com/repos/huggingface/diffusers/issues/3684 | How to finetune img2img model? | I want to train the text guided image2image model. Is there anyone know how to revise the text2img code？
Many thanks.
| closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-06T04:39:01Z | 2023-07-14T15:03:37Z | 2023-07-14T15:03:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yftongbupt | 74,176,966 | MDQ6VXNlcjc0MTc2OTY2 | User | false |
huggingface/diffusers | 1,743,579,692 | I_kwDOHa8MBc5n7Oos | 3,688 | https://github.com/huggingface/diffusers/issues/3688 | https://api.github.com/repos/huggingface/diffusers/issues/3688 | [LoRA] Ensure text encoder LoRA also uses PT 2.0 SPDA | Similar to https://github.com/huggingface/diffusers/pull/3594, the text encoder LoRA should also use PT 2.0 SPDA.
Sorry, this slipped my mind in https://github.com/huggingface/diffusers/pull/3594. | closed | completed | false | 0 | [] | [
"sayakpaul"
] | 2023-06-06T10:34:08Z | 2023-06-06T20:20:55Z | 2023-06-06T20:20:55Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,743,619,007 | I_kwDOHa8MBc5n7YO_ | 3,689 | https://github.com/huggingface/diffusers/issues/3689 | https://api.github.com/repos/huggingface/diffusers/issues/3689 | [LORA] How to unload lora weight after load_lora_weights | **Is your feature request related to a problem? Please describe.**
I use pipe.load_lora_weights to load a lora weight, but don't know how to unload it.
I tried pipe.unet.set_default_attn_processor(), but it didn't work.
**Describe the solution you'd like**
I need a clear function to unload lora weight.
| closed | completed | false | 19 | [] | [] | 2023-06-06T10:58:09Z | 2025-01-22T10:34:49Z | 2023-06-11T07:01:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yuananf | 4,101,738 | MDQ6VXNlcjQxMDE3Mzg= | User | false |
huggingface/diffusers | 1,362,938,034 | I_kwDOHa8MBc5RPMiy | 369 | https://github.com/huggingface/diffusers/issues/369 | https://api.github.com/repos/huggingface/diffusers/issues/369 | Represent learnt concept in textual inversion with more than one token | ### Describe the bug
As we discuss in #266
> The original textual inversion support [using more than one vector](https://github.com/rinongal/textual_inversion/blob/main/ldm/modules/embedding_manager.py#L39) to represent the learnt concept.
For the current implementation, if we just extend the learned vocab and CL... | closed | completed | false | 11 | [
"enhancement"
] | [
"patil-suraj"
] | 2022-09-06T08:56:18Z | 2024-11-21T11:33:16Z | 2024-11-21T11:33:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luvata | 17,178,612 | MDQ6VXNlcjE3MTc4NjEy | User | false |
huggingface/diffusers | 1,744,432,439 | I_kwDOHa8MBc5n-e03 | 3,693 | https://github.com/huggingface/diffusers/issues/3693 | https://api.github.com/repos/huggingface/diffusers/issues/3693 | Issues in the docstrings for Kandinsky pipeline(s) | ### Describe the bug
I noticed a few issues with the docstrings in Kandinsky pipelines:
1. Wrong model name in the [pipeline_kandinsky.py example](https://github.com/huggingface/diffusers/blob/8669e8313dfedd5da1fd84e8abab92eda82623c0/src/diffusers/pipelines/kandinsky/pipeline_kandinsky.py#LL44C48-L44C48):
```py
p... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-06-06T19:08:23Z | 2023-06-07T03:38:39Z | 2023-06-06T21:18:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | freespirit | 222,401 | MDQ6VXNlcjIyMjQwMQ== | User | false |
huggingface/diffusers | 1,744,451,130 | I_kwDOHa8MBc5n-jY6 | 3,694 | https://github.com/huggingface/diffusers/issues/3694 | https://api.github.com/repos/huggingface/diffusers/issues/3694 | Unable to locate model_index.json and config.json while running the text_to_image/train_text_to_image.py finetuning pokemon example. | ### Describe the bug
Using a Google Colab notebook I ran the steps of the text_to_image fine-tuning example using the pokemon data provided. I successfully fine-tuned the model for 500 steps and see the checkpoint-500 output in my directory. When I attempt to follow the readme and run:
```
from diffusers import Sta... | closed | completed | false | 9 | [
"bug"
] | [] | 2023-06-06T19:23:01Z | 2024-02-11T16:15:48Z | 2023-06-16T06:26:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | benjaminwfriedman | 11,875,832 | MDQ6VXNlcjExODc1ODMy | User | false |
huggingface/diffusers | 1,744,929,326 | I_kwDOHa8MBc5oAYIu | 3,696 | https://github.com/huggingface/diffusers/issues/3696 | https://api.github.com/repos/huggingface/diffusers/issues/3696 | loading state_dict for UNet2DConditionModel when convert safetensors to diffusers model | `RuntimeError: Error(s) in loading state_dict for UNet2DConditionModel:
size mismatch for conv_in.weight: copying a param with shape torch.Size([320, 9, 3, 3]) from checkpoint, the shape in current model is torch.Size([320, 4, 3, 3]).`
when I try to convert the safetensors model from civitai (https://civitai.com/a... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-06-07T02:44:30Z | 2023-07-15T15:04:40Z | 2023-07-15T15:04:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FrancisDacian | 24,215,706 | MDQ6VXNlcjI0MjE1NzA2 | User | false |
huggingface/diffusers | 1,744,984,640 | I_kwDOHa8MBc5oAlpA | 3,698 | https://github.com/huggingface/diffusers/issues/3698 | https://api.github.com/repos/huggingface/diffusers/issues/3698 | convert the inpainting model to diffusers | **Is your feature request related to a problem? Please describe.**
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
**Describe the solution you'd like**
A clear and concise description of what you want to happen.
**Describe alternatives you've considered**
A clear an... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-07T03:59:00Z | 2023-07-15T15:04:39Z | 2023-07-15T15:04:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FrancisDacian | 24,215,706 | MDQ6VXNlcjI0MjE1NzA2 | User | false |
huggingface/diffusers | 1,745,042,786 | I_kwDOHa8MBc5oAz1i | 3,699 | https://github.com/huggingface/diffusers/issues/3699 | https://api.github.com/repos/huggingface/diffusers/issues/3699 | OSX 13.4 M2 AssertionError: Torch not compiled with CUDA enabled | ### Describe the bug
I am trying to run `img2img` A.I. on my M2 MBP 2023. The terminal raises up an error regarding CUDA is not enabled on my system.
### Reproduction
Run this script.
```python
import cv2
from PIL import Image
from diffusers import StableDiffusionControlNetPipeline, ControlNetModel, UniPCMultist... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-06-07T05:03:39Z | 2023-06-07T08:49:50Z | 2023-06-07T08:49:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | elcolie | 18,206,728 | MDQ6VXNlcjE4MjA2NzI4 | User | false |
huggingface/diffusers | 1,363,044,648 | I_kwDOHa8MBc5RPmko | 370 | https://github.com/huggingface/diffusers/issues/370 | https://api.github.com/repos/huggingface/diffusers/issues/370 | Lack of functionality in stable diffusion inpainting | The img2img can handle init_image as FloatTensor and PIL image with
```
if not isinstance(init_image, torch.FloatTensor):
init_image = preprocess(init_image)
```
But inpainting cannot work like this with init_image and mask_image. It's only working with FloatTensor. Although the typing states that it can work w... | closed | completed | false | 3 | [] | [
"patil-suraj"
] | 2022-09-06T10:11:30Z | 2022-09-16T15:35:42Z | 2022-09-16T15:35:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Inkorak | 52,286,717 | MDQ6VXNlcjUyMjg2NzE3 | User | false |
huggingface/diffusers | 1,745,111,918 | I_kwDOHa8MBc5oBEtu | 3,700 | https://github.com/huggingface/diffusers/issues/3700 | https://api.github.com/repos/huggingface/diffusers/issues/3700 | Not able to use min_snr argument in train_text_to_image_lora.py | ### Describe the bug
I am trying to finetune train_text_to_image_lora.py on custom dataset , I am trying to use min_snr also for training and setting it to default value of 5.0 , however while running its giving some error .
### Reproduction
```
export MODEL_NAME="CompVis/stable-diffusion-v1-4"
export TRAI... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-06-07T06:10:25Z | 2023-06-08T10:30:41Z | 2023-06-08T10:30:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | atharmzaalo2023 | 134,913,498 | U_kgDOCAqd2g | User | false |
huggingface/diffusers | 1,745,308,000 | I_kwDOHa8MBc5oB0lg | 3,701 | https://github.com/huggingface/diffusers/issues/3701 | https://api.github.com/repos/huggingface/diffusers/issues/3701 | Major update on Controlnet v1.1, Feature Reference-only Control | sd-webui-controlnet now supports reference-only control, which does not require any control models. it guides diffusion directly using images as references. It would be nice to have it on diffusers.
https://github.com/Mikubill/sd-webui-controlnet/pull/1235/files
I think the method for reference-only is continuall... | closed | completed | false | 9 | [] | [] | 2023-06-07T08:05:26Z | 2023-08-01T07:32:19Z | 2023-06-08T02:38:17Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jinwonkim93 | 26,476,095 | MDQ6VXNlcjI2NDc2MDk1 | User | false |
huggingface/diffusers | 1,745,390,847 | I_kwDOHa8MBc5oCIz_ | 3,702 | https://github.com/huggingface/diffusers/issues/3702 | https://api.github.com/repos/huggingface/diffusers/issues/3702 | How to merge controlnet with sd into one model? | Does unet can be merged with controlnet? | closed | completed | false | 4 | [
"stale"
] | [] | 2023-06-07T08:54:27Z | 2023-07-15T15:04:38Z | 2023-07-15T15:04:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yftongbupt | 74,176,966 | MDQ6VXNlcjc0MTc2OTY2 | User | false |
huggingface/diffusers | 1,745,638,863 | I_kwDOHa8MBc5oDFXP | 3,703 | https://github.com/huggingface/diffusers/issues/3703 | https://api.github.com/repos/huggingface/diffusers/issues/3703 | ValueError: invalid version number '0.17.0rc0' | ### Describe the bug
Loading a diffusion model broke suddenly.
