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 | 1,950,290,574 | I_kwDOHa8MBc50PxKO | 5,444 | https://github.com/huggingface/diffusers/issues/5444 | https://api.github.com/repos/huggingface/diffusers/issues/5444 | Integrate Pixart Alpha | Pixart Alpha is a new generative model that makes 1024x1024 images with just 600m parameters (!).
Paper: https://pixart-alpha.github.io/
> The most advanced text-to-image (T2I) models require significan... | closed | completed | false | 7 | [] | [] | 2023-10-18T17:30:43Z | 2023-11-03T12:13:27Z | 2023-11-03T12:13:27Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AmericanPresidentJimmyCarter | 110,263,573 | U_kgDOBpJ9FQ | User | false |
huggingface/diffusers | 1,951,092,769 | I_kwDOHa8MBc50S1Ah | 5,447 | https://github.com/huggingface/diffusers/issues/5447 | https://api.github.com/repos/huggingface/diffusers/issues/5447 | StableDiffusionXLImg2ImgPipeline: denoising_start doesn't work correctly with 2nd order schedulers | ### Describe the bug
We were setting up SDXL with the refiner in an ensemble of experts configuration, where we split the generation between the base model and the refiner.
When using `denoising_start` for the refiner step, we noticed that the result images were either completely black, or not fully denoised for ... | closed | completed | false | 2 | [
"bug"
] | [
"yiyixuxu"
] | 2023-10-19T02:50:15Z | 2023-10-25T14:24:18Z | 2023-10-25T14:24:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nhnt11 | 485,488 | MDQ6VXNlcjQ4NTQ4OA== | User | false |
huggingface/diffusers | 1,951,585,267 | I_kwDOHa8MBc50UtPz | 5,449 | https://github.com/huggingface/diffusers/issues/5449 | https://api.github.com/repos/huggingface/diffusers/issues/5449 | AttributeError: Can't pickle local object 'main.<locals>.preprocess_train' | Running train_text_to_image_lora.py with the following command:
['C:\\Python3109\\python.exe', 'train_text_to_image_lora.py', '--pretrained_model_name_or_path=stabilityai/stable-diffusion-2-1', '--train_data_dir=G:\\Project Files\\Kohya_SS - Data\\Training\\JasonXBLoRA', '--dataloader_num_workers=8', '--resolution=5... | closed | completed | false | 9 | [
"stale"
] | [] | 2023-10-19T08:25:46Z | 2024-01-03T15:07:36Z | 2024-01-03T15:07:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | RhodianSK | 99,561,171 | U_kgDOBe8u0w | User | false |
huggingface/diffusers | 1,951,588,776 | I_kwDOHa8MBc50UuGo | 5,450 | https://github.com/huggingface/diffusers/issues/5450 | https://api.github.com/repos/huggingface/diffusers/issues/5450 | ValueError on FlaxStableDiffusionPipeline | ### Describe the bug
`FlaxStableDiffusionPipeline.__call__()` raises `ValueError: assignment destination is read-only`.
Probably NSFW filters are involved.
### Reproduction
- FlaxStableDiffusionPipeline
- 'CompVis/stable-diffusion-v1-4'
- revision='bf16'
- dtype=jnp.bfloat16
- prompt 'pretty asian girl' (no n... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-10-19T08:27:08Z | 2023-11-14T11:16:54Z | 2023-11-14T11:16:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | YusukeSuzuki | 657,060 | MDQ6VXNlcjY1NzA2MA== | User | false |
huggingface/diffusers | 1,951,613,082 | I_kwDOHa8MBc50U0Ca | 5,451 | https://github.com/huggingface/diffusers/issues/5451 | https://api.github.com/repos/huggingface/diffusers/issues/5451 | QUESTIONS about SDXL INPAINTING | I found that the SDXL-INPAINTING model is named version 0.1 Hence, may I ask if you guys want to further improve the model?
Thanks! | closed | completed | false | 1 | [
"stale"
] | [] | 2023-10-19T08:39:39Z | 2023-11-28T15:05:43Z | 2023-11-28T15:05:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | George0726 | 38,740,075 | MDQ6VXNlcjM4NzQwMDc1 | User | false |
huggingface/diffusers | 1,952,487,499 | I_kwDOHa8MBc50YJhL | 5,455 | https://github.com/huggingface/diffusers/issues/5455 | https://api.github.com/repos/huggingface/diffusers/issues/5455 | subprocess.CalledProcessError when running example notebook for dreambooth+lora | ### Describe the bug
subprocess.CalledProcessError occurs when running an example notebook https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/SDXL_DreamBooth_LoRA_.ipynb
in colab free tier with t4 runtime.
### Reproduction
run the notebook till training part.
https://colab.research.... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-10-19T15:04:51Z | 2023-10-20T01:43:52Z | 2023-10-20T01:43:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nvnarayna | 98,206,357 | U_kgDOBdqClQ | User | false |
huggingface/diffusers | 1,952,816,312 | I_kwDOHa8MBc50ZZy4 | 5,457 | https://github.com/huggingface/diffusers/issues/5457 | https://api.github.com/repos/huggingface/diffusers/issues/5457 | What is function of `attention_mask` in `get_attention_scores`? | What is function of `attention_mask` in `get_attention_scores`? I guess it is used to ignore some value when calculating the attention map
I can not find a example in diffusers library that actually use this `attention_mask`. Could you provide an example on how to use it?
https://github.com/huggingface/diffusers/bl... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-19T18:14:38Z | 2023-11-28T15:05:41Z | 2023-11-28T15:05:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | g-jing | 44,223,191 | MDQ6VXNlcjQ0MjIzMTkx | User | false |
huggingface/diffusers | 1,952,880,779 | I_kwDOHa8MBc50ZpiL | 5,458 | https://github.com/huggingface/diffusers/issues/5458 | https://api.github.com/repos/huggingface/diffusers/issues/5458 | LoraLoaderMixin.lora_state_dict does not respect argument local_files_only | ### Describe the bug
I assume Lora loading is going to be subject to a lot of changes but on 0.21.4 I am unable to load cached weights
without a network connection at least for the model 'sayakpaul/sd-model-finetuned-lora-t4'
When attempting to use the forwarded argument: local_files_only=True
There is a networ... | closed | completed | false | 7 | [
"bug",
"stale"
] | [
"sayakpaul"
] | 2023-10-19T18:58:40Z | 2023-11-28T15:05:39Z | 2023-11-28T15:05:39Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Teriks | 14,919,098 | MDQ6VXNlcjE0OTE5MDk4 | User | false |
huggingface/diffusers | 1,953,211,968 | I_kwDOHa8MBc50a6ZA | 5,460 | https://github.com/huggingface/diffusers/issues/5460 | https://api.github.com/repos/huggingface/diffusers/issues/5460 | Loading models from cache extra slow due to extra conversion | ### Describe the bug
I have 2 Python environments, one on Windows and another on Linux (over WSL), both using diffusers. To avoid having mutliple copies of the same model on disk, I try to make these two installations share a single diffusers model cache.
1. Ubuntu on WSL, Python 3.10.12, cache is a symlink poin... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-10-19T23:37:28Z | 2023-10-24T17:01:53Z | 2023-10-24T17:01:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | joe-chiu | 14,063,642 | MDQ6VXNlcjE0MDYzNjQy | User | false |
huggingface/diffusers | 1,953,222,034 | I_kwDOHa8MBc50a82S | 5,461 | https://github.com/huggingface/diffusers/issues/5461 | https://api.github.com/repos/huggingface/diffusers/issues/5461 | Enable Flax Stable Diffusion 2.1 model in the example | ### Describe the bug
In the current stable diffusion code [example](https://github.com/huggingface/diffusers/tree/main/examples/text_to_image), it supports model `CompVis/stable-diffusion-v1-4-flax`.
Request to add a `from_pt` bool arg and pass in `CLIPTokenizer.from_pretrained`, `FlaxCLIPTextModel.from_pretrai... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-10-19T23:50:22Z | 2023-10-25T21:19:43Z | 2023-10-25T21:19:43Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | RissyRan | 20,385,466 | MDQ6VXNlcjIwMzg1NDY2 | User | false |
huggingface/diffusers | 1,953,373,150 | I_kwDOHa8MBc50bhve | 5,462 | https://github.com/huggingface/diffusers/issues/5462 | https://api.github.com/repos/huggingface/diffusers/issues/5462 | img2img for Latent Consistency Models | **Is your feature request related to a problem? Please describe.**
Add support for img2img on the LCM pipeline
**Describe the solution you'd like**
img2img support
| closed | completed | false | 5 | [
"stale"
] | [] | 2023-10-20T02:20:28Z | 2023-11-21T14:24:08Z | 2023-11-21T14:24:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | arcanite24 | 3,642,929 | MDQ6VXNlcjM2NDI5Mjk= | User | false |
huggingface/diffusers | 1,953,575,065 | I_kwDOHa8MBc50cTCZ | 5,463 | https://github.com/huggingface/diffusers/issues/5463 | https://api.github.com/repos/huggingface/diffusers/issues/5463 | UnboundLocalError: local variable 'image_latents' referenced before assignment | ### Describe the bug
Getting this error with strength = 1.0 (is_strength_max) other wise it's working fine in the StableDiffusionXLControlNetInpaintPipeline
### Reproduction
```
sdxl_inpaint_pipe(
prompt=prompt,
negative_prompt= negative_prompt,
width=1024,
height=1024,
... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-10-20T05:45:11Z | 2023-10-24T13:59:02Z | 2023-10-24T13:59:01Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shubhdotai | 6,880,985 | MDQ6VXNlcjY4ODA5ODU= | User | false |
huggingface/diffusers | 1,953,779,818 | I_kwDOHa8MBc50dFBq | 5,465 | https://github.com/huggingface/diffusers/issues/5465 | https://api.github.com/repos/huggingface/diffusers/issues/5465 | DiffEdit pipeline does not support “inpaint_strength” values other than 0.8 | ### Describe the bug
The diff Edit throws an exception when passed inpaint_strength different from 0.8.
It does however function with strength of 0.8.