I've already tried pip reinstalling diffusers (from pip as well as this github repo)
### Reproduction
from diffusers import DiffusionPipeline
import torch
model_id = "runwayml/stable-diffusion-v1-5"
ref_pipe = DiffusionPipeline.from_pretrained(... | closed | completed | false | 29 | [
"bug",
"stale"
] | [] | 2023-06-07T11:02:57Z | 2023-07-15T15:04:37Z | 2023-07-15T15:04:37Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nagolinc | 7,775,917 | MDQ6VXNlcjc3NzU5MTc= | User | false |
huggingface/diffusers | 1,745,651,159 | I_kwDOHa8MBc5oDIXX | 3,704 | https://github.com/huggingface/diffusers/issues/3704 | https://api.github.com/repos/huggingface/diffusers/issues/3704 | Load StableDiffusionInpaintPipeline using from_ckpt | Hi, I tried to load an inpaint pipeline but i got an error that says: 'StableDiffusionInpaintPipeline' has no attribute 'from_ckpt' .
if i am understanding #3533 correctly: we cannot load an inpaint pipeline using from_ckpt. we have to load txt2img pipeline with from_ckpt and load a inpaint pipeline using txt2img pipe... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-07T11:10:25Z | 2023-07-15T15:04:35Z | 2023-07-15T15:04:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aycaecemgul | 50,622,243 | MDQ6VXNlcjUwNjIyMjQz | User | false |
huggingface/diffusers | 1,745,651,995 | I_kwDOHa8MBc5oDIkb | 3,705 | https://github.com/huggingface/diffusers/issues/3705 | https://api.github.com/repos/huggingface/diffusers/issues/3705 | The model never saves to HF | ### Describe the bug
When running the full https://huggingface.co/docs/diffusers/main/en/tutorials/basic_training training, all that gets saved are the images, the logs.
No `unet` and no `scheduler` is saved. Basically, the model is missing, which is the whole point of the exercise.
### Reproduction
Just run th... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-06-07T11:11:01Z | 2023-07-24T15:03:49Z | 2023-07-24T15:03:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | smileBeda | 11,800,236 | MDQ6VXNlcjExODAwMjM2 | User | false |
huggingface/diffusers | 1,745,806,168 | I_kwDOHa8MBc5oDuNY | 3,706 | https://github.com/huggingface/diffusers/issues/3706 | https://api.github.com/repos/huggingface/diffusers/issues/3706 | New Discrete Diffusion Model: Wuerstchen | ### Model/Pipeline/Scheduler description
The amazing @dome272 published a new diffusion model, called Wuerstchen: https://github.com/dome272/Wuerstchen
I think it would be a very nice addition to the library given it's superior speed.
Anybody interested in picking this one up?
### Open source status
- [X... | closed | completed | false | 18 | [
"New pipeline/model",
"Good second issue"
] | [] | 2023-06-07T12:37:42Z | 2023-11-28T11:20:52Z | 2023-11-28T11:20:52Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,747,345,843 | I_kwDOHa8MBc5oJmGz | 3,717 | https://github.com/huggingface/diffusers/issues/3717 | https://api.github.com/repos/huggingface/diffusers/issues/3717 | OOM when use multi-gpu by 3090Ti | ### Describe the bug
Is there any attempt to finetune a stable diffusion model using multi 3090ti with 24Mem by running train_text_to_image.py?
I tried lots of accurate configs but OOM still occurs. I use another card and found using only one gpu takes 25G while use multiple GPUs takes about 35G/card. Is there an... | closed | completed | false | 12 | [
"bug",
"stale"
] | [] | 2023-06-08T08:14:57Z | 2024-07-18T04:18:11Z | 2023-07-19T15:03:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wangmiaowei | 27,915,265 | MDQ6VXNlcjI3OTE1MjY1 | User | false |
huggingface/diffusers | 1,747,346,066 | I_kwDOHa8MBc5oJmKS | 3,718 | https://github.com/huggingface/diffusers/issues/3718 | https://api.github.com/repos/huggingface/diffusers/issues/3718 | StableDiffusionPipeline.from_ckpt() cannot recognition remote(Hugging Face) model (SD Checkpoint) | ### Describe the bug
my code is:
```python
pipe = StableDiffusionPipeline.from_ckpt(
"https://huggingface.co/andite/anything-v4.0/blob/main/anything-v4.0-pruned-fp16.ckpt",
# torch_dtype=torch.float16
)
```
just like [Diffusers Doc](https://huggingface.co/docs/diffusers/v0.16.0/en/api/pipelines... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-06-08T08:15:06Z | 2023-07-16T15:03:24Z | 2023-07-16T15:03:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AndrewHan0457 | 44,436,825 | MDQ6VXNlcjQ0NDM2ODI1 | User | false |
huggingface/diffusers | 1,747,396,273 | I_kwDOHa8MBc5oJyax | 3,719 | https://github.com/huggingface/diffusers/issues/3719 | https://api.github.com/repos/huggingface/diffusers/issues/3719 | Implement ViCo: Detail-Preserving Visual Condition for Personalized Text-to-Image Generation for diffusers | I would like to use new advancements for personalisation
To the best of my knowledge [ViCo](https://github.com/haoosz/ViCo) is the first available open source implementation of similar ideas that also use an image encoding for conditioning diffusion models.
https://github.com/haoosz/ViCo
| closed | completed | false | 15 | [
"community-examples",
"Good second issue"
] | [] | 2023-06-08T08:46:40Z | 2024-11-24T03:30:04Z | 2024-11-24T03:30:04Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | okaris | 1,448,702 | MDQ6VXNlcjE0NDg3MDI= | User | false |
huggingface/diffusers | 1,363,229,673 | I_kwDOHa8MBc5RQTvp | 372 | https://github.com/huggingface/diffusers/issues/372 | https://api.github.com/repos/huggingface/diffusers/issues/372 | MPS: models require an initial pass for reproducibility | I have no idea what's causing this, I suspect something related to RNGs, as we know they don't work properly in `mps` yet.
This is a pending task from our initial `mps` support (#355) that we need to investigate. For now, we have adopted two workarounds:
- Perform an extra pass as appropriate so tests pass.
- Reco... | closed | completed | false | 22 | [
"stale"
] | [
"pcuenca"
] | 2022-09-06T12:44:07Z | 2023-03-23T12:37:31Z | 2023-03-10T15:05:11Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pcuenca | 1,177,582 | MDQ6VXNlcjExNzc1ODI= | User | false |
huggingface/diffusers | 1,748,604,216 | I_kwDOHa8MBc5oOZU4 | 3,722 | https://github.com/huggingface/diffusers/issues/3722 | https://api.github.com/repos/huggingface/diffusers/issues/3722 | StableDiffusionControlNetPipeline.from_ckpt() does not exist |
StableDiffusionControlNetPipeline doesn't appear to inherit from FromCkptMixin so there is no ability to load safetensors like we can with StableDiffusionPipeline.from_ckpt(). Adding the class doesn't fix it as there is another failure in the from_ckpt code.
It would be cool to support from_ckpt() in StableDiffus... | closed | completed | false | 14 | [
"stale"
] | [] | 2023-06-08T20:39:46Z | 2023-10-18T15:17:00Z | 2023-10-18T15:17:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tripzero | 1,034,370 | MDQ6VXNlcjEwMzQzNzA= | User | false |
huggingface/diffusers | 1,749,021,197 | I_kwDOHa8MBc5oP_IN | 3,724 | https://github.com/huggingface/diffusers/issues/3724 | https://api.github.com/repos/huggingface/diffusers/issues/3724 | Merit in adding a Watermarking module to diffusion models ? | **Is your feature request related to a problem? Please describe.**
Right now, there is no way to watermark images generated by models in diffusers out of the box.
**Describe the solution you'd like**
The recent paper, [Tree ring watermarking](https://arxiv.org/abs/2305.20030) describes a way to watermark an image ... | closed | completed | false | 8 | [
"stale"
] | [] | 2023-06-09T03:01:42Z | 2023-11-13T08:15:25Z | 2023-07-24T15:03:46Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Abhinay1997 | 24,771,261 | MDQ6VXNlcjI0NzcxMjYx | User | false |
huggingface/diffusers | 1,749,079,532 | I_kwDOHa8MBc5oQNXs | 3,725 | https://github.com/huggingface/diffusers/issues/3725 | https://api.github.com/repos/huggingface/diffusers/issues/3725 | failed to use the feature of supporting for A1111 LoRA | I am glad to see the diffusers add the supporting for A1111 LoRA. However, i failed to ues this feature after I update the diffusers.
It did not exporting error as before, but the lora have no effect in the generated images.
I use it as following:
```
pipe = StableDiffusionPipeline.from_pretrained(repo_id, torch_... | closed | completed | false | 21 | [] | [] | 2023-06-09T03:58:34Z | 2023-07-25T22:24:21Z | 2023-07-25T22:24:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | icech | 16,433,164 | MDQ6VXNlcjE2NDMzMTY0 | User | false |
huggingface/diffusers | 1,749,081,742 | I_kwDOHa8MBc5oQN6O | 3,726 | https://github.com/huggingface/diffusers/issues/3726 | https://api.github.com/repos/huggingface/diffusers/issues/3726 | VAE training sample script | I believe the current lack of easy access to VAE training is stopping diffusion models from disrupting even more industries.