### Reproduction
```
import PIL
import requests
import torch
from io import BytesIO
import numpy as np
from diffusers import StableDiffusionDiffEditPipel... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-10-20T08:19:39Z | 2023-12-27T15:06:33Z | 2023-12-27T15:06:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | exx8 | 8,540,180 | MDQ6VXNlcjg1NDAxODA= | User | false |
huggingface/diffusers | 1,953,868,876 | I_kwDOHa8MBc50daxM | 5,466 | https://github.com/huggingface/diffusers/issues/5466 | https://api.github.com/repos/huggingface/diffusers/issues/5466 | sdxl image-to-image load lora file: ValueError: Cannot load because down.weight expected shape | I am trying to run image-to-image with sdxl model and lora file, but when pipe.load_lora_weights is error.
same model and same lora file, text-to-image is well.
### Reproduction
pipe = SDXLLongPromptWeightingPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0",
... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-10-20T09:12:32Z | 2024-03-11T16:41:39Z | 2023-10-20T10:15:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kolar1988 | 39,948,604 | MDQ6VXNlcjM5OTQ4NjA0 | User | false |
huggingface/diffusers | 1,953,994,229 | I_kwDOHa8MBc50d5X1 | 5,467 | https://github.com/huggingface/diffusers/issues/5467 | https://api.github.com/repos/huggingface/diffusers/issues/5467 | [New Pipeline]: Rerender-A-Video: Zero-Shot Video-to-Video Translation | ### Model/Pipeline/Scheduler description
We propose a zero-shot video-to-video translation framework for SD. (https://www.mmlab-ntu.com/project/rerender/)
Features:
* Temporal consistency: cross-frame constraints for low-level temporal consistency.
* Zero-shot: no training or fine-tuning required.
* Flexibilit... | closed | completed | false | 13 | [
"stale"
] | [] | 2023-10-20T10:24:05Z | 2024-01-13T08:12:46Z | 2024-01-10T06:58:09Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | SingleZombie | 47,652,064 | MDQ6VXNlcjQ3NjUyMDY0 | User | false |
huggingface/diffusers | 1,954,047,232 | I_kwDOHa8MBc50eGUA | 5,468 | https://github.com/huggingface/diffusers/issues/5468 | https://api.github.com/repos/huggingface/diffusers/issues/5468 | Redesign SchedulerCommonTest to use less hardcoded timesteps | **What API design would you like to have changed or added to the library? Why?**
I'm not sure if this counts as an API design issue but I think it would be more convenient if the tests and utility functions in `SchedulerCommonTest` had less dependency on hardcoded timesteps. For example, some tests that use hardcode... | closed | completed | false | 4 | [
"stale"
] | [
"yiyixuxu"
] | 2023-10-20T10:56:45Z | 2023-12-27T15:06:30Z | 2023-12-27T15:06:30Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dg845 | 58,458,699 | MDQ6VXNlcjU4NDU4Njk5 | User | false |
huggingface/diffusers | 1,376,874,162 | I_kwDOHa8MBc5SEW6y | 547 | https://github.com/huggingface/diffusers/issues/547 | https://api.github.com/repos/huggingface/diffusers/issues/547 | [Documentation] Better documentation that shows how to load stable diffusion locally | ### Describe the bug
We've gotten much feedback about the bad UX of either having to pass an `auth_token` to `from_pretrained(...)` or about the cache being to complicated to figure out how to load the model locally => we need to improve the documentation here.
E.g. - see: https://github.com/huggingface/diffusers/i... | closed | completed | false | 1 | [
"bug"
] | [
"patrickvonplaten"
] | 2022-09-17T21:10:12Z | 2022-11-07T19:22:28Z | 2022-11-07T19:22:28Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,954,383,006 | I_kwDOHa8MBc50fYSe | 5,471 | https://github.com/huggingface/diffusers/issues/5471 | https://api.github.com/repos/huggingface/diffusers/issues/5471 | SDXL inpaint UNet2DConditionModel: TyperError in regards with added_cond_kwargs.get("text_embeds") | ### Describe the bug
Hello,
I am trying to utilize the UNet2DConditionModel from diffusers and when I pass the parameters needed it throws an error. From my understanding StabilityAI uses the same unet for "stabilityai/stable-diffusion-2-inpainting" and when I call the UNet2DConditionModel using the equivalent code... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-10-20T14:14:43Z | 2024-06-02T08:42:30Z | 2024-01-09T15:08:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | IoannisDem | 42,682,824 | MDQ6VXNlcjQyNjgyODI0 | User | false |
huggingface/diffusers | 1,955,453,151 | I_kwDOHa8MBc50jdjf | 5,474 | https://github.com/huggingface/diffusers/issues/5474 | https://api.github.com/repos/huggingface/diffusers/issues/5474 | [New Pipeline] Stable Diffusion Reference Only. Zero-shot secondary creation of any character, line art coloring, style transfer. Different from ControlNet Reference Only and more stable powerful. | ### Model/Pipeline/Scheduler description
This is my latest work. Stable Diffusion Reference Only is an Imgs2Img pipeline. Only 4G/8G VRAM is needed to implement secondary creation of any character, line drawing coloring, and style transfer. It contains a new pipeline and a new UNetModel. Are diffusers interested in ad... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-10-21T11:33:31Z | 2023-12-05T02:42:41Z | 2023-12-05T02:42:41Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aihao2000 | 51,043,929 | MDQ6VXNlcjUxMDQzOTI5 | User | false |
huggingface/diffusers | 1,955,691,846 | I_kwDOHa8MBc50kX1G | 5,475 | https://github.com/huggingface/diffusers/issues/5475 | https://api.github.com/repos/huggingface/diffusers/issues/5475 | LaVie: High-Quality Video Generation with Cascaded Latent Diffusion Models | ### Model/Pipeline/Scheduler description
This work aims to learn a high-quality text-to-video (T2V) generative model by leveraging a pre-trained text-to-image (T2I) model as a basis. It is a highly desirable yet challenging task to simultaneously a) accomplish the synthesis of visually realistic and temporally coher... | closed | completed | false | 4 | [
"New pipeline/model"
] | [] | 2023-10-21T23:01:17Z | 2024-11-25T13:07:07Z | 2024-11-25T13:07:07Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | clarencechen | 8,482,341 | MDQ6VXNlcjg0ODIzNDE= | User | false |
huggingface/diffusers | 1,955,693,440 | I_kwDOHa8MBc50kYOA | 5,476 | https://github.com/huggingface/diffusers/issues/5476 | https://api.github.com/repos/huggingface/diffusers/issues/5476 | Is a DreamBooth model wegiths trained with xformers_memory_efficient_attention could be loaded into a model without xformers_memory_efficient_attention? | I am training DreamBooth following https://github.com/huggingface/diffusers/blob/main/examples/custom_diffusion/README.md. Could a DreamBooth model wegiths trained with xformers_memory_efficient_attention be loaded into a model without xformers_memory_efficient_attention? Thanks. | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-21T23:09:07Z | 2023-11-29T15:05:33Z | 2023-11-29T15:05:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | g-jing | 44,223,191 | MDQ6VXNlcjQ0MjIzMTkx | User | false |
huggingface/diffusers | 1,376,950,403 | I_kwDOHa8MBc5SEpiD | 548 | https://github.com/huggingface/diffusers/issues/548 | https://api.github.com/repos/huggingface/diffusers/issues/548 | Apple MPS image generation times take much longer with 0.4.0 | ### Describe the bug
It used to take about 90 seconds to generate a 768 x 512 image with diffusers 0.3.0. I checked out the main branch with 0.4.0dev0 today and the time has increased to about 600 seconds and that too for only 20 steps, as opposed to 75 steps initially. It feels as if this build uses both the CPU an... | closed | completed | false | 17 | [
"bug",
"stale"
] | [
"pcuenca"
] | 2022-09-18T05:25:35Z | 2022-11-23T15:03:32Z | 2022-11-23T15:03:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FahimF | 181,110 | MDQ6VXNlcjE4MTExMA== | User | false |
huggingface/diffusers | 1,955,794,840 | I_kwDOHa8MBc50kw-Y | 5,480 | https://github.com/huggingface/diffusers/issues/5480 | https://api.github.com/repos/huggingface/diffusers/issues/5480 | Request implementation of EDM | ### Model/Pipeline/Scheduler description
https://github.com/NVlabs/edm
These schedulers looks great! The results in paper demonstrated better performance than popular scheduler DDPM/DDIM.
### Open source status
- [X] The model implementation is available
- [ ] The model weights are available (Only relevant if add... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-10-22T07:27:17Z | 2023-11-29T15:05:30Z | 2023-11-29T15:05:29Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luciennnnnnn | 20,135,317 | MDQ6VXNlcjIwMTM1MzE3 | User | false |
huggingface/diffusers | 1,955,815,006 | I_kwDOHa8MBc50k15e | 5,481 | https://github.com/huggingface/diffusers/issues/5481 | https://api.github.com/repos/huggingface/diffusers/issues/5481 | SDXL T2I Adapter Inpainting Support | ### Model/Pipeline/Scheduler description
Greetings contributors!
I was wondering if `diffusers` currently supports inpainting with T2I adapters (with SDXL in my case)? If not, would a new pipeline be required to achieve this functionality, or is it not technically possible?
Thanks! | closed | completed | false | 3 | [
"stale"
] | [] | 2023-10-22T08:40:34Z | 2024-01-09T15:07:54Z | 2024-01-09T15:07:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jakebabbidge | 26,162,335 | MDQ6VXNlcjI2MTYyMzM1 | User | false |
huggingface/diffusers | 1,956,317,307 | I_kwDOHa8MBc50mwh7 | 5,484 | https://github.com/huggingface/diffusers/issues/5484 | https://api.github.com/repos/huggingface/diffusers/issues/5484 | [xformers] training is broken with `xformers` and PyTorch 2.1 | Related issue: https://github.com/huggingface/diffusers/issues/5368
When using PyTorch 2.1 and the latest stable build of `xformers`, our [DreamBooth LoRA script for SDXL](https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/README_sdxl.md) doesn't work. #5368 provides more details.