I'm talking about consistent details on things that are less represented in the original training data. 64x64 res can only carry so much detail. Very often I get good result from latent space... | closed | completed | false | 99 | [
"community-examples",
"Good second issue"
] | [] | 2023-06-09T04:01:20Z | 2025-05-14T11:16:26Z | 2024-11-25T13:11:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhuliyi0 | 48,817,897 | MDQ6VXNlcjQ4ODE3ODk3 | User | false |
huggingface/diffusers | 1,749,424,706 | I_kwDOHa8MBc5oRhpC | 3,728 | https://github.com/huggingface/diffusers/issues/3728 | https://api.github.com/repos/huggingface/diffusers/issues/3728 | [New Pipeline]: Shifted Diffusion for Text-to-image Generation | ### Model/Pipeline/Scheduler description
Github: https://github.com/drboog/Shifted_Diffusion
Paper: https://arxiv.org/abs/2211.15388.pdf
### Open source status
- [X] The model implementation is available
- [X] The model weights are available (Only relevant if addition is not a scheduler).
### Provide useful links ... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-06-09T08:39:56Z | 2023-07-17T15:03:27Z | 2023-07-17T15:03:27Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kadirnar | 36,204,372 | MDQ6VXNlcjM2MjA0Mzcy | User | false |
huggingface/diffusers | 1,749,535,439 | I_kwDOHa8MBc5oR8rP | 3,729 | https://github.com/huggingface/diffusers/issues/3729 | https://api.github.com/repos/huggingface/diffusers/issues/3729 | StableDiffusionPipeline.from_ckpt() is unable to load checkpoints offline | When the network connection is disabled, StableDiffusionPipeline.from_ckpt() is unable to load checkpoints from local_files_only because it tries to download the .yaml configuration from the GitHub repository.
This problem can be solved by adding this line to from_ckpt() in [loaders.py](https://github.com/huggingfac... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-09T09:38:42Z | 2023-07-17T15:03:25Z | 2023-07-17T15:03:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nimadez | 111,365,948 | U_kgDOBqNPPA | User | false |
huggingface/diffusers | 1,749,730,164 | I_kwDOHa8MBc5oSsN0 | 3,730 | https://github.com/huggingface/diffusers/issues/3730 | https://api.github.com/repos/huggingface/diffusers/issues/3730 | Wrong time step in inpainting blending | I might be wrong, but at the first glance, `init_latents_proper` should be diffused to step `t-1` instead of `t`. (`1` here means generalized step size). Can someone explain the current implementation for me? Thanks
https://github.com/huggingface/diffusers/blob/c42f6ee43e0408c5fe8a1d3dc3cdeb9eb3a02fa6/src/diffusers/... | closed | completed | false | 2 | [] | [] | 2023-06-09T11:38:12Z | 2023-06-12T12:16:07Z | 2023-06-12T12:16:07Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ernestchu | 51,432,514 | MDQ6VXNlcjUxNDMyNTE0 | User | false |
huggingface/diffusers | 1,749,749,532 | I_kwDOHa8MBc5oSw8c | 3,731 | https://github.com/huggingface/diffusers/issues/3731 | https://api.github.com/repos/huggingface/diffusers/issues/3731 | ModuleNotFoundError: No module named 'models.unet_3d_condition' | ### Describe the bug
Hi, I'm receiving this error:
```
/home/anton/.local/lib/python3.8/site-packages/diffusers/pipelines/pipeline_utils.py:318 in │
│ get_class_obj_and_candidates │
│ ... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-06-09T11:53:30Z | 2023-06-09T12:06:13Z | 2023-06-09T12:06:13Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | NotNANtoN | 19,983,153 | MDQ6VXNlcjE5OTgzMTUz | User | false |
huggingface/diffusers | 1,749,880,847 | I_kwDOHa8MBc5oTRAP | 3,732 | https://github.com/huggingface/diffusers/issues/3732 | https://api.github.com/repos/huggingface/diffusers/issues/3732 | How to use local safetensors model for inference? | I've known that from_pretraind function is able to load online model. But I have a local safetensors stable diffusion model (xxx.safetensors). How to use this file to generate images without AUTOMATIC1111 webui? | closed | completed | false | 4 | [
"stale"
] | [] | 2023-06-09T13:13:35Z | 2023-07-24T15:03:44Z | 2023-07-24T15:03:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Eugeoter | 35,489,075 | MDQ6VXNlcjM1NDg5MDc1 | User | false |
huggingface/diffusers | 1,750,915,570 | I_kwDOHa8MBc5oXNny | 3,736 | https://github.com/huggingface/diffusers/issues/3736 | https://api.github.com/repos/huggingface/diffusers/issues/3736 | Hyper pigmentation since Diffusers 15 (breaks on certain VAEs?) | ### Describe the bug
(edits: added more affected checkpoint examples)
I was hoping a savant could take a look at this picture and in one second describe what we're doing wrong
warning, mild cartoon nudity:
https://imgur.com/a/Y672o4R
Literally, if we run the same prompt in Diffusers 15, it works, on an... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-06-10T10:39:18Z | 2023-08-29T15:04:24Z | 2023-08-29T15:04:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CustodialMatter | 136,163,227 | U_kgDOCB2vmw | User | false |
huggingface/diffusers | 1,751,057,671 | I_kwDOHa8MBc5oXwUH | 3,737 | https://github.com/huggingface/diffusers/issues/3737 | https://api.github.com/repos/huggingface/diffusers/issues/3737 | StableDiffusionControlNetImg2ImgPipeline is missing from library | ### Describe the bug
The `StableDiffusionControlNetImg2ImgPipeline` seems to be missing from the library when I try to import it.
I get this error:
`ImportError: cannot import name 'StableDiffusionControlNetImg2ImgPipeline' from 'diffusers'
(/usr/local/lib/python3.10/dist-packages/diffusers/__init__.py)`
To ... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-06-10T17:51:06Z | 2023-06-11T12:39:51Z | 2023-06-11T12:39:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | redromnon | 74,495,920 | MDQ6VXNlcjc0NDk1OTIw | User | false |
huggingface/diffusers | 1,751,097,016 | I_kwDOHa8MBc5oX564 | 3,738 | https://github.com/huggingface/diffusers/issues/3738 | https://api.github.com/repos/huggingface/diffusers/issues/3738 | Issue during training diffusion model with custom dataset | While training the CompVis/stable-diffusion-v1-4 model I get an issue of a byte error with the images
`UnidentifiedImageError: cannot identify image file <_io.BytesIO object at 0x7f57670893a0>`
I get the same issue with the runwayml/stable-diffusion-v1-5 model1
There where no issues when I pushed my dataset t... | closed | completed | false | 10 | [] | [] | 2023-06-10T19:49:34Z | 2023-06-22T11:39:01Z | 2023-06-22T11:39:01Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | neramas1221 | 17,319,053 | MDQ6VXNlcjE3MzE5MDUz | User | false |
huggingface/diffusers | 1,363,529,829 | I_kwDOHa8MBc5RRdBl | 374 | https://github.com/huggingface/diffusers/issues/374 | https://api.github.com/repos/huggingface/diffusers/issues/374 | [Community] cancelable and asynchronous pipelines | In [#developers-corner](https://discord.com/channels/1002292111942635562/1011228667659178055/1016173296930271282) we have a request for cancellation support in the pipeline API.
That could also be something to consider in the context of a more asynchronous API in general; some applications want to avoid blocking the... | closed | completed | false | 21 | [
"good first issue",
"Good Example PR"
] | [] | 2022-09-06T16:20:58Z | 2023-07-25T09:01:14Z | 2023-02-23T23:05:35Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | keturn | 83,819 | MDQ6VXNlcjgzODE5 | User | false |
huggingface/diffusers | 1,751,458,124 | I_kwDOHa8MBc5oZSFM | 3,740 | https://github.com/huggingface/diffusers/issues/3740 | https://api.github.com/repos/huggingface/diffusers/issues/3740 | Legacy Attention Block conversion doesn't work with device_map "sequential" from accelerate | ### Describe the bug
As far as I understand, checkpoints from before [this refactor of the Attention module](https://github.com/huggingface/diffusers/pull/3387) have to be converted with [this function](https://github.com/huggingface/diffusers/blob/abbfe4b5b7a9508169e69476a951a1a6dd86c927/src/diffusers/models/modeling... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-06-11T14:45:00Z | 2023-06-16T17:39:31Z | 2023-06-16T17:39:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | stefan-canva | 90,706,016 | MDQ6VXNlcjkwNzA2MDE2 | User | false |
huggingface/diffusers | 1,751,521,184 | I_kwDOHa8MBc5oZheg | 3,741 | https://github.com/huggingface/diffusers/issues/3741 | https://api.github.com/repos/huggingface/diffusers/issues/3741 | Wrong link about LoRA in the release page of v0.17.0 | Notice the following link doesn't point to a PR:
`Thanks to @takuma104 for contributing this feature via [this PR](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/lora_non_lora_comparison.png).`
https://github.com/huggingface/diffusers/releases/tag/v0.17.0
![image](https://g... | closed | completed | false | 4 | [] | [] | 2023-06-11T17:13:54Z | 2023-06-13T04:13:31Z | 2023-06-12T17:03:50Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | offchan42 | 15,215,732 | MDQ6VXNlcjE1MjE1NzMy | User | false |
huggingface/diffusers | 1,752,062,241 | I_kwDOHa8MBc5oblkh | 3,743 | https://github.com/huggingface/diffusers/issues/3743 | https://api.github.com/repos/huggingface/diffusers/issues/3743 | How I can setup AUTO1111 formatted LoRA alpha ? | from source code I found network_alpha , but how I can setup network_alpha ?
# Convert kohya-ss Style LoRA attn procs to diffusers attn procs
network_alpha = None
if all((k.startswith("lora_te_") or k.startswith("lora_unet_")) for k in state_dict.keys()):
state_dict, network_alpha = self._conv... | closed | completed | false | 12 | [
"stale"
] | [] | 2023-06-12T07:10:50Z | 2023-08-23T15:04:35Z | 2023-08-23T15:04:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | stevensu1977 | 6,694,421 | MDQ6VXNlcjY2OTQ0MjE= | User | false |
huggingface/diffusers | 1,752,316,811 | I_kwDOHa8MBc5ocjuL | 3,748 | https://github.com/huggingface/diffusers/issues/3748 | https://api.github.com/repos/huggingface/diffusers/issues/3748 | Lora QueryKey float16/float32 mismatch when loading safetensor from CivitAI | ### Describe the bug
I'm trying to load LORAs following the [docs](https://huggingface.co/docs/diffusers/main/en/training/lora#supporting-a1111-themed-lora-checkpoints-from-diffusers). I've downloaded a couple LORAs so far, and ensured that they are marked as `fp16` on civitai.
This throws the following error:
... | closed | completed | false | 8 | [
"bug"
] | [] | 2023-06-12T09:28:47Z | 2023-06-14T06:05:30Z | 2023-06-14T06:05:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | codejudas | 6,273,743 | MDQ6VXNlcjYyNzM3NDM= | User | false |
huggingface/diffusers | 1,752,642,094 | I_kwDOHa8MBc5odzIu | 3,750 | https://github.com/huggingface/diffusers/issues/3750 | https://api.github.com/repos/huggingface/diffusers/issues/3750 | `StableDiffusionKDiffusionPipeline` does not support LoRA loading | LoRA loading is supported in the `StableDiffusionPipeline` thanks to the last release (v0.17.0), however it's not yet fully supported in other pipelines like `StableDiffusionKDiffusionPipeline`, this is probably under work anyway given [this PR](https://github.com/huggingface/diffusers/pull/3626/) , but I'm adding a PR... | closed | completed | false | 1 | [] | [] | 2023-06-12T12:16:51Z | 2023-06-15T14:26:17Z | 2023-06-15T14:26:17Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tripathiarpan20 | 42,506,819 | MDQ6VXNlcjQyNTA2ODE5 | User | false |
huggingface/diffusers | 1,752,871,225 | I_kwDOHa8MBc5oerE5 | 3,752 | https://github.com/huggingface/diffusers/issues/3752 | https://api.github.com/repos/huggingface/diffusers/issues/3752 | error during inference with the finetuned deepfloyd if using dreambooth | I am trying to finetune **deepfloyd if (stage 1)** using train_dreambooth.py and an error was encountered during inference with the trained model following
```
from diffusers import DiffusionPipeline
import torch
stage_1 = DiffusionPipeline.from_pretrained("xxxx", torch_dtype=torch.float16)
stage_1.enable_model... | closed | completed | false | 1 | [] | [] | 2023-06-12T14:16:34Z | 2023-06-12T14:35:38Z | 2023-06-12T14:35:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | garychan22 | 108,175,311 | U_kgDOBnKfzw | User | false |
huggingface/diffusers | 1,752,882,640 | I_kwDOHa8MBc5oet3Q | 3,753 | https://github.com/huggingface/diffusers/issues/3753 | https://api.github.com/repos/huggingface/diffusers/issues/3753 | Reshape in Unet_3D seems to be wrong | ### Describe the bug
Looking at the [reshape operation](https://github.com/huggingface/diffusers/blob/38adcd21bd9cc69f84710deb161c578b6478944f/src/diffusers/models/unet_3d_condition.py#L466) of the input sample within `unet_3d_condition.py,` it looks as if it is not correct.
### Reproduction
``` python
impor... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-06-12T14:22:02Z | 2023-07-20T15:03:31Z | 2023-07-20T15:03:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rob-hen | 115,473,528 | U_kgDOBuH8eA | User | false |
huggingface/diffusers | 1,753,456,994 | I_kwDOHa8MBc5og6Fi | 3,758 | https://github.com/huggingface/diffusers/issues/3758 | https://api.github.com/repos/huggingface/diffusers/issues/3758 | controlnet_guidance_start and controlnet_guidance_end is missing in StableDiffusionControlNetImg2ImgPipeline | In the old community controlnet img2img pipeline, there are controlnet_guidance_start and controlnet_guidance_end setting
https://github.com/huggingface/diffusers/blob/34d14d78489f82b4abebdb62d90545312171c033/examples/community/stable_diffusion_controlnet_img2img.py#L911-L920
However, controlnet_guidance_start... | closed | completed | false | 4 | [] | [] | 2023-06-12T19:46:34Z | 2023-06-14T17:54:32Z | 2023-06-14T17:54:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Bochu3 | 121,131,123 | U_kgDOBzhQcw | User | false |
huggingface/diffusers | 1,753,635,200 | I_kwDOHa8MBc5ohlmA | 3,759 | https://github.com/huggingface/diffusers/issues/3759 | https://api.github.com/repos/huggingface/diffusers/issues/3759 | [New feature] Add ControlNet `conditioning_scale` scheduling to StableDiffusionControlNetPipeline | **Is your feature request related to a problem? Please describe.**
In many use case a fixed ControlNet `conditioning_scale` is not appropriate. For example in [this](https://huggingface.co/spaces/blanchon/qrcode-diffusion) project, I'm trying to constraint a stable diffusion model to overlay a QR code using ControlN... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-06-12T21:37:59Z | 2024-08-26T01:44:52Z | 2023-07-21T15:03:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | julien-blanchon | 11,278,197 | MDQ6VXNlcjExMjc4MTk3 | User | false |
huggingface/diffusers | 1,753,724,806 | I_kwDOHa8MBc5oh7eG | 3,761 | https://github.com/huggingface/diffusers/issues/3761 | https://api.github.com/repos/huggingface/diffusers/issues/3761 | Can't load 2 community pipelines in same process | ### Describe the bug
Loading 2 community pipelines leads to it confusing the 2 classes.
Reproducible steps below
### Reproduction
I'm trying to avoid loading community pipelines through the hub so I'm only loading community pipelines locally now.
So I'm doing something like:
```
mkdir -p /usr/local/src/d... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-06-12T23:01:28Z | 2023-07-21T15:03:40Z | 2023-07-21T15:03:40Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | reimager | 20,918,094 | MDQ6VXNlcjIwOTE4MDk0 | User | false |
huggingface/diffusers | 1,753,912,531 | I_kwDOHa8MBc5oipTT | 3,762 | https://github.com/huggingface/diffusers/issues/3762 | https://api.github.com/repos/huggingface/diffusers/issues/3762 | TypeError: mul got incompatible shapes for broadcasting, FlaxStableDiffusionImg2ImgPipeline | ### Describe the bug
Certain image dimensions for the seed image generate "TypeError: mul got incompatible shapes for broadcasting" when using FlaxStableDiffusionImg2ImgPipeline
Does not seem related to divisibility by 8 or failing to use 1:1 or 1.5 aspect ratio or power of two. The example in the source code uses ... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-06-13T02:24:23Z | 2023-08-19T15:03:29Z | 2023-08-19T15:03:29Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Teriks | 14,919,098 | MDQ6VXNlcjE0OTE5MDk4 | User | false |
huggingface/diffusers | 1,753,937,416 | I_kwDOHa8MBc5oivYI | 3,763 | https://github.com/huggingface/diffusers/issues/3763 | https://api.github.com/repos/huggingface/diffusers/issues/3763 | Postprocess for better result | Can diffusers support postprocess like https://github.com/jquesnelle/txt2imghd/tree/master in the future?
First generating an image from a prompt, upscaling it, and then running img2img on smaller pieces of the upscaled image, and blending the result back into the original image. | closed | completed | false | 2 | [
"community-examples",
"stale"
] | [] | 2023-06-13T02:55:57Z | 2023-07-21T15:03:38Z | 2023-07-21T15:03:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lzj9072 | 26,057,369 | MDQ6VXNlcjI2MDU3MzY5 | User | false |
huggingface/diffusers | 1,753,989,543 | I_kwDOHa8MBc5oi8Gn | 3,764 | https://github.com/huggingface/diffusers/issues/3764 | https://api.github.com/repos/huggingface/diffusers/issues/3764 | Using torch.compile as instructed in 0.17 release note results in BackendCompilerFailed: cannot schedule new futures after interpreter shutdown | ### Describe the bug
Strange Numba errors print out upon running inference if I've compiled the models as described in 0.17
### Reproduction
Unfortunately it is quite a big codebase and I have not reproduced it yet in a small minimal project. I will do so ASAP and update this issue, however I am posting it in ... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-06-13T04:04:30Z | 2023-06-13T15:10:27Z | 2023-06-13T15:10:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | oxysoft | 3,945,277 | MDQ6VXNlcjM5NDUyNzc= | User | false |
huggingface/diffusers | 1,754,045,803 | I_kwDOHa8MBc5ojJ1r | 3,765 | https://github.com/huggingface/diffusers/issues/3765 | https://api.github.com/repos/huggingface/diffusers/issues/3765 | control net：Images that produced stacking effects during training | ### Describe the bug
I tried to train control net using my own dataset, but I found the resulting images very strange. I haven't made any changes to the code (only the data input part), and the dataset I used is a 512 resolution image cut from Cityscape. But I don't think this is related to the dataset because I tried... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2023-06-13T05:05:13Z | 2023-07-21T15:03:37Z | 2023-07-21T15:03:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | YuanSnowing | 53,067,358 | MDQ6VXNlcjUzMDY3MzU4 | User | false |
huggingface/diffusers | 1,754,144,735 | I_kwDOHa8MBc5ojh_f | 3,767 | https://github.com/huggingface/diffusers/issues/3767 | https://api.github.com/repos/huggingface/diffusers/issues/3767 | FrozenDict is not "frozen" | ### Describe the bug
ForzenDict is not "frozen", I can change the context in this dict, and different changing methods get different results.