But when us... | closed | completed | false | 2 | [] | [] | 2023-10-23T04:14:52Z | 2023-12-26T15:10:15Z | 2023-12-26T15:10:15Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,956,424,451 | I_kwDOHa8MBc50nKsD | 5,485 | https://github.com/huggingface/diffusers/issues/5485 | https://api.github.com/repos/huggingface/diffusers/issues/5485 | ControlNet batch inference takes forever | ### Describe the bug
I tried to do batch inference for ControlNet following [the official example](https://github.com/huggingface/diffusers/tree/main/examples/controlnet#performing-inference-with-the-trained-controlnet). It works fine to infer only 1 image as in the example. However, batch inference (_e.g._, `batch_si... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-10-23T06:10:45Z | 2024-01-15T15:07:54Z | 2024-01-15T15:07:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | doem97 | 19,631,039 | MDQ6VXNlcjE5NjMxMDM5 | User | false |
huggingface/diffusers | 1,956,590,711 | I_kwDOHa8MBc50nzR3 | 5,487 | https://github.com/huggingface/diffusers/issues/5487 | https://api.github.com/repos/huggingface/diffusers/issues/5487 | segmind/SSD-1B training on train_dreambooth_lora_sdxl.py & train_text_to_image_lora_sdxl.py | Can allow the script to run on distilled SDXL? the script can't run because of fp16 safetensor in the text_encoders,vae and unets of segmind/SSD-1B. https://huggingface.co/segmind/SSD-1B | closed | completed | false | 4 | [] | [] | 2023-10-23T08:06:34Z | 2023-10-24T04:34:08Z | 2023-10-24T01:33:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | danieltanhx | 60,249,161 | MDQ6VXNlcjYwMjQ5MTYx | User | false |
huggingface/diffusers | 1,956,714,953 | I_kwDOHa8MBc50oRnJ | 5,488 | https://github.com/huggingface/diffusers/issues/5488 | https://api.github.com/repos/huggingface/diffusers/issues/5488 | the demo code generate one image is not work | ### Describe the bug
my diffusers version is
```
diffusers 0.21.4
nvidia-cublas-cu12 12.1.3.1
nvidia-cuda-cupti-cu12 12.1.105
nvidia-cuda-nvrtc-cu12 12.1.105
nvidia-cuda-runtime-cu12 12.1.105
nvidia-cudnn-cu12 8.9.2.26
nvidia-cufft-cu12 11.0.2.54
nvidia-curand-cu12 ... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-10-23T09:15:11Z | 2024-01-22T08:54:43Z | 2024-01-09T06:38:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ryzn0518 | 19,700,467 | MDQ6VXNlcjE5NzAwNDY3 | User | false |
huggingface/diffusers | 1,956,798,843 | I_kwDOHa8MBc50omF7 | 5,489 | https://github.com/huggingface/diffusers/issues/5489 | https://api.github.com/repos/huggingface/diffusers/issues/5489 | 🤗 PEFT x 🧨 diffusers -- integration alert 🔥 | Dear community members,
Over the past few weeks, @younesbelkada and @pacman100 helped us integrate `peft` into `diffusers`. Taking advantage of this integration, users can easily perform multi-adapter inference with control over `scale`, switching between different adapters, weighted adapter inference, etc.
Belo... | closed | completed | false | 31 | [
"stale"
] | [] | 2023-10-23T10:00:37Z | 2024-08-12T12:57:47Z | 2023-12-11T03:38:12Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,956,914,236 | I_kwDOHa8MBc50pCQ8 | 5,490 | https://github.com/huggingface/diffusers/issues/5490 | https://api.github.com/repos/huggingface/diffusers/issues/5490 | [Request]Add swiglu activation function support | Add in swiglu support for attention.py and activations.py please. swiglu is the go to activation function for many SOTA LLM models and audio models. | closed | completed | false | 5 | [
"stale"
] | [] | 2023-10-23T11:08:05Z | 2024-01-15T15:07:52Z | 2024-01-15T15:07:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ezra-ch | 69,337,319 | MDQ6VXNlcjY5MzM3MzE5 | User | false |
huggingface/diffusers | 1,957,099,647 | I_kwDOHa8MBc50pvh_ | 5,492 | https://github.com/huggingface/diffusers/issues/5492 | https://api.github.com/repos/huggingface/diffusers/issues/5492 | stabilityai/stable-diffusion-2-base is missing fp16 files | Hi Team,
My team is trying to fine-tune a sample and we are using model **stabilityai/stable-diffusion-2-base** . Due to this we are stuck at the **Training setting** phase.
Command we are using
!accelerate launch train_dreambooth.py \
--pretrained_model_name_or_path=$MODEL_NAME \
--pretrained_vae_name_... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-10-23T12:53:28Z | 2023-12-02T15:05:48Z | 2023-12-02T15:05:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ranjansingh11 | 145,988,344 | U_kgDOCLOa-A | User | false |
huggingface/diffusers | 1,957,133,349 | I_kwDOHa8MBc50p3wl | 5,494 | https://github.com/huggingface/diffusers/issues/5494 | https://api.github.com/repos/huggingface/diffusers/issues/5494 | inpaint background with kandinsky2.2 | the inpaint background is failed, what's wrong?
![download (7)](https://github.com/huggingface/diffusers/assets/11... | closed | completed | false | 4 | [
"stale"
] | [
"yiyixuxu"
] | 2023-10-23T13:11:26Z | 2024-01-15T15:07:50Z | 2024-01-15T15:07:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | crapthings | 1,147,704 | MDQ6VXNlcjExNDc3MDQ= | User | false |
huggingface/diffusers | 1,957,169,653 | I_kwDOHa8MBc50qAn1 | 5,495 | https://github.com/huggingface/diffusers/issues/5495 | https://api.github.com/repos/huggingface/diffusers/issues/5495 | Unclear sentence in docs about torch.compile and image size | I was reading the [docs](https://huggingface.co/docs/diffusers/optimization/torch2.0#using-accelerated-transformers-and-torchcompile) about using `torch.compile` and saw the following sentence:
> Calling the compiled pipeline on a different image size will re-trigger compilation which can be expensive.
What exact... | closed | completed | false | 3 | [] | [] | 2023-10-23T13:27:30Z | 2023-10-24T06:39:04Z | 2023-10-24T06:39:03Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | holwech | 7,474,295 | MDQ6VXNlcjc0NzQyOTU= | User | false |
huggingface/diffusers | 1,958,578,537 | I_kwDOHa8MBc50vYlp | 5,502 | https://github.com/huggingface/diffusers/issues/5502 | https://api.github.com/repos/huggingface/diffusers/issues/5502 | `train_custom_diffusion.py` does not support `fp16` | when I run train_custom_diffusion.py by following args, there will be a bug report
```shell
python3 custom_diffusion/train_custom_diffusion.py
--pretrained_model_name_or_path CompVis/stable-diffusion-v1-4
--instance_data_dir diffusers/cat_toy_example
--prior_loss_weight 1.0
--class_prompt "cat"
--... | closed | completed | false | 4 | [
"stale"
] | [
"DN6"
] | 2023-10-24T06:05:55Z | 2023-12-04T07:13:48Z | 2023-12-04T07:13:48Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiaqiw09 | 60,021,713 | MDQ6VXNlcjYwMDIxNzEz | User | false |
huggingface/diffusers | 1,958,580,681 | I_kwDOHa8MBc50vZHJ | 5,503 | https://github.com/huggingface/diffusers/issues/5503 | https://api.github.com/repos/huggingface/diffusers/issues/5503 | ControlNet Scribble with SDXL | Hello! I would like to ask if there will be support for Controlnet scribble with SDXL. Currently I have only seen models for canny, depth and openpose. | closed | completed | false | 3 | [] | [
"DN6"
] | 2023-10-24T06:08:18Z | 2023-11-09T07:51:42Z | 2023-11-09T07:51:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rafationgson | 70,418,702 | MDQ6VXNlcjcwNDE4NzAy | User | false |
huggingface/diffusers | 1,958,688,497 | I_kwDOHa8MBc50vzbx | 5,504 | https://github.com/huggingface/diffusers/issues/5504 | https://api.github.com/repos/huggingface/diffusers/issues/5504 | PEFT LoRA not working with xformers. | ### Describe the bug
PEFT LoRA not working with xformers.
### Reproduction
```python
from diffusers import DiffusionPipeline
import torch
pipe = DiffusionPipeline.from_pretrained("./models/checkpoint/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16)
pipe.load_lora_weights("CiroN2022/toy-face", weight_n... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-10-24T07:41:05Z | 2023-10-25T06:21:58Z | 2023-10-25T06:21:58Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AnyISalIn | 16,937,838 | MDQ6VXNlcjE2OTM3ODM4 | User | false |
huggingface/diffusers | 1,958,753,630 | I_kwDOHa8MBc50wDVe | 5,507 | https://github.com/huggingface/diffusers/issues/5507 | https://api.github.com/repos/huggingface/diffusers/issues/5507 | `torch.aucast()` only adapat to 'cuda' | in diffusers/examples, the tranning files always use `torch.autocast('cuda') ` or `torch.cuda.amp.autocast()`, which lacks some flexbility.
I think it's better to replace these functions with `torch.autocast(device_type=accelerator.device.type)`
@patrickvonplaten
If you need, I can submit a PR quickly. | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-24T08:23:36Z | 2023-12-02T15:05:40Z | 2023-12-02T15:05:40Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiaqiw09 | 60,021,713 | MDQ6VXNlcjYwMDIxNzEz | User | false |
huggingface/diffusers | 1,959,136,044 | I_kwDOHa8MBc50xgss | 5,509 | https://github.com/huggingface/diffusers/issues/5509 | https://api.github.com/repos/huggingface/diffusers/issues/5509 | Using np.ndarray or cv2 image as image doesn't work: image batch size must be same as prompt batch size | ### Describe the bug
image parameter accepts np.ndarray but doesn't work with images loaded from `cv2.imread()` or PIL images converted with `np.array(image)`
https://huggingface.co/docs/diffusers/v0.21.0/en/api/pipelines/controlnet#diffusers.StableDiffusionControlNetPipeline
> image (torch.FloatTensor, PIL.Imag... | closed | completed | false | 3 | [
"bug",
"stale"
] | [
"DN6"
] | 2023-10-24T12:20:22Z | 2024-01-15T15:07:48Z | 2024-01-15T15:07:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | geroldmeisinger | 112,266,044 | U_kgDOBrELPA | User | false |
huggingface/diffusers | 1,377,042,621 | I_kwDOHa8MBc5SFAC9 | 551 | https://github.com/huggingface/diffusers/issues/551 | https://api.github.com/repos/huggingface/diffusers/issues/551 | Merging Stable diffusion pipelines just makes sense | [Following the Philosophy](https://github.com/huggingface/diffusers/issues/307), it has been decided to keep different pipelines for Stable Diffusion for txt-to-img, img-to-img and inpainting.