### Reproduction
```
import diffusers
from diffusers import DDIMScheduler
scheduler = DDIMScheduler.from_pretrained("runwayml/stable-diffusion-v1-5", subfolder='scheduler')... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-06-13T06:33:54Z | 2023-07-21T15:03:36Z | 2023-07-21T15:03:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LOOKCC | 22,993,065 | MDQ6VXNlcjIyOTkzMDY1 | User | false |
huggingface/diffusers | 1,754,167,013 | I_kwDOHa8MBc5ojnbl | 3,768 | https://github.com/huggingface/diffusers/issues/3768 | https://api.github.com/repos/huggingface/diffusers/issues/3768 | DDPO Example | **Is your feature request related to a problem? No.
A clear and concise description of what the problem is.
Current implementation of DDPO supports only TPU.
A general implementation that also supports GPU would be a great addition to the existing examples of diffusers library.
Supporting links:
- https://gi... | closed | completed | false | 7 | [
"community-examples",
"stale"
] | [] | 2023-06-13T06:50:15Z | 2023-08-23T15:04:32Z | 2023-08-23T15:04:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | abhijitpal1247 | 69,110,711 | MDQ6VXNlcjY5MTEwNzEx | User | false |
huggingface/diffusers | 1,754,269,715 | I_kwDOHa8MBc5okAgT | 3,769 | https://github.com/huggingface/diffusers/issues/3769 | https://api.github.com/repos/huggingface/diffusers/issues/3769 | Quatered image produced, StableDiffusionPanoramaPipeline | ### Describe the bug
Hi,
Anyone experienced the following when running the [multidifussion tutorial](https://huggingface.co/docs/diffusers/main/en/api/pipelines/panorama)?
It generates a quater dolomites with the default setup but, correct one with `height=1024`
### Reproduction
Just did as shown in https://hug... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-06-13T07:55:08Z | 2023-06-16T00:28:43Z | 2023-06-16T00:28:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | inkyusa | 5,215,050 | MDQ6VXNlcjUyMTUwNTA= | User | false |
huggingface/diffusers | 1,754,548,128 | I_kwDOHa8MBc5olEeg | 3,771 | https://github.com/huggingface/diffusers/issues/3771 | https://api.github.com/repos/huggingface/diffusers/issues/3771 | CKPT to Converted Diffusers are not able to Inference | Hey @patrickvonplaten @sayakpaul, i think there's a bug in the diffusers can you help me to clear this bug, I'm giving an extensive flow on how to reproduce this error.
I have a [script](https://github.com/huggingface/diffusers/blob/main/scripts/convert_original_stable_diffusion_to_diffusers.py) from diffusers which... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-06-13T10:23:12Z | 2023-08-15T15:03:27Z | 2023-08-15T15:03:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | satwiksunnam19 | 83,942,187 | MDQ6VXNlcjgzOTQyMTg3 | User | false |
huggingface/diffusers | 1,754,642,381 | I_kwDOHa8MBc5olbfN | 3,772 | https://github.com/huggingface/diffusers/issues/3772 | https://api.github.com/repos/huggingface/diffusers/issues/3772 | Feature Request: Add resize method crop_and_resize | Are you considering adding resize methods for each pipeline?
| closed | completed | false | 3 | [
"stale"
] | [] | 2023-06-13T11:15:23Z | 2023-08-02T15:03:40Z | 2023-08-02T15:03:40Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kadirnar | 36,204,372 | MDQ6VXNlcjM2MjA0Mzcy | User | false |
huggingface/diffusers | 1,754,988,082 | I_kwDOHa8MBc5omv4y | 3,774 | https://github.com/huggingface/diffusers/issues/3774 | https://api.github.com/repos/huggingface/diffusers/issues/3774 | No any GPU memory saved when using zero3 to train train_text_to_image.py | ### Describe the bug
In order to save GPU memory, I adopt the zero3 FSDP strategy in deepspeed. And then full finetune train train_text_to_image.py. However, I do not see any memory savings. Is there any problem here?
### Reproduction
My default_config.yaml
```
compute_environment: LOCAL_MACHINE
deepspeed_config:... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-06-13T14:10:11Z | 2023-11-14T16:05:45Z | 2023-07-21T15:03:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wangmiaowei | 27,915,265 | MDQ6VXNlcjI3OTE1MjY1 | User | false |
huggingface/diffusers | 1,755,129,553 | I_kwDOHa8MBc5onSbR | 3,775 | https://github.com/huggingface/diffusers/issues/3775 | https://api.github.com/repos/huggingface/diffusers/issues/3775 | Benchmark for max-autotune VS reduce-overhead ? | Reading the docs for `torch.compile`, I couldn't help but notice this 'max-autotune' option. Real-time addicts like me are always on the lookout for any millisecond they can scrape, so I think it would be interesting to include this in the benchmarks on this page https://huggingface.co/docs/diffusers/main/en/optimizati... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-13T15:15:39Z | 2023-07-23T15:03:12Z | 2023-07-23T15:03:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | oxysoft | 3,945,277 | MDQ6VXNlcjM5NDUyNzc= | User | false |
huggingface/diffusers | 1,363,934,121 | I_kwDOHa8MBc5RS_up | 378 | https://github.com/huggingface/diffusers/issues/378 | https://api.github.com/repos/huggingface/diffusers/issues/378 | Textual Inversion DiagonalGaussianDistribution | ### Describe the bug
When running `accelerate launch textual_inversion.py`
getting the following error:
```
Traceback (most recent call last):
File "textual_inversion.py", line 579, in <module>
main()
File "textual_inversion.py", line 501, in main
latents = vae.encode(batch["pixel_values"]).latent_d... | closed | completed | false | 2 | [
"bug"
] | [] | 2022-09-07T00:13:11Z | 2022-09-07T16:12:58Z | 2022-09-07T16:12:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AmanKishore | 10,477,380 | MDQ6VXNlcjEwNDc3Mzgw | User | false |
huggingface/diffusers | 1,756,021,875 | I_kwDOHa8MBc5oqsRz | 3,781 | https://github.com/huggingface/diffusers/issues/3781 | https://api.github.com/repos/huggingface/diffusers/issues/3781 | Cannot load multiple LoRAs. Need to revert PR #3649 | ### Describe the bug
https://github.com/huggingface/diffusers/pull/3649 Introduced a bug. It automatically reverts any lora loaded when calling load_lora_weights. I can no longer load multiple loras. Rather than automatically unloading loras, we should have a method to unload them individually and a method to unload t... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-06-14T04:15:09Z | 2023-06-15T12:23:42Z | 2023-06-15T12:23:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JemiloII | 5,192,788 | MDQ6VXNlcjUxOTI3ODg= | User | false |
huggingface/diffusers | 1,756,812,624 | I_kwDOHa8MBc5ottVQ | 3,784 | https://github.com/huggingface/diffusers/issues/3784 | https://api.github.com/repos/huggingface/diffusers/issues/3784 | Correspondence of Text-To-Video pipeline to Modelscope | The text-to-video pipeline `TextToVideoSDPipeline` described [here](https://huggingface.co/docs/diffusers/api/pipelines/text_to_video) is used to implement [VideoFusion](https://arxiv.org/pdf/2303.08320.pdf) if I understand correct. However, VideoFusion uses a base model, base noise and a residual model with a residual... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-06-14T12:38:41Z | 2023-07-31T15:03:32Z | 2023-07-31T15:03:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rob-hen | 115,473,528 | U_kgDOBuH8eA | User | false |
huggingface/diffusers | 1,757,052,034 | I_kwDOHa8MBc5ounyC | 3,787 | https://github.com/huggingface/diffusers/issues/3787 | https://api.github.com/repos/huggingface/diffusers/issues/3787 | ImportError: cannot import name 'DiffusionPipeline' from partially initialized module 'diffusers' | ### Describe the bug
When attempting to import any pipeline from the diffusers module, I get the import error:
> ImportError: cannot import name 'DiffusionPipeline' from partially initialized module 'diffusers'
### Reproduction
Create a new python environment
> python -m venv .venv
> source .venv/bin/activate... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-06-14T14:32:24Z | 2023-06-14T15:54:30Z | 2023-06-14T15:54:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | awmartin | 57,281 | MDQ6VXNlcjU3Mjgx | User | false |
huggingface/diffusers | 1,757,276,333 | I_kwDOHa8MBc5oveit | 3,788 | https://github.com/huggingface/diffusers/issues/3788 | https://api.github.com/repos/huggingface/diffusers/issues/3788 | [New Pipeline]: UniControl for Unified Controllable Visual Generation | ### Model/Pipeline/Scheduler description
Github: https://github.com/salesforce/UniControl
Paper: https://arxiv.org/abs/2305.11147
### Open source status
- [X] The model implementation is available
- [X] The model weights are available (Only relevant if addition is not a scheduler).
### Provide useful links for t... | closed | completed | false | 17 | [
"stale"
] | [] | 2023-06-14T16:34:01Z | 2024-03-25T02:57:12Z | 2023-11-22T15:07:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | canqin001 | 46,471,476 | MDQ6VXNlcjQ2NDcxNDc2 | User | false |
huggingface/diffusers | 1,757,669,443 | I_kwDOHa8MBc5ow-hD | 3,790 | https://github.com/huggingface/diffusers/issues/3790 | https://api.github.com/repos/huggingface/diffusers/issues/3790 | has_nsfw_concept is a bool | Hi, I noticed that most of the implementations of the smooth zoom in SD died because of an error that comes from
pipeline_stable_diffusion_inpaint.py, line 1064
"TypeError: 'bool' object is not iterable"
Wonder if anybody knows what caused it and if it's a bug | closed | completed | false | 5 | [
"stale"
] | [] | 2023-06-14T21:22:22Z | 2023-08-24T10:30:10Z | 2023-07-24T15:03:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | orenong | 52,457,663 | MDQ6VXNlcjUyNDU3NjYz | User | false |
huggingface/diffusers | 1,757,873,478 | I_kwDOHa8MBc5oxwVG | 3,792 | https://github.com/huggingface/diffusers/issues/3792 | https://api.github.com/repos/huggingface/diffusers/issues/3792 | StableDiffusionControlNetImg2ImgPipeline Not Working As it Should | ### Describe the bug
I'm using StableDiffusionControlNetImg2ImgPipeline and I believe it to be a "generalized" pipeline:
- If I want to do text2image, I set strength=1.0
- If I want to do image2image, I set strength<1.0
- If I want to use ControlNet, I set controlnet_conditioning_scale to > 1.0
- If I want to disa... | closed | completed | false | 5 | [
"bug"
] | [
"yiyixuxu"
] | 2023-06-15T01:21:24Z | 2023-06-16T16:05:17Z | 2023-06-16T04:06:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xiankgx | 4,113,258 | MDQ6VXNlcjQxMTMyNTg= | User | false |
huggingface/diffusers | 1,758,020,403 | I_kwDOHa8MBc5oyUMz | 3,793 | https://github.com/huggingface/diffusers/issues/3793 | https://api.github.com/repos/huggingface/diffusers/issues/3793 | [SDPA vs. xformers] Discussions on benchmarking SDPA and xformers and implications | Recently, we have been receiving issues from users complaining that SDPA leads to OOMs whereas xformers doesn't not.
So, to help identify the root cause of this, I started a simple benchmark to compare the timings of the different efficient implementations of attention provided by SDPA and xformers. I have restrict... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-06-15T04:38:37Z | 2023-07-24T15:03:40Z | 2023-07-24T15:03:40Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,758,153,486 | I_kwDOHa8MBc5oy0sO | 3,794 | https://github.com/huggingface/diffusers/issues/3794 | https://api.github.com/repos/huggingface/diffusers/issues/3794 | Support lora with bucket training which enable model to use image of different shape? | Great work~. Normally we have images with different shape,I hope the training script support the image of different shape, not only center-crop.This method place images with similar aspect ratios in the same bucket when feed to model. The feature is the same as “Aspect Ratio Bucketing” in https://github.com/kohya-ss/sd... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-06-15T06:47:57Z | 2023-07-24T15:03:39Z | 2023-07-24T15:03:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Anychnn | 15,360,646 | MDQ6VXNlcjE1MzYwNjQ2 | User | false |
huggingface/diffusers | 1,759,113,547 | I_kwDOHa8MBc5o2fFL | 3,799 | https://github.com/huggingface/diffusers/issues/3799 | https://api.github.com/repos/huggingface/diffusers/issues/3799 | Incorrect Inferred type for DDPMScheduler.from_pretrained model_id | ### Describe the bug
When using DDPMScheduler.from_pretrained to load up the noise scheduler the param pretrained_model_name_or_path is set to type Dict[str, Any] [https://github.com/huggingface/diffusers/blob/afcca3916fafccff1419c45f3de00be57db12dfc/src/diffusers/schedulers/scheduling_utils.py#LL80C62-L80C62](url) w... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-06-15T15:42:25Z | 2023-07-24T15:03:38Z | 2023-07-24T15:03:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | qnlbnsl | 46,204,848 | MDQ6VXNlcjQ2MjA0ODQ4 | User | false |
huggingface/diffusers | 1,363,976,246 | I_kwDOHa8MBc5RTKA2 | 380 | https://github.com/huggingface/diffusers/issues/380 | https://api.github.com/repos/huggingface/diffusers/issues/380 | Unused parameters in scheduler `init`'s | **What API design would you like to have changed or added to the library? Why?**
I see some of the [schedulers](https://github.com/huggingface/diffusers/blob/7bd50cabafc60bf45ebbe1957b125d3f4c758ba8/src/diffusers/schedulers/scheduling_lms_discrete.py#L34) had these parameters added to init
```
trained_beta... | closed | completed | false | 12 | [
"stale"
] | [
"anton-l"
] | 2022-09-07T01:34:10Z | 2022-11-29T15:04:00Z | 2022-11-29T15:04:00Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | natolambert | 10,695,622 | MDQ6VXNlcjEwNjk1NjIy | User | false |
huggingface/diffusers | 1,759,498,726 | I_kwDOHa8MBc5o39Hm | 3,802 | https://github.com/huggingface/diffusers/issues/3802 | https://api.github.com/repos/huggingface/diffusers/issues/3802 | --checkpoints_total_limit argument doesn't work in training examples | ### Describe the bug
When training examples from ```diffusers/examples``` (e.g. ```diffusers/examples/unconditional_image_generation```) and using ```--checkpoints_total_limit```, the desired behavior of limiting the number of checkpoints doesn't work. Instead, there is no limit to the number of checkpoints and the... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-06-15T20:22:27Z | 2023-06-15T22:39:10Z | 2023-06-15T22:39:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ShlomoLibo | 49,692,869 | MDQ6VXNlcjQ5NjkyODY5 | User | false |
huggingface/diffusers | 1,759,574,378 | I_kwDOHa8MBc5o4Plq | 3,803 | https://github.com/huggingface/diffusers/issues/3803 | https://api.github.com/repos/huggingface/diffusers/issues/3803 | Multi-GPU LoRA training -- model is not shared across multiple gpus | ### Describe the bug
**When launching `accelerate launch train_text_to_image_lora.py` with multi-GPU training (under examples/text-to-image folder), model is not correctly shared across multiple gpus.**
To reproduce the test result, add some test code like this after [this line](https://github.com/huggingface/di... | closed | completed | false | 17 | [
"bug"
] | [] | 2023-06-15T21:29:02Z | 2023-06-16T23:30:05Z | 2023-06-16T21:55:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yingfan-bot | 32,705,605 | MDQ6VXNlcjMyNzA1NjA1 | User | false |
huggingface/diffusers | 1,759,635,299 | I_kwDOHa8MBc5o4edj | 3,805 | https://github.com/huggingface/diffusers/issues/3805 | https://api.github.com/repos/huggingface/diffusers/issues/3805 | unable to run on colab CalledProcessErro | ### Describe the bug
I try to run [lora model](https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth_lora.py) in colab with accelerate library but it doesn't work.
### Reproduction
```bash
from accelerate.utils import write_basic_config
write_basic_config(mixed_precision="fp... | closed | completed | false | 11 | [
"bug",
"stale"
] | [] | 2023-06-15T22:36:32Z | 2023-08-17T13:40:59Z | 2023-08-17T13:40:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pure-rgb | 45,315,076 | MDQ6VXNlcjQ1MzE1MDc2 | User | false |
huggingface/diffusers | 1,759,974,496 | I_kwDOHa8MBc5o5xRg | 3,808 | https://github.com/huggingface/diffusers/issues/3808 | https://api.github.com/repos/huggingface/diffusers/issues/3808 | [DeepFloyd IF] Unable to train IF stage 2 of LoRA Dreambooth and Full Dreambooth. | ### Describe the bug
I have trained IF stage I LoRA Dreambooth but when training stage 2 I am getting following error: **ValueError: value should be one of int, float, str, bool, or torch.Tensor.** I am getting this error in both LoRA and Dreambooth.
### Reproduction
# first cell
import os
os.environ["MODEL_NAME... | closed | completed | false | 5 | [
"bug"
] | [
"williamberman"
] | 2023-06-16T05:59:53Z | 2023-07-25T04:10:56Z | 2023-07-25T04:10:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | firoz47 | 97,146,450 | U_kgDOBcpWUg | User | false |
huggingface/diffusers | 1,760,391,589 | I_kwDOHa8MBc5o7XGl | 3,809 | https://github.com/huggingface/diffusers/issues/3809 | https://api.github.com/repos/huggingface/diffusers/issues/3809 | Running `diffusers` with `torch`-DDP leads to warning on stride mismatch | ### Describe the bug
Hey folks!
First off, thanks a lot for the amazing work with `diffusers`!
Recently, we have observed a warning while using `diffusers`-based models and running scripts.
We also observe similar warning while using *only* the models provided in `diffusers/models/`.
While this error might not... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-06-16T10:59:43Z | 2023-06-20T05:33:54Z | 2023-06-20T05:33:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eeshandhekane | 26,852,397 | MDQ6VXNlcjI2ODUyMzk3 | User | false |
huggingface/diffusers | 1,364,164,290 | I_kwDOHa8MBc5RT37C | 381 | https://github.com/huggingface/diffusers/issues/381 | https://api.github.com/repos/huggingface/diffusers/issues/381 | Error in section `3. How to write your own inference pipeline with diffusers` of the notebook `diffusers/stable_diffusion.ipynb` | ### Describe the bug
Just open the notebook in colab, and you will get:
```
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
[<ipython-input-27-912da26b456b>](https://localhost:8080/#) in <module>
1 #ima... | closed | completed | false | 3 | [
"bug"
] | [] | 2022-09-07T06:34:48Z | 2022-09-07T15:25:53Z | 2022-09-07T15:25:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pootow | 688,742 | MDQ6VXNlcjY4ODc0Mg== | User | false |
huggingface/diffusers | 1,761,579,286 | I_kwDOHa8MBc5o_5EW | 3,811 | https://github.com/huggingface/diffusers/issues/3811 | https://api.github.com/repos/huggingface/diffusers/issues/3811 | Lora loaded with load_lora_weights has no effect on generation. | ### Describe the bug
Trying to load lora with diffusers for the first time given the new update, but they do not seem to have any effect. In this case, I'm trying a Lora that leads to an messy oily painting style and worked perfectly well in Auto1111.