Here is the result:
* PR #549 : **code duplicated 4 times (onnx included)**
* PR #537 : **modification only applied to t... | closed | completed | false | 39 | [
"stale",
"Good Example PR"
] | [
"patrickvonplaten"
] | 2022-09-18T12:37:18Z | 2024-04-01T17:41:25Z | 2022-11-13T22:54:32Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | leszekhanusz | 4,641,499 | MDQ6VXNlcjQ2NDE0OTk= | User | false |
huggingface/diffusers | 1,959,368,807 | I_kwDOHa8MBc50yZhn | 5,510 | https://github.com/huggingface/diffusers/issues/5510 | https://api.github.com/repos/huggingface/diffusers/issues/5510 | RuntimeError: Input type (c10::Half) and bias type (float) should be the same when running StableDiffusionXLInstructPix2PixPipeline | ### Describe the bug
when I run the example of StableDiffusionXLInstructPix2PixPipeline, an error has occurred.
As I found that the previous commit https://github.com/huggingface/diffusers/pull/4796 had already resolved this problem, so I used the latest code from the main branch, but there was still the same error
... | closed | completed | false | 4 | [
"bug",
"stale"
] | [
"DN6"
] | 2023-10-24T14:18:16Z | 2023-12-26T15:09:54Z | 2023-12-26T15:09:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kisisjrlly | 34,215,214 | MDQ6VXNlcjM0MjE1MjE0 | User | false |
huggingface/diffusers | 1,960,170,820 | I_kwDOHa8MBc501dVE | 5,515 | https://github.com/huggingface/diffusers/issues/5515 | https://api.github.com/repos/huggingface/diffusers/issues/5515 | LCM pipeline fails when initialized without scheduler | ### Describe the bug
I installed the latest version of diffusers on colab after #5448 was merged. The LCM pipeline fails to initialize when using the code present in [docs](https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/pipelines/latent_consistency_models.md) with the error:
```
ValueError... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-10-24T22:19:51Z | 2023-10-25T06:33:03Z | 2023-10-25T06:33:03Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | a-r-r-o-w | 72,266,394 | MDQ6VXNlcjcyMjY2Mzk0 | User | false |
huggingface/diffusers | 1,960,276,947 | I_kwDOHa8MBc5013PT | 5,516 | https://github.com/huggingface/diffusers/issues/5516 | https://api.github.com/repos/huggingface/diffusers/issues/5516 | Lora not functioning when used with t2i adapters Pipeline | ### Describe the bug
Hello,
Thank you for this useful library.
I have a small problem, I managed to use the code to generate images using SDXL with t2i, then an Image with a Lora. But for some reasons not with both.
When I load the lora using the StableDiffusionXLPipeline It does work but not when I do so with ... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-10-25T00:04:58Z | 2023-11-01T20:25:40Z | 2023-11-01T20:25:40Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ilisparrow | 4,880,273 | MDQ6VXNlcjQ4ODAyNzM= | User | false |
huggingface/diffusers | 1,960,612,060 | I_kwDOHa8MBc503JDc | 5,518 | https://github.com/huggingface/diffusers/issues/5518 | https://api.github.com/repos/huggingface/diffusers/issues/5518 | Using Loras AttributeError: 'ModuleList' object has no attribute 'time' | ### Describe the bug
I have this code
```
vae = AutoencoderKL.from_pretrained("stabilityai/sd-vae-ft-mse")
pipe = StableDiffusionPipeline.from_pretrained(
'/content/drive/MyDrive/model',
safety_checker=None,
vae=vae
)
pipe.load_lora_weights(".", weight_name="/content/drive/MyDrive/loras/tomhol... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-10-25T06:06:01Z | 2023-10-29T05:37:58Z | 2023-10-29T05:37:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jamesnogra | 6,120,605 | MDQ6VXNlcjYxMjA2MDU= | User | false |
huggingface/diffusers | 1,960,674,111 | I_kwDOHa8MBc503YM_ | 5,519 | https://github.com/huggingface/diffusers/issues/5519 | https://api.github.com/repos/huggingface/diffusers/issues/5519 | Network unreachable | Hi,
I'm using dreambooth on HPC machine. After installing and configuration as required, when launched on compute node dreambooth get "Network unreachable" error. I think is trying to download the models, but compute node are not opened to internet. Is there a way to download such models before start and to indicate w... | closed | completed | false | 3 | [
"stale"
] | [
"DN6"
] | 2023-10-25T06:50:23Z | 2023-12-02T15:05:35Z | 2023-12-02T15:05:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | unrue | 51,996,285 | MDQ6VXNlcjUxOTk2Mjg1 | User | false |
huggingface/diffusers | 1,960,703,166 | I_kwDOHa8MBc503fS- | 5,520 | https://github.com/huggingface/diffusers/issues/5520 | https://api.github.com/repos/huggingface/diffusers/issues/5520 | Error while deserializing header: HeaderTooLarge | Hi,
I'm using dreambooth on HPC machine, using dog test case after all installation and configuration steps, according with the manual. At launch, I get the following error:
```
File "train_dreambooth.py", line 1422, in <module>
main(args)
File "train_dreambooth.py", line 914, in main
text_encoder =... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-25T07:09:13Z | 2023-12-02T15:05:33Z | 2023-12-02T15:05:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | unrue | 51,996,285 | MDQ6VXNlcjUxOTk2Mjg1 | User | false |
huggingface/diffusers | 1,960,945,649 | I_kwDOHa8MBc504afx | 5,522 | https://github.com/huggingface/diffusers/issues/5522 | https://api.github.com/repos/huggingface/diffusers/issues/5522 | TypeError: PeftAdapterMixin.load_adapter() got an unexpected keyword argument 'adapter_state_dict' | ### Describe the bug
When loading LoRAs with the new multiple LoRAs adapter, some SDXL LoRAs fail loading with this error:
`TypeError: PeftAdapterMixin.load_adapter() got an unexpected keyword argument 'adapter_state_dict'`
Ex.
https://civitai.com/models/57437/andrei-tarkovsky-style
https://civitai.com/models/1... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-10-25T09:24:50Z | 2023-10-30T18:35:47Z | 2023-10-30T18:35:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,961,038,241 | I_kwDOHa8MBc504xGh | 5,523 | https://github.com/huggingface/diffusers/issues/5523 | https://api.github.com/repos/huggingface/diffusers/issues/5523 | Results were inconsistent for each run | ### Describe the bug
When i run the same code, the sd result is slight different each time, even if set the random seed.
```
import os
import torch
import numpy as np
import cv2
import random
seed = 0
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.cu... | closed | completed | false | 2 | [
"bug"
] | [
"DN6"
] | 2023-10-25T10:14:28Z | 2023-10-30T07:29:31Z | 2023-10-30T07:29:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Dengshunge | 40,339,713 | MDQ6VXNlcjQwMzM5NzEz | User | false |
huggingface/diffusers | 1,961,211,770 | I_kwDOHa8MBc505bd6 | 5,524 | https://github.com/huggingface/diffusers/issues/5524 | https://api.github.com/repos/huggingface/diffusers/issues/5524 | Loading ex. Würstchen: AttributeError: 'Linear' object has no attribute 'lora_layer' | ### Describe the bug
Loading "other" models like Wüerstchen worked fine before, but now an error is raised:
`AttributeError: 'Linear' object has no attribute 'lora_layer'`
### Reproduction
```
from diffusers import AutoPipelineForText2Image
from diffusers.pipelines.wuerstchen import DEFAULT_STAGE_C_TIMESTEPS
pi... | closed | completed | false | 17 | [
"bug"
] | [] | 2023-10-25T11:52:27Z | 2024-03-12T05:12:52Z | 2023-11-13T17:37:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,961,423,849 | I_kwDOHa8MBc506PPp | 5,529 | https://github.com/huggingface/diffusers/issues/5529 | https://api.github.com/repos/huggingface/diffusers/issues/5529 | Bug in train_text_to_image_sdxl.py | ### Describe the bug
Every time I try to run `train_text_to_image_sdxl.py` I get the following error during mapping:
```
pyarrow.lib.ArrowInvalid: Column 2 named prompt_embeds expected length 1000 but got length 999
```
I tried to add batch_size=1 but it caused even more errors further.
### Reproduction
... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-10-25T13:24:13Z | 2023-10-29T18:42:30Z | 2023-10-29T18:42:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | julkaztwittera | 96,252,562 | U_kgDOBbyykg | User | false |
huggingface/diffusers | 1,377,116,710 | I_kwDOHa8MBc5SFSIm | 553 | https://github.com/huggingface/diffusers/issues/553 | https://api.github.com/repos/huggingface/diffusers/issues/553 | `test_ddpm_ddim_equality` fails if manual seed changed to 4 | ### Describe the bug
Test `PipelineTesterMixin::test_ddpm_ddim_equality` in `test_pipelines.py` fails if `generator` is set to `torch.manual_seed(4)` instead of `torch.manual_seed(0)`.
So the fact that the test passes for seed equal 0 is most likely accidental, and the difference be fluctuate above or below `0.1... | closed | completed | false | 5 | [
"bug"
] | [
"anton-l"
] | 2022-09-18T16:52:05Z | 2022-11-02T13:50:34Z | 2022-11-02T13:50:34Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sgrigory | 23,200,558 | MDQ6VXNlcjIzMjAwNTU4 | User | false |
huggingface/diffusers | 1,961,924,845 | I_kwDOHa8MBc508Jjt | 5,532 | https://github.com/huggingface/diffusers/issues/5532 | https://api.github.com/repos/huggingface/diffusers/issues/5532 | Bug in DPMSolverSDEScheduler | ### Describe the bug
Inside scheduler function `self.noise_sampler = BrownianTreeNoiseSampler` is caching and if you run pipe with images with different shapes, then you got an error:
```
prev_sample = prev_sample + self.noise_sampler(sigma_fn(t), sigma_fn(t_next)) * s_noise * sigma_up
RuntimeError: The size of t... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-10-25T17:44:48Z | 2023-12-08T15:05:48Z | 2023-12-08T15:05:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | attashe | 8,243,605 | MDQ6VXNlcjgyNDM2MDU= | User | false |
huggingface/diffusers | 1,962,558,288 | I_kwDOHa8MBc50-kNQ | 5,534 | https://github.com/huggingface/diffusers/issues/5534 | https://api.github.com/repos/huggingface/diffusers/issues/5534 | dataset.transform() hangs indefinitely while finetuning the stable diffusion XL | ### Describe the bug
Multi-GPU fine-tuning the stable diffusion X by following https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/README_sdxl.md hangs indefinitely.
### Reproduction
accelerate launch train_text_to_image_sdxl.py --pretrained_model_name_or_path=$MODEL_NAME --pretrained_v... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-10-26T01:49:03Z | 2023-10-30T17:39:38Z | 2023-10-30T17:35:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bhosalems | 10,846,405 | MDQ6VXNlcjEwODQ2NDA1 | User | false |
huggingface/diffusers | 1,962,829,405 | I_kwDOHa8MBc50_mZd | 5,537 | https://github.com/huggingface/diffusers/issues/5537 | https://api.github.com/repos/huggingface/diffusers/issues/5537 | SDXL_DreamBooth_LoRA_.ipynb no longer working | ### Describe the bug
Hello !
SDXL_DreamBooth_LoRA_.ipynb is no longer working.
Seems to be incompatible with latest pytorch or xformers.
### Reproduction
Try to launch the default notebook on Google Colab T4 or copy paste the code in Kaggle T4.
### Logs
```shell
On Google Colab :
WARNING[XFORMERS]: xForm... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-10-26T06:52:22Z | 2023-12-08T15:05:46Z | 2023-12-08T15:05:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghost | 10,137 | MDQ6VXNlcjEwMTM3 | User | false |
huggingface/diffusers | 1,963,238,174 | I_kwDOHa8MBc51BKMe | 5,538 | https://github.com/huggingface/diffusers/issues/5538 | https://api.github.com/repos/huggingface/diffusers/issues/5538 | Why is the pipeline_stable_diffusion_upscale.py file not using the encoder-decoder latent? | ### Describe the bug
There is no training script for pipeline_stable_diffusion_upscale.py because the authors chose not to utilize the latent domain for the Super-resolution task. Additionally, the U-Net implemented in pipeline_stable_diffusion_upscale.py only accepts 7 channels. How is this achieved?
### Reproductio... | closed | completed | false | 3 | [
"question",
"stale"
] | [] | 2023-10-26T10:47:10Z | 2023-12-08T15:05:44Z | 2023-12-08T15:05:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AnasHXH | 65,374,131 | MDQ6VXNlcjY1Mzc0MTMx | User | false |
huggingface/diffusers | 1,963,450,661 | I_kwDOHa8MBc51B-El | 5,540 | https://github.com/huggingface/diffusers/issues/5540 | https://api.github.com/repos/huggingface/diffusers/issues/5540 | Inference out of memory | I successfully trained Dreambooth with dog dataset test using HPc machine having Tesla V100 card, and pytorch 1.10.1, and now I'm trying to make inference from the manual, having a Cuda out of memory error:
```
from diffusers import StableDiffusionPipeline
import torch
model_id = "out_train"
pipe = StableDiffu... | closed | completed | false | 9 | [
"stale"
] | [] | 2023-10-26T12:25:34Z | 2023-12-26T15:09:46Z | 2023-12-26T15:09:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | unrue | 51,996,285 | MDQ6VXNlcjUxOTk2Mjg1 | User | false |
huggingface/diffusers | 1,963,705,183 | I_kwDOHa8MBc51C8Nf | 5,542 | https://github.com/huggingface/diffusers/issues/5542 | https://api.github.com/repos/huggingface/diffusers/issues/5542 | Add DPM++ 2M Karras sampler to DiffusionPipeline | I need to use the DPM++ 2M Karras sampler when using DiffusionPipeline with custom pipeline "long-prompt-weighting-stable-diffusion". i.e.:
from diffusers import DiffusionPipeline
...
image = pipe.text2img(...)
The text2img call should use DPM++ 2M Karras sampler
| closed | completed | false | 4 | [] | [] | 2023-10-26T14:26:40Z | 2023-10-26T18:54:34Z | 2023-10-26T18:54:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Kenshiro-28 | 31,041,399 | MDQ6VXNlcjMxMDQxMzk5 | User | false |
huggingface/diffusers | 1,963,765,294 | I_kwDOHa8MBc51DK4u | 5,544 | https://github.com/huggingface/diffusers/issues/5544 | https://api.github.com/repos/huggingface/diffusers/issues/5544 | Disabling FreeU in StableDiffusionXLPipeline if its not enabled throws an error | ### Describe the bug
If you disable FreeU before enabling it you get an error, at least of SDXL not relied any others
```
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
[<ipython-input-5-210568d4bac8>](https:... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-10-26T14:56:20Z | 2023-10-27T15:59:13Z | 2023-10-27T15:59:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Vargol | 62,868 | MDQ6VXNlcjYyODY4 | User | false |
huggingface/diffusers | 1,963,862,901 | I_kwDOHa8MBc51Dit1 | 5,545 | https://github.com/huggingface/diffusers/issues/5545 | https://api.github.com/repos/huggingface/diffusers/issues/5545 | I have a problem when using StableDiffusionInpaintPipeline() | ### Describe the bug
pipeline = StableDiffusionInpaintPipeline.from_pretrained(
"stable-diffusion-inpainting"
)
Loading pipeline components...: 29%|██▊ | 2/7 [02:26<06:06, 73.22s/it]
Traceback (most recent call last):
File "/home/bentle/anaconda3/envs/aigc/lib/python3.8/site-packages/diffusers/models... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-10-26T15:48:17Z | 2023-12-08T15:05:38Z | 2023-12-08T15:05:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bent1e | 50,535,454 | MDQ6VXNlcjUwNTM1NDU0 | User | false |
huggingface/diffusers | 1,963,911,095 | I_kwDOHa8MBc51Due3 | 5,546 | https://github.com/huggingface/diffusers/issues/5546 | https://api.github.com/repos/huggingface/diffusers/issues/5546 | Wrong amount of steps for some schedulers with base+refiner ensemble when use_karras_sigmas=True | ### Describe the bug
**DEISMultistepScheduler, UniPCMultistepScheduler, DPMSolverSinglestepScheduler, EulerDiscreteScheduler, DPMSolverMultistepScheduler, LMSDiscreteScheduler: Does 33 steps total instead of 40**
Base: 25, Refiner: 8
**HeunDiscreteScheduler: Does 33 steps total and reports strange steps**
Base:... | closed | completed | false | 12 | [
"bug",
"stale"
] | [
"yiyixuxu"
] | 2023-10-26T16:17:55Z | 2023-12-26T15:09:44Z | 2023-12-26T15:09:44Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TimothyAlexisVass | 55,708,319 | MDQ6VXNlcjU1NzA4MzE5 | User | false |
huggingface/diffusers | 1,964,126,856 | I_kwDOHa8MBc51EjKI | 5,548 | https://github.com/huggingface/diffusers/issues/5548 | https://api.github.com/repos/huggingface/diffusers/issues/5548 | [HacktoberFest][Documentation] Diffusers documentation to Portuguese | Hi!
I am opening this issue to encourage people translating this project, and other AI knowledge to Portuguese.
I would like to contribute to this too, specially because of this year's hacktoberfest 🎃.
For this, you'll need:
- copy english docs from [source/en](https://github.com/huggingface/diffusers/blob/m... | closed | completed | false | 1 | [] | [] | 2023-10-26T18:30:30Z | 2023-10-27T17:51:37Z | 2023-10-27T17:51:37Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | SirMonteiro | 29,388,757 | MDQ6VXNlcjI5Mzg4NzU3 | User | false |
huggingface/diffusers | 1,964,699,537 | I_kwDOHa8MBc51Gu-R | 5,551 | https://github.com/huggingface/diffusers/issues/5551 | https://api.github.com/repos/huggingface/diffusers/issues/5551 | Can any pre-trained model be inverted? | I'm trying to invert face images, similar to diffae. I want to know if all pre-trained models have the ability to do image2latent, and then latent2image | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-27T03:29:40Z | 2023-12-08T15:05:35Z | 2023-12-08T15:05:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pikeyang | 59,758,624 | MDQ6VXNlcjU5NzU4NjI0 | User | false |
huggingface/diffusers | 1,965,091,976 | I_kwDOHa8MBc51IOyI | 5,553 | https://github.com/huggingface/diffusers/issues/5553 | https://api.github.com/repos/huggingface/diffusers/issues/5553 | DiffusionPipeline fails to load custom model | ### Describe the bug
I was exploring [this](https://github.com/SUDO-AI-3D/zero123plus) repository. It works well when using `diffusers==0.20.2`. To make it work with the latest diffusers release (0.21.4), you have to modify the pipeline file a bit, but that's not part of what's causing the issue.