Trying Auto1111, we get this:
![visual](https://github.com/hu... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-06-17T01:46:43Z | 2023-06-30T14:19:28Z | 2023-06-30T14:19:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lucmaki | 26,182,176 | MDQ6VXNlcjI2MTgyMTc2 | User | false |
huggingface/diffusers | 1,761,653,919 | I_kwDOHa8MBc5pALSf | 3,812 | https://github.com/huggingface/diffusers/issues/3812 | https://api.github.com/repos/huggingface/diffusers/issues/3812 | {'controlnet'} was not found in config. Values will be initialized to default values. | ### Describe the bug
I tried the training of the ControlNet in the main branch right away.
In function log_validation, The following prompts appear:{'controlnet'} was not found in config. Values will be initialized to default values.
### Reproduction
def log_validation(batchs, vae, text_encoder, tokenizer, unet,... | closed | completed | false | 10 | [
"bug",
"stale"
] | [] | 2023-06-17T05:32:36Z | 2023-09-03T15:04:09Z | 2023-09-03T15:04:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dgr9527 | 49,510,355 | MDQ6VXNlcjQ5NTEwMzU1 | User | false |
huggingface/diffusers | 1,761,666,852 | I_kwDOHa8MBc5pAOck | 3,813 | https://github.com/huggingface/diffusers/issues/3813 | https://api.github.com/repos/huggingface/diffusers/issues/3813 | How to finetune stable diffusion with multi-gpu under the mode of model parallelism | I wonder if there is any solution to run the demo in https://github.com/huggingface/diffusers/tree/main/examples/text_to_image with multi-gpu, where VRAM of each gpu is 16G and the total VRAM is 64G.
I don't want to use Lora or 8-bit optimazation method | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-17T06:10:51Z | 2023-07-25T15:03:10Z | 2023-07-25T15:03:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chaojiewang94 | 16,972,919 | MDQ6VXNlcjE2OTcyOTE5 | User | false |
huggingface/diffusers | 1,761,675,774 | I_kwDOHa8MBc5pAQn- | 3,814 | https://github.com/huggingface/diffusers/issues/3814 | https://api.github.com/repos/huggingface/diffusers/issues/3814 | Trying to save model locally returns ListConfig is not JSON serializable | ``` python
# model_to_train_from_scratch.py
from diffusers import UNet2DModel
class MyLightningClass(L.LightningModule):
model = UNet2DModel (...)
self.model = model
self.output_path = "./outputs/"
def train_step(self, batch, batch_idx):
...
def validation_step(self):
... # some ... | closed | completed | false | 2 | [] | [] | 2023-06-17T06:40:34Z | 2023-07-06T11:45:25Z | 2023-07-06T11:45:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fujistoo | 14,198,032 | MDQ6VXNlcjE0MTk4MDMy | User | false |
huggingface/diffusers | 1,761,894,149 | I_kwDOHa8MBc5pBF8F | 3,816 | https://github.com/huggingface/diffusers/issues/3816 | https://api.github.com/repos/huggingface/diffusers/issues/3816 | AttributeError: module transformers has no attribute CLIPImageProcessor | ### Describe the bug
Getting `AttributeError: module transformers has no attribute CLIPImageProcessor` with [code](https://github.com/carlthome/stable-diffusion/blob/main/main.py) that used to work six months ago. No changes on my end as far as I know.
### Reproduction
```
git clone git@github.com:carlthome/s... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-06-17T15:23:50Z | 2023-07-27T15:03:24Z | 2023-07-27T15:03:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | carlthome | 1,595,907 | MDQ6VXNlcjE1OTU5MDc= | User | false |
huggingface/diffusers | 1,762,013,135 | I_kwDOHa8MBc5pBi_P | 3,817 | https://github.com/huggingface/diffusers/issues/3817 | https://api.github.com/repos/huggingface/diffusers/issues/3817 | EDGE: Editable Dance Generation From Music | ### Model/Pipeline/Scheduler description
The authors of this paper propose a diffusion-based method for generating human dance sequences (represented by joint angle trajectories) conditioned on music audio embeddings. The weights are stored in Google Docs and are open to the public, but translating the weights into a ... | open | null | false | 0 | [
"New pipeline/model"
] | [] | 2023-06-17T22:12:02Z | 2023-06-19T15:30:16Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | clarencechen | 8,482,341 | MDQ6VXNlcjg0ODIzNDE= | User | false |
huggingface/diffusers | 1,762,654,490 | I_kwDOHa8MBc5pD_ka | 3,822 | https://github.com/huggingface/diffusers/issues/3822 | https://api.github.com/repos/huggingface/diffusers/issues/3822 | StableDiffusionPipeline() and CLIP "cooperation" | Hello!
First of all, amazing work, guys! Absolutely love the whole `transformers`, `diffusers` and others for its simplicity! ❤️
I would suggest an improvement to a `StableDiffusionPipeline()` function to allow for portable & offline execution on any machine if the required files are already in specified locations... | closed | completed | false | 2 | [] | [] | 2023-06-19T03:17:13Z | 2023-06-28T13:29:51Z | 2023-06-28T13:29:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghost | 10,137 | MDQ6VXNlcjEwMTM3 | User | false |
huggingface/diffusers | 1,762,847,111 | I_kwDOHa8MBc5pEumH | 3,823 | https://github.com/huggingface/diffusers/issues/3823 | https://api.github.com/repos/huggingface/diffusers/issues/3823 | class_labels should be provided when num_class_embeds > 0 | Hi @patrickvonplaten @sayakpaul
I'm training controlnet for Edge. I followed the readme and executed the scripts and tried to train the circle filling dataset with the base model as stable diffusion v2.1. But facing the following error:
ValueError: class_labels should be provided when num_class_embeds > 0
I s... | closed | completed | false | 3 | [] | [] | 2023-06-19T06:34:16Z | 2023-06-20T05:02:58Z | 2023-06-20T05:02:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rakesh-reddy95 | 20,672,248 | MDQ6VXNlcjIwNjcyMjQ4 | User | false |
huggingface/diffusers | 1,763,141,236 | I_kwDOHa8MBc5pF2Z0 | 3,824 | https://github.com/huggingface/diffusers/issues/3824 | https://api.github.com/repos/huggingface/diffusers/issues/3824 | Question about StableDiffusionInpaintPipelineLegacy and StableDiffusionInpaintPipeline | Did you not use original latent in StableDiffusionInpaintPipeline intentionally? If yes I think it is hard to say StableDiffusionInpaintPipeline can achieve exactly the same functionality as StableDiffusionInpaintPipelineLegacy. What do think?
https://github.com/huggingface/diffusers/blob/f7cc9adc05c36985c8b3f3e95a5... | closed | completed | false | 4 | [] | [] | 2023-06-19T09:35:21Z | 2023-06-21T10:39:29Z | 2023-06-20T05:10:37Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jinwonkim93 | 26,476,095 | MDQ6VXNlcjI2NDc2MDk1 | User | false |
huggingface/diffusers | 1,764,170,901 | I_kwDOHa8MBc5pJxyV | 3,826 | https://github.com/huggingface/diffusers/issues/3826 | https://api.github.com/repos/huggingface/diffusers/issues/3826 | [New Pipeline]: Audio-Journey: Visual+LLM-aided Audio Encodec Diffusion | ### Model/Pipeline/Scheduler description
We efficiently trained an Audio Diffusion model with the aid of Alpaca augmented audio captions using AudioSet labels;
[website](https://audiojourney.github.io/)
[preprint](https://github.com/audiojourney/audiojourney.github.io/blob/main/neurIPS_2023_v1.2.pdf)
[Appendix](... | open | reopened | false | 3 | [
"New pipeline/model",
"community-examples"
] | [] | 2023-06-19T20:16:02Z | 2023-07-03T18:49:03Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lijuncheng16 | 6,412,709 | MDQ6VXNlcjY0MTI3MDk= | User | false |
huggingface/diffusers | 1,764,314,454 | I_kwDOHa8MBc5pKU1W | 3,827 | https://github.com/huggingface/diffusers/issues/3827 | https://api.github.com/repos/huggingface/diffusers/issues/3827 | Tokenizer error using civitai models as a starting point | I'm getting this error:
```
ValueError: Calling CLIPTokenizer.from_pretrained() with the path to a single file or url is not supported for this tokenizer. Use a model identifier or the path to a directory
instead.
```
I see that at least one person on Reddit ran into the same issue, but that was months ago an... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-19T22:41:49Z | 2023-07-28T15:03:46Z | 2023-07-28T15:03:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | elbamos | 10,103,420 | MDQ6VXNlcjEwMTAzNDIw | User | false |
huggingface/diffusers | 1,764,363,334 | I_kwDOHa8MBc5pKgxG | 3,828 | https://github.com/huggingface/diffusers/issues/3828 | https://api.github.com/repos/huggingface/diffusers/issues/3828 | TypeError when using LayoutLMv3 in pipeline for Document QA | I get the following error: ```forward() got an unexpected keyword argument 'image'``` when I run the code below:
```
img = Image.open("test.png").convert("RGB")
nlp = pipeline("document-question-answering","microsoft/layoutlmv3-base")
nlp(img, 'what is the number of tabs open in this window?')