The issue is mai... | closed | completed | false | 7 | [
"bug"
] | [
"patrickvonplaten"
] | 2023-10-27T09:06:35Z | 2023-11-08T10:59:31Z | 2023-10-30T18:32:13Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | a-r-r-o-w | 72,266,394 | MDQ6VXNlcjcyMjY2Mzk0 | User | false |
huggingface/diffusers | 1,965,159,708 | I_kwDOHa8MBc51IfUc | 5,554 | https://github.com/huggingface/diffusers/issues/5554 | https://api.github.com/repos/huggingface/diffusers/issues/5554 | Converting bin models to fp16? | Hi!
can someone please tell me how can i convert fp32 diffuser models [the ones that have unet vae tokenizer etc.. folders that are used for stable diffusion] to fp16?
i can't find any tutorial or script but i can see some people converted their bin files in smaller size.
i need it cause i want to convert LCM drea... | closed | completed | false | 4 | [] | [] | 2023-10-27T09:47:46Z | 2023-10-28T13:23:01Z | 2023-10-28T13:23:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Amin456789 | 145,983,752 | U_kgDOCLOJCA | User | false |
huggingface/diffusers | 1,965,363,393 | I_kwDOHa8MBc51JRDB | 5,557 | https://github.com/huggingface/diffusers/issues/5557 | https://api.github.com/repos/huggingface/diffusers/issues/5557 | ValueError: Unexpected keyword arguments: use_reentrant | Hi,
I'm training Dreambooth with dog dataset test, (Prior-reservation loss case) using HPC machine having Tesla V100 card, pytorch 1.10.1,diffusers 0.22.0.dev0, accelerate 0.20.3. The following is the launch command as the manual suggest:
```
accelerate launch train_dreambooth.py \
--pretrained_model_name_or... | closed | completed | false | 4 | [] | [] | 2023-10-27T11:53:32Z | 2023-11-02T00:23:26Z | 2023-10-27T13:04:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | unrue | 51,996,285 | MDQ6VXNlcjUxOTk2Mjg1 | User | false |
huggingface/diffusers | 1,965,384,747 | I_kwDOHa8MBc51JWQr | 5,558 | https://github.com/huggingface/diffusers/issues/5558 | https://api.github.com/repos/huggingface/diffusers/issues/5558 | Artefacts When Using StableDiffusionControlNetInpaintPipeline with Multiple ControlNets | ### Describe the bug
I'm encountering an issue with the **StableDiffusionControlNetInpaintPipeline** when I use multiple ControlNet models, specifically the inpaint with openpose model. The images I'm getting have unexpected openpose patterns popping up. I'm seeing shapes and patterns that I didn't anticipate.
... | closed | completed | false | 5 | [
"bug",
"stale"
] | [
"yiyixuxu"
] | 2023-10-27T12:05:00Z | 2023-11-27T13:00:06Z | 2023-11-27T13:00:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vadsuh | 34,363,498 | MDQ6VXNlcjM0MzYzNDk4 | User | false |
huggingface/diffusers | 1,377,287,307 | I_kwDOHa8MBc5SF7yL | 556 | https://github.com/huggingface/diffusers/issues/556 | https://api.github.com/repos/huggingface/diffusers/issues/556 | Image output tiling for seamless textures with Stable Diffusion | **Is your feature request related to a problem? Please describe.**
Currently there is no way to create seamless textures with Stable Diffusion, a crucial feature that is missing.
**Describe the solution you'd like**
Something similar to this pull on the sd-webui repo: https://github.com/sd-webui/stable-diffusion-... | closed | completed | false | 95 | [
"community-examples",
"contributions-welcome"
] | [] | 2022-09-19T02:04:33Z | 2024-12-20T10:57:02Z | 2024-11-25T17:08:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | torrinworx | 82,110,564 | MDQ6VXNlcjgyMTEwNTY0 | User | false |
huggingface/diffusers | 1,965,642,482 | I_kwDOHa8MBc51KVLy | 5,560 | https://github.com/huggingface/diffusers/issues/5560 | https://api.github.com/repos/huggingface/diffusers/issues/5560 | Poor quality results | Hi,
I'm using this manual on 1 GPU V100:
`https://huggingface.co/docs/diffusers/main/en/training/dreambooth#performing-inference-using-a-saved-checkpoint`
using dog dataset and finetuning the text encoder. "A dog in a bucket" inference case, retrieve very low quality results, very far from the expected result.... | closed | completed | false | 10 | [
"stale"
] | [
"sayakpaul"
] | 2023-10-27T14:22:54Z | 2023-12-08T15:05:29Z | 2023-12-08T15:05:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | unrue | 51,996,285 | MDQ6VXNlcjUxOTk2Mjg1 | User | false |
huggingface/diffusers | 1,965,816,334 | I_kwDOHa8MBc51K_oO | 5,562 | https://github.com/huggingface/diffusers/issues/5562 | https://api.github.com/repos/huggingface/diffusers/issues/5562 | Make 🤗 Diffusers tests device-agnostic | ## Feature Request
Make the Diffusers test suite device-agnostic, so alternative accelerator devices are able to easily run the repo tests.
### Motivation
At the moment, Diffusers test suite explicitly supports `cuda` and `mps` devices. #4673 has made this less explicit, and other devices supporting PyTorch can ... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-10-27T16:09:45Z | 2023-12-08T15:05:27Z | 2023-12-08T15:05:27Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | arsalanu | 49,171,201 | MDQ6VXNlcjQ5MTcxMjAx | User | false |
huggingface/diffusers | 1,966,266,697 | I_kwDOHa8MBc51MtlJ | 5,563 | https://github.com/huggingface/diffusers/issues/5563 | https://api.github.com/repos/huggingface/diffusers/issues/5563 | SDXL DreamBooth LoRA training throws errors | ### Describe the bug
SDXL DreamBooth LoRA training colab notebook doesn't work properly. It throws an error at the `!accelerate launch train_dreambooth_lora_sdxl.py` section.
### Reproduction
You can just ran the [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/SDXL_Dream... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-10-27T22:33:27Z | 2023-12-14T00:23:11Z | 2023-12-08T15:05:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | unalkalkan | 9,661,442 | MDQ6VXNlcjk2NjE0NDI= | User | false |
huggingface/diffusers | 1,966,458,413 | I_kwDOHa8MBc51NcYt | 5,564 | https://github.com/huggingface/diffusers/issues/5564 | https://api.github.com/repos/huggingface/diffusers/issues/5564 | Benchmarking SDXL with the new TensorRT compilation | During the PyTorch conference, `torch.compile()` support with TensorRT was introduced. See the following:
* Video: https://youtu.be/eGDMJ3MY4zk?si=MhkbgwAPVQSFZEha
* Slides: https://static.sched.com/hosted_files/pytorch2023/9e/Torch_TRT_PyTorch_Conference.pdf
In Slide 7, it's mentioned that:
> Stable Diffusio... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-10-28T06:17:42Z | 2024-01-10T15:07:24Z | 2024-01-10T15:07:24Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,966,499,490 | I_kwDOHa8MBc51Nmai | 5,565 | https://github.com/huggingface/diffusers/issues/5565 | https://api.github.com/repos/huggingface/diffusers/issues/5565 | Not able to run the lora sdxl dreambooth script in the examples directory | ### Describe the bug
When I run the dreambooth script for lora sdxl, I get the following error,
"stabilityai/stable-diffusion-xl-base-1.0 does not appear to have a file named diffusion_pytorch_model.bin"
The full error being, Entry Not Found for url: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-10-28T08:21:31Z | 2023-10-28T13:08:02Z | 2023-10-28T13:08:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mahendra0203 | 1,389,460 | MDQ6VXNlcjEzODk0NjA= | User | false |
huggingface/diffusers | 1,966,520,423 | I_kwDOHa8MBc51Nrhn | 5,566 | https://github.com/huggingface/diffusers/issues/5566 | https://api.github.com/repos/huggingface/diffusers/issues/5566 | Saving multiple models into one file | For [these two lines](https://github.com/huggingface/diffusers/blob/main/examples/instruct_pix2pix/train_instruct_pix2pix.py#L492-L493), why the model is saved multiple times and all to the same file, without filename including the information of the variable `i`? It seems that accelerator can report an error (somethin... | closed | completed | false | 1 | [
"stale"
] | [
"DN6"
] | 2023-10-28T09:34:41Z | 2023-12-08T15:05:23Z | 2023-12-08T15:05:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | deJQK | 25,779,973 | MDQ6VXNlcjI1Nzc5OTcz | User | false |
huggingface/diffusers | 1,966,615,047 | I_kwDOHa8MBc51OCoH | 5,567 | https://github.com/huggingface/diffusers/issues/5567 | https://api.github.com/repos/huggingface/diffusers/issues/5567 | accelerator.prepare overwrites .config method in unet (unet.config) | ### Describe the bug
To preface, this is possibly something only encountered through a very fringe use case.
Accelerate uses the .config method to store it's own internal logic dict (batch size, optimizer params, etc), overriding the one created by diffusers. This causes running the pipeline to fail after putting... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-10-28T14:25:13Z | 2023-12-08T15:05:21Z | 2023-12-08T15:05:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Xynonners | 63,993,670 | MDQ6VXNlcjYzOTkzNjcw | User | false |
huggingface/diffusers | 1,377,350,187 | I_kwDOHa8MBc5SGLIr | 557 | https://github.com/huggingface/diffusers/issues/557 | https://api.github.com/repos/huggingface/diffusers/issues/557 | diffusers DDIM gives worst results than Deforum DDIM | ### Describe the bug
diffusers DDIM gives worst results than Deforum DDIM when using small inference steps of size 8.
My understanding is that ddim is a very fast sampler, and
should yield a decent result even at a low step size like 8 or 16.
it's possible that I didn't set up the scheduler correctly. So, I've inc... | closed | completed | false | 2 | [
"bug"
] | [] | 2022-09-19T03:38:47Z | 2022-09-21T20:22:18Z | 2022-09-21T20:22:18Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AbdullahAlfaraj | 7,842,232 | MDQ6VXNlcjc4NDIyMzI= | User | false |
huggingface/diffusers | 1,966,894,957 | I_kwDOHa8MBc51PG9t | 5,570 | https://github.com/huggingface/diffusers/issues/5570 | https://api.github.com/repos/huggingface/diffusers/issues/5570 | 'compute_snr' function is missing in 'diffusers/training_utils.py' in diffusers v0.21.4 package | ### Describe the bug
When I run [the code for fine-tuning SDXL](https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image_lora_sdxl.py) in diffusers v0.21.4 installed from pip using `pip install diffusers`, I am encountered with the error `ImportError: cannot import name 'compute_... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-10-29T08:29:03Z | 2023-12-08T15:05:19Z | 2023-12-08T15:05:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | leyangjin | 82,919,244 | MDQ6VXNlcjgyOTE5MjQ0 | User | false |
huggingface/diffusers | 1,967,008,720 | I_kwDOHa8MBc51PivQ | 5,572 | https://github.com/huggingface/diffusers/issues/5572 | https://api.github.com/repos/huggingface/diffusers/issues/5572 | AttributeError: 'FrozenDict' object has no attribute 'scaling_factor' | ### Describe the bug
@thedarkzeno
in line 715 of the file i get this error:
vae.config has no attribute scaling-factor
what should i do?
### Reproduction
latents = vae.encode(batch["pixel_values"].to(dtype=weight_dtype)).latent_dist.sample()
latents = latents * vae.config.scaling_factor
### Logs... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-10-29T14:15:34Z | 2024-01-10T15:07:22Z | 2024-01-10T15:07:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hosnasn1987 | 145,333,136 | U_kgDOCKmbkA | User | false |
huggingface/diffusers | 1,967,054,549 | I_kwDOHa8MBc51Pt7V | 5,573 | https://github.com/huggingface/diffusers/issues/5573 | https://api.github.com/repos/huggingface/diffusers/issues/5573 | Keep initializing controlnet weights from unet for a long time | ### Describe the bug
I use the given script to train controlnet however it stuck in initializing weights from unet for a long long time
And I watch the vram and notice that it load the weights again and again, e.g. memory used from 6.5GB to 7.6GB, and then fell down to 6.5GB, and rised to 7.6GB again, and keep re... | closed | completed | false | 20 | [
"bug"
] | [] | 2023-10-29T16:18:35Z | 2024-01-20T21:20:56Z | 2023-11-06T13:05:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | OrangeSodahub | 54,439,582 | MDQ6VXNlcjU0NDM5NTgy | User | false |
huggingface/diffusers | 1,967,371,589 | I_kwDOHa8MBc51Q7VF | 5,574 | https://github.com/huggingface/diffusers/issues/5574 | https://api.github.com/repos/huggingface/diffusers/issues/5574 | How to train a part of UNet attention parameters with LoRA | ### Describe the bug
I adapt the LoRA training code in # to train my model.
And I only want to update the parameters in "down block", so I comment out the code for other attention blocks:
<img width="909" alt="image" src="https://github.com/huggingface/diffusers/assets/52530394/6b204ad8-e201-43b0-ab97-5d29a936e3c... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-10-30T02:58:07Z | 2023-12-08T15:05:16Z | 2023-12-08T15:05:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lxycopper | 52,530,394 | MDQ6VXNlcjUyNTMwMzk0 | User | false |
huggingface/diffusers | 1,967,422,955 | I_kwDOHa8MBc51RH3r | 5,575 | https://github.com/huggingface/diffusers/issues/5575 | https://api.github.com/repos/huggingface/diffusers/issues/5575 | How to set the "transformer_in" layer's hidden size in LoRA training? | ### Describe the bug
I modify the code for text-to-image [lora](https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image_lora.py) as Figure 1,
<img width="908" alt="image" src="https://github.com/huggingface/diffusers/assets/52530394/0639998b-8106-49d9-8761-c58014095e7e">
Howev... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-10-30T03:44:32Z | 2024-01-10T15:07:20Z | 2024-01-10T15:07:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lxycopper | 52,530,394 | MDQ6VXNlcjUyNTMwMzk0 | User | false |
huggingface/diffusers | 1,967,596,192 | I_kwDOHa8MBc51RyKg | 5,576 | https://github.com/huggingface/diffusers/issues/5576 | https://api.github.com/repos/huggingface/diffusers/issues/5576 | Request to implement FreeNoise, a new diffusion scheduler | ### Model/Pipeline/Scheduler description
FreeNoise is a tuning-free and time-efficient paradigm for longer video generation based on pretrained video diffusion models. In other words, it is another *free lunch* that can generate longer videos given multiple prompts, with negligible time cost overhead.
![hd01](https:... | closed | completed | false | 15 | [
"community-examples"
] | [] | 2023-10-30T06:22:16Z | 2024-08-07T05:05:19Z | 2024-08-07T05:05:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | arthur-qiu | 28,619,658 | MDQ6VXNlcjI4NjE5NjU4 | User | false |
huggingface/diffusers | 1,967,674,017 | I_kwDOHa8MBc51SFKh | 5,577 | https://github.com/huggingface/diffusers/issues/5577 | https://api.github.com/repos/huggingface/diffusers/issues/5577 | KeyError: 'time_embed.0.weight' when using `ControlNetModel.from_single_file` | ### Describe the bug
Hi, I had one problem on loading SXDL ControlNet from local files, using the `from_single_file` method.
The model in safetensors format and the config file are downloaded from https://huggingface.co/diffusers/controlnet-canny-sdxl-1.0-small/tree/main .
When calling `ControlNetModel.from_sing... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-10-30T07:23:05Z | 2023-10-30T07:29:35Z | 2023-10-30T07:29:35Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xuyxu | 22,359,569 | MDQ6VXNlcjIyMzU5NTY5 | User | false |
huggingface/diffusers | 1,967,716,288 | I_kwDOHa8MBc51SPfA | 5,578 | https://github.com/huggingface/diffusers/issues/5578 | https://api.github.com/repos/huggingface/diffusers/issues/5578 | OSError: (random module) does not appear to have a file named model_index.json. | ### Describe the bug
Using this example:
```
from diffusers import StableDiffusionControlNetImg2ImgPipeline, ControlNetModel, UniPCMultistepScheduler
import torch
controlnet = ControlNetModel.from_pretrained("lllyasviel/control_v11f1p_sd15_depth", torch_dtype=torch.float16, use_safetensors=True)
pipe = Stable... | closed | completed | false | 9 | [
"bug"
] | [] | 2023-10-30T07:49:14Z | 2023-11-09T13:11:09Z | 2023-11-09T13:11:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,968,091,610 | I_kwDOHa8MBc51TrHa | 5,580 | https://github.com/huggingface/diffusers/issues/5580 | https://api.github.com/repos/huggingface/diffusers/issues/5580 | ControlNets not working correctly when FreeU is enabled | ControlNets not working correctly when FreeU is enabled. Does anybody have the same issue? | closed | completed | false | 2 | [] | [] | 2023-10-30T11:11:53Z | 2023-10-31T02:45:42Z | 2023-10-31T02:45:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | simonnxren | 20,294,218 | MDQ6VXNlcjIwMjk0MjE4 | User | false |
huggingface/diffusers | 1,969,592,517 | I_kwDOHa8MBc51ZZjF | 5,592 | https://github.com/huggingface/diffusers/issues/5592 | https://api.github.com/repos/huggingface/diffusers/issues/5592 | UniDiffuser fail to load from checkpoints | ### Describe the bug
The current version of diffusers (I've tested on both 0.21.4 and 0.21.0) seems to falsely name the `image_processor` as `clip_image_processor`, where the latter is not expected by `UniDiffuserPipeline`, resulting in the `ValueError` indicating missing `image_processor`.
### Reproduction
im... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-10-31T03:57:40Z | 2023-10-31T06:49:34Z | 2023-10-31T06:49:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jinhong-ni | 87,235,406 | MDQ6VXNlcjg3MjM1NDA2 | User | false |
huggingface/diffusers | 1,969,606,714 | I_kwDOHa8MBc51ZdA6 | 5,593 | https://github.com/huggingface/diffusers/issues/5593 | https://api.github.com/repos/huggingface/diffusers/issues/5593 | `AutoPipeline.from_pipe` fails when passed a ControlNet pipeline | ### Describe the bug
The implementation of `AutoPipeline` does a `replace` of `"Pipeline"` with `"ControlNetPipeline"` when the `controlnet` argument is passed.
When `from_pipe` is called with an existing `ControlNet` pipeline, the following transformation occurs.
```python
StableDiffusionControlNetPipeline -->... | closed | completed | false | 3 | [
"bug"
] | [
"yiyixuxu"
] | 2023-10-31T04:16:24Z | 2023-12-23T17:43:48Z | 2023-11-03T18:15:14Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | carson-katri | 13,581,484 | MDQ6VXNlcjEzNTgxNDg0 | User | false |
huggingface/diffusers | 1,969,729,486 | I_kwDOHa8MBc51Z6_O | 5,594 | https://github.com/huggingface/diffusers/issues/5594 | https://api.github.com/repos/huggingface/diffusers/issues/5594 | Saving the memory usage for SDXL fine-tuning using LoRA | **Is your feature request related to a problem? Please describe.**
Fine-tuning SDXL using LoRA consumes too much memory. The script I use is: https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image_lora_sdxl.py.
**Describe the solution you'd like**
Supporting methods like putt... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-10-31T06:22:40Z | 2024-01-03T06:06:43Z | 2023-12-26T15:09:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | leyangjin | 82,919,244 | MDQ6VXNlcjgyOTE5MjQ0 | User | false |
huggingface/diffusers | 1,969,797,845 | I_kwDOHa8MBc51aLrV | 5,595 | https://github.com/huggingface/diffusers/issues/5595 | https://api.github.com/repos/huggingface/diffusers/issues/5595 | raise ValueError(f"`height` and `width` have to be divisible by 8 but are 427 and 640 | ### Describe the bug
Hi, I am using diffuser for inpainting, and the size of input image is (427, 640), to keep the input and output the same resolution, i set the `height=427` and `width=640` in the pipeline(e.g., `img = pipeline(prompt=prompt, image=init_image, mask_image=mask_image, height=427, width=640).images... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-10-31T07:15:50Z | 2023-12-26T15:09:25Z | 2023-12-26T15:09:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | BeijingBlueSky | 46,390,876 | MDQ6VXNlcjQ2MzkwODc2 | User | false |
huggingface/diffusers | 1,969,809,654 | I_kwDOHa8MBc51aOj2 | 5,596 | https://github.com/huggingface/diffusers/issues/5596 | https://api.github.com/repos/huggingface/diffusers/issues/5596 | AttributeError: 'dict' object has no attribute 'convert' When are building pokemon-blip-captions | ### Describe the bug
@sayakpaul @patrickvonplaten
I follow this tutorial(https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/README_sdxl.md) to build SDXL LoRA based on Pokemon, but failed.