```
Haven't... | closed | completed | false | 2 | [] | [] | 2023-06-19T23:34:13Z | 2023-06-27T01:43:42Z | 2023-06-27T01:43:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FFFiend | 96,851,409 | U_kgDOBcXV0Q | User | false |
huggingface/diffusers | 1,764,550,477 | I_kwDOHa8MBc5pLOdN | 3,829 | https://github.com/huggingface/diffusers/issues/3829 | https://api.github.com/repos/huggingface/diffusers/issues/3829 | ONNX pipeline Stable Diffusion for text2image task with EulerAncestralDiscreteScheduler, fixed weight, prompts and seed would generate different images | ### Describe the bug
ONNX pipeline Stable Diffusion for text2image task with EulerAncestralDiscreteScheduler, fixed weight, prompts and seed would generate different images
### Reproduction
**PreConfig:** Fix all configurations for sd model: fixed seed value; numbers of steps; scheduler; prompt text
**What ... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-06-20T03:26:18Z | 2023-07-28T15:03:45Z | 2023-07-28T15:03:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kotomiDu | 22,112,323 | MDQ6VXNlcjIyMTEyMzIz | User | false |
huggingface/diffusers | 1,364,356,233 | I_kwDOHa8MBc5RUmyJ | 383 | https://github.com/huggingface/diffusers/issues/383 | https://api.github.com/repos/huggingface/diffusers/issues/383 | Question about: `ConfigMixin` vs `XYZConfig class` | I see that [ConfigMixin](https://github.com/huggingface/diffusers/blob/d799084a9a7d5196ae708da72b46dd8e84604194/src/diffusers/configuration_utils.py#L38) & [@register_to_config](https://github.com/huggingface/diffusers/blob/d799084a9a7d5196ae708da72b46dd8e84604194/src/diffusers/configuration_utils.py#L300) are used to ... | closed | completed | false | 2 | [] | [] | 2022-09-07T09:18:27Z | 2022-09-07T10:04:05Z | 2022-09-07T10:03:55Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mishig25 | 11,827,707 | MDQ6VXNlcjExODI3NzA3 | User | false |
huggingface/diffusers | 1,764,601,319 | I_kwDOHa8MBc5pLa3n | 3,830 | https://github.com/huggingface/diffusers/issues/3830 | https://api.github.com/repos/huggingface/diffusers/issues/3830 | Image variation training | Hi,
I can't find the source code for image variations training. Would you please add it to the repo?
Thank you | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-20T04:14:58Z | 2023-07-28T15:03:43Z | 2023-07-28T15:03:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shnamin | 53,882,576 | MDQ6VXNlcjUzODgyNTc2 | User | false |
huggingface/diffusers | 1,765,903,664 | I_kwDOHa8MBc5pQY0w | 3,834 | https://github.com/huggingface/diffusers/issues/3834 | https://api.github.com/repos/huggingface/diffusers/issues/3834 | Add controlnet module to instruct pix2pix training code | How can I add controlnet module to the instruct pix2pix training code?
Should I use it this way:
```
controlnet = ControlNetModel.from_pretrained(args.controlnet_model_name_or_path)
unet = UNet2DConditionModel.from_pretrained("timbrooks/instruct-pix2pix")
.
.
.
down_block_res_samples, mid_block_res_sample... | closed | completed | false | 4 | [] | [] | 2023-06-20T18:11:18Z | 2023-06-22T03:49:31Z | 2023-06-22T03:49:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mahdip72 | 42,680,708 | MDQ6VXNlcjQyNjgwNzA4 | User | false |
huggingface/diffusers | 1,766,625,740 | I_kwDOHa8MBc5pTJHM | 3,837 | https://github.com/huggingface/diffusers/issues/3837 | https://api.github.com/repos/huggingface/diffusers/issues/3837 | Pass large images/masks to StableDiffusionInpaintPipeline | Currently the `StableDiffusionInpaintPipeline` implementation only allows for 512*512 images/masks. However, in some community applications, I noticed that larger images/masks are also passed to the inpainting model (most probably not through this pipeline).
The ask is to add this a feature to the `StableDiffusionI... | closed | completed | false | 4 | [] | [
"yiyixuxu"
] | 2023-06-21T03:44:46Z | 2023-08-02T15:03:36Z | 2023-08-02T15:03:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hosseinsarshar | 4,457,205 | MDQ6VXNlcjQ0NTcyMDU= | User | false |
huggingface/diffusers | 1,766,640,529 | I_kwDOHa8MBc5pTMuR | 3,838 | https://github.com/huggingface/diffusers/issues/3838 | https://api.github.com/repos/huggingface/diffusers/issues/3838 | Video Composer | ### Model/Pipeline/Scheduler description
VideoComposer: Compositional Video Synthesis
with Motion Controllability
### Open source status
- [ ] The model implementation is available
- [ ] The model weights are... | open | null | false | 2 | [
"New pipeline/model"
] | [] | 2023-06-21T04:05:44Z | 2026-06-08T05:31:05Z | null | NONE | null | 20260608T053515Z | 2026-06-08T05:35:15Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,364,357,911 | I_kwDOHa8MBc5RUnMX | 384 | https://github.com/huggingface/diffusers/issues/384 | https://api.github.com/repos/huggingface/diffusers/issues/384 | AttributeError: 'DiagonalGaussianDistribution' object has no attribute 'latent_dist' | ### Describe the bug
Hey I get this error for inverting images
AttributeError: 'DiagonalGaussianDistribution' object has no attribute 'latent_dist'
### Reproduction
accelerate launch textual_inversion.py --pretrained_model_name_or_path=$MODEL_NAME --use_auth_token --train_data_dir=$DATA_DIR --learnable_property=... | closed | completed | false | 2 | [
"bug"
] | [] | 2022-09-07T09:19:46Z | 2023-04-12T12:22:02Z | 2022-09-07T09:28:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | UglyStupidHonest | 68,284,461 | MDQ6VXNlcjY4Mjg0NDYx | User | false |
huggingface/diffusers | 1,768,005,716 | I_kwDOHa8MBc5pYaBU | 3,840 | https://github.com/huggingface/diffusers/issues/3840 | https://api.github.com/repos/huggingface/diffusers/issues/3840 | [UnCLIPImageVariationPipeline] Possible Mismatch in Implementation vs Paper | Hi Team,
There is a possible mismatch in the implementation of [UnCLIPImageVariationPipeline](UnCLIPImageVariationPipeline) w.r.t the Dalle-2 [paper](https://cdn.openai.com/papers/dall-e-2.pdf)
### Context:
Any image can be represented by a bipartite representation $(z_i, X_T)$ where $z_i$ is the CLIP Image embe... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-21T16:45:22Z | 2023-07-30T15:03:49Z | 2023-07-30T15:03:49Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Abhinay1997 | 24,771,261 | MDQ6VXNlcjI0NzcxMjYx | User | false |
huggingface/diffusers | 1,768,193,014 | I_kwDOHa8MBc5pZHv2 | 3,843 | https://github.com/huggingface/diffusers/issues/3843 | https://api.github.com/repos/huggingface/diffusers/issues/3843 | global_step key not found in model. | ### Describe the bug
Last updates on the convert_from_ckpt.py broke converting pre trained models from places like civitai to diffusers.
### Reproduction
Load any model from civitai using safetensors with the StableDiffusionPipeline.from_ckpt("location of the model") or similar
### Logs
```shell
2023-06-21 18:08:3... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-06-21T18:39:08Z | 2023-06-24T09:48:04Z | 2023-06-24T09:48:04Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | VincentNeemie | 92,559,302 | U_kgDOBYRXxg | User | false |
huggingface/diffusers | 1,769,411,684 | I_kwDOHa8MBc5pdxRk | 3,851 | https://github.com/huggingface/diffusers/issues/3851 | https://api.github.com/repos/huggingface/diffusers/issues/3851 | DDIMScheduler silently breaking when using epsilon prediction and rescaling betas to zero SNR | ### Describe the bug
When using the `DDIMScheduler` with arguments `prediction_type="epsilon"` (which is default) and activating `rescale_betas_zero_snr=True`, and then trying to start sampling with it, one encounters `Inf` values in the returned prediction immediately at the first step.
This occurs because of [th... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-06-22T10:38:10Z | 2023-08-06T15:03:26Z | 2023-08-06T15:03:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | f90 | 23,065,103 | MDQ6VXNlcjIzMDY1MTAz | User | false |
huggingface/diffusers | 1,769,791,678 | I_kwDOHa8MBc5pfOC- | 3,853 | https://github.com/huggingface/diffusers/issues/3853 | https://api.github.com/repos/huggingface/diffusers/issues/3853 | LoRA weights won't load (HeaderTooLarge) | I'm following [train_text_to_image_lora.py](https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image_lora.py) example.
Saving LoRA weights with:
```
unet = unet.to(torch.float32)
unet.save_attn_procs("path/to/pytorch_lora_weights.bin")
```
Then, loading weights with:
```
... | closed | completed | false | 8 | [
"stale"
] | [] | 2023-06-22T14:21:44Z | 2023-07-30T15:03:47Z | 2023-07-30T15:03:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jorgejgnz | 60,674,219 | MDQ6VXNlcjYwNjc0MjE5 | User | false |
huggingface/diffusers | 1,769,888,199 | I_kwDOHa8MBc5pflnH | 3,854 | https://github.com/huggingface/diffusers/issues/3854 | https://api.github.com/repos/huggingface/diffusers/issues/3854 | pipe = StableDiffusionPipeline.from_ckpt(model_p) locked | ### Describe the bug
pipe = StableDiffusionPipeline.from_ckpt(model_p)
Can not load safetygensor from local
### Reproduction
```
import torch
from diffusers import DDIMScheduler
from diffusers.pipelines.stable_diffusion import StableDiffusionPipeline
import sys
import os
"""
python -m pip install ... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-06-22T15:05:07Z | 2023-07-31T15:03:26Z | 2023-07-31T15:03:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lucasjinreal | 21,303,438 | MDQ6VXNlcjIxMzAzNDM4 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.