If anyone met smiliar issue when building.
Thanks,
Jiancong
### Reproduction
!cd diffusers... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-10-31T07:24:12Z | 2024-01-10T15:07:17Z | 2024-01-10T15:07:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jiancong | 701,159 | MDQ6VXNlcjcwMTE1OQ== | User | false |
huggingface/diffusers | 1,969,869,585 | I_kwDOHa8MBc51adMR | 5,598 | https://github.com/huggingface/diffusers/issues/5598 | https://api.github.com/repos/huggingface/diffusers/issues/5598 | Bug in the `_modify_text_encoder` method | ### Describe the bug
The bottom code in diffusers/loaders.py would cause AttributeError: 'int' object has no attribute 'pop'， when patch_mlp is True and the rank is an int variable instead of a dict
```Python
current_rank_fc1 = rank.pop(f"{name}.fc1.lora_linear_layer.up.weight")
current_rank_fc2 = rank.pop(f"{n... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-10-31T08:05:19Z | 2023-12-26T15:09:19Z | 2023-12-26T15:09:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eezywu | 39,943,728 | MDQ6VXNlcjM5OTQzNzI4 | User | false |
huggingface/diffusers | 1,970,930,623 | I_kwDOHa8MBc51egO_ | 5,599 | https://github.com/huggingface/diffusers/issues/5599 | https://api.github.com/repos/huggingface/diffusers/issues/5599 | How I can train instrct-pix2pix + lora? | How I can train instrct-pix2pix + lora? | closed | completed | false | 3 | [] | [] | 2023-10-31T17:17:35Z | 2023-11-02T08:02:58Z | 2023-11-02T08:02:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mzeynali | 95,717,811 | U_kgDOBbSJsw | User | false |
huggingface/diffusers | 1,377,796,412 | I_kwDOHa8MBc5SH4E8 | 560 | https://github.com/huggingface/diffusers/issues/560 | https://api.github.com/repos/huggingface/diffusers/issues/560 | `UNet2DOutput` becomes nested when wrapped with `accelerate` | ### Describe the bug
When either `UNet2DModel` or `UNet2DConditionModel` are prepared with `accelerate`, their outputs become nested, i.e. to get the `sample` you have to do `outputs.sample['sample']` instead of just `outputs.sample`.
However, it works as expected with `return_dict=False`.
Linked to #499 #434
... | closed | completed | false | 1 | [
"bug"
] | [] | 2022-09-19T11:19:12Z | 2022-09-20T16:32:18Z | 2022-09-20T16:32:18Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | anton-l | 26,864,830 | MDQ6VXNlcjI2ODY0ODMw | User | false |
huggingface/diffusers | 1,971,028,751 | I_kwDOHa8MBc51e4MP | 5,600 | https://github.com/huggingface/diffusers/issues/5600 | https://api.github.com/repos/huggingface/diffusers/issues/5600 | Over Grow issue for SDXL inpainting model for outpainting task. | I Have been exploring SDXL from sometime now, However I'm not able to understand in inpainting model I'm facing overgrow issue in some cases, I would love to now how I can fix this issue and why does it happen? | closed | completed | false | 12 | [
"stale",
"inpainting"
] | [
"DN6"
] | 2023-10-31T18:03:13Z | 2024-01-19T15:06:15Z | 2024-01-19T15:06:15Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rajdeep2804 | 53,480,855 | MDQ6VXNlcjUzNDgwODU1 | User | false |
huggingface/diffusers | 1,971,036,340 | I_kwDOHa8MBc51e6C0 | 5,601 | https://github.com/huggingface/diffusers/issues/5601 | https://api.github.com/repos/huggingface/diffusers/issues/5601 | Fine Tuning SDXL inpainting weights. | I was wondering how can train SDXL inpainting model also how to build custom dataset for such task and feed it to model. It'll be really helpful if someone could guide with the process. | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-31T18:06:58Z | 2023-12-26T15:09:17Z | 2023-12-26T15:09:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rajdeep2804 | 53,480,855 | MDQ6VXNlcjUzNDgwODU1 | User | false |
huggingface/diffusers | 1,971,610,270 | I_kwDOHa8MBc51hGKe | 5,603 | https://github.com/huggingface/diffusers/issues/5603 | https://api.github.com/repos/huggingface/diffusers/issues/5603 | ImportError: cannot import name 'is_safetensors_available' from 'diffusers.utils' | ### Describe the bug
Hello, I have the following error: ImportError: cannot import name 'is_safetensors_available' from 'diffusers.utils'
### Reproduction
Traceback (most recent call last):
File "C:\Users\USUARIO\Desktop\wonder3d\test_mvdiffusion_seq.py", line 40, in <module>
from mvdiffusion.models.unet_mv2... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-11-01T03:04:54Z | 2024-03-09T04:50:08Z | 2023-11-08T00:43:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | peepeepeepoopoopoo | 137,663,161 | U_kgDOCDSSuQ | User | false |
huggingface/diffusers | 1,971,826,927 | I_kwDOHa8MBc51h7Dv | 5,604 | https://github.com/huggingface/diffusers/issues/5604 | https://api.github.com/repos/huggingface/diffusers/issues/5604 | Unfusing multiple LoRA is not possible. | ### Describe the bug
I need to run following repro code snippet as the user request is comming. That means each time, I need to unfuse_lora for pipe first, then to load two lora weights checkpoints.
After several repeats, I've found the LoRA seems not unfuse as even I set the lora_scale to 0.0, the effect is still ... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-11-01T07:07:02Z | 2023-11-01T08:34:21Z | 2023-11-01T08:34:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jiancong | 701,159 | MDQ6VXNlcjcwMTE1OQ== | User | false |
huggingface/diffusers | 1,971,857,894 | I_kwDOHa8MBc51iCnm | 5,605 | https://github.com/huggingface/diffusers/issues/5605 | https://api.github.com/repos/huggingface/diffusers/issues/5605 | Add per image caption support for SDXL dreambooth. | **Is your feature request related to a problem? Please describe.**
No.
**Describe the solution you'd like**
Implement per image captions for SDXL dreambooth similar to Everydream2 (https://github.com/victorchall/EveryDream-trainer) which only works for SD-1.5.
I have tried to make a preliminary implementation h... | closed | completed | false | 6 | [
"stale"
] | [
"DN6"
] | 2023-11-01T07:33:35Z | 2023-12-05T06:57:55Z | 2023-12-05T06:57:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aravindhv10 | 10,844,204 | MDQ6VXNlcjEwODQ0MjA0 | User | false |
huggingface/diffusers | 1,971,935,592 | I_kwDOHa8MBc51iVlo | 5,606 | https://github.com/huggingface/diffusers/issues/5606 | https://api.github.com/repos/huggingface/diffusers/issues/5606 | Bug when loading LoRA with no alpha value | ### Describe the bug
I got an error message when loading LoRA in civitai using `pipeline.load_lora_weights()`
```
File ".../python3.10/site-packages/diffusers/utils/peft_utils.py", line 147, in get_peft_kwargs
lora_alpha = set(network_alpha_dict.values()).pop()
KeyError: 'pop from an empty set
```
whe... | closed | completed | false | 7 | [
"bug"
] | [
"younesbelkada"
] | 2023-11-01T08:34:23Z | 2023-11-02T05:15:38Z | 2023-11-01T21:08:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Eun0 | 33,515,697 | MDQ6VXNlcjMzNTE1Njk3 | User | false |
huggingface/diffusers | 1,972,037,675 | I_kwDOHa8MBc51iugr | 5,607 | https://github.com/huggingface/diffusers/issues/5607 | https://api.github.com/repos/huggingface/diffusers/issues/5607 | RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 64 but got size 512 for tensor number 2 in the list. | ### Describe the bug
Hi,
I'm encountering a problem when I call the __call__ function of StableDiffusionInpaintingPipeline.
It's throwing a RuntimeError. This issue only occurs with the latest versions of the library (I tested with 0.21.1, 0.21.4, 0.22.0dev); it works fine with version 0.20.0.
What am I missing... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-11-01T09:45:50Z | 2023-11-10T10:08:36Z | 2023-11-01T20:02:20Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fabiorigano | 57,982,783 | MDQ6VXNlcjU3OTgyNzgz | User | false |
huggingface/diffusers | 1,972,235,491 | I_kwDOHa8MBc51jezj | 5,609 | https://github.com/huggingface/diffusers/issues/5609 | https://api.github.com/repos/huggingface/diffusers/issues/5609 | is it reasonable to add controlent to instruct pix2pix? | I want to use instructpix2pix for arranging items on store shelves, I gather 200 pair before and after images, the before images are empty items (shelves **without** items) and the after images are full items (shelves **with** items), The train was I train 5000 steps, the train was successful, but in the inference time... | closed | completed | false | 1 | [] | [] | 2023-11-01T11:59:16Z | 2023-11-02T08:03:10Z | 2023-11-02T08:03:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mzeynali | 95,717,811 | U_kgDOBbSJsw | User | false |
huggingface/diffusers | 1,972,839,843 | I_kwDOHa8MBc51lyWj | 5,613 | https://github.com/huggingface/diffusers/issues/5613 | https://api.github.com/repos/huggingface/diffusers/issues/5613 | Need a way to switch pipeline type when using custom pipelines | With standard SD pipelines, you can switch from txt2img to img2img easily:
```py
if new_pipe_type == DiffusersTaskType.TEXT_2_IMAGE:
new_pipe = diffusers.AutoPipelineForText2Image.from_pipe(pipe)
elif new_pipe_type == DiffusersTaskType.IMAGE_2_IMAGE:
new_pipe = diffusers.A... | closed | completed | false | 3 | [] | [] | 2023-11-01T17:58:23Z | 2023-11-02T18:51:59Z | 2023-11-02T18:51:59Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.