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 | 1,816,657,535 | I_kwDOHa8MBc5sR_5_ | 4,206 | https://github.com/huggingface/diffusers/issues/4206 | https://api.github.com/repos/huggingface/diffusers/issues/4206 | the main branch is broken for controlnet training with sdxl | ### Describe the bug
the main branch (all PRs up to #4205) failed the controlnet training with sdxl. Some kind of disalignment of arguments.
### Reproduction
export MODEL_DIR="stabilityai/stable-diffusion-xl-base-0.9"
export VAE_DIR="madebyollin/sdxl-vae-fp16-fix"
export OUTPUT_DIR="product_train_output_ext... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-07-22T07:10:03Z | 2023-07-25T00:02:24Z | 2023-07-25T00:01:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yutongli | 12,973,127 | MDQ6VXNlcjEyOTczMTI3 | User | false |
huggingface/diffusers | 1,816,745,631 | I_kwDOHa8MBc5sSVaf | 4,208 | https://github.com/huggingface/diffusers/issues/4208 | https://api.github.com/repos/huggingface/diffusers/issues/4208 | Added the ability to set SDXL `Micro-Conditioning` embeddings as 0 | **Is your feature request related to a problem? Please describe.**
During the SDXL training process, it may be necessary to pass in a zero embedding as `Micro-Conditioning` embeddings:
https://github.com/Stability-AI/generative-models/blob/e25e4c0df1d01fb9720f62c73b4feab2e4003e3f/sgm/modules/encoders/modules.py#L... | closed | completed | false | 6 | [] | [] | 2023-07-22T12:19:52Z | 2023-08-26T03:43:45Z | 2023-08-26T03:43:45Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | budui | 16,448,529 | MDQ6VXNlcjE2NDQ4NTI5 | User | false |
huggingface/diffusers | 1,816,778,946 | I_kwDOHa8MBc5sSdjC | 4,209 | https://github.com/huggingface/diffusers/issues/4209 | https://api.github.com/repos/huggingface/diffusers/issues/4209 | Implement using RetNet instead of transformers | New RetNet is introduced i think it should be implemented in diffusers . I think you already knew about it but it may be helpful for faster inference.
| closed | completed | false | 3 | [
"stale"
] | [] | 2023-07-22T14:10:44Z | 2023-09-02T15:03:23Z | 2023-09-02T15:03:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tronizx | 87,223,249 | MDQ6VXNlcjg3MjIzMjQ5 | User | false |
huggingface/diffusers | 1,816,839,580 | I_kwDOHa8MBc5sSsWc | 4,210 | https://github.com/huggingface/diffusers/issues/4210 | https://api.github.com/repos/huggingface/diffusers/issues/4210 | How to use "attention_mask" in "forward" function of "UNet2DConditionModel" defined in "diffusers/src/diffusers/models /unet_2d_condition.py"? | ### Describe the bug
How to use the "attention_mask" in UNet2DConditionModel? What should the size of "attention_mask" look like?
And "attention_mask" can not be used when opening "enable_xformers_memory_efficient_attention" in "examples/text_to_image/train_text_to_image.py"?
` File "/usr/local/lib/python3.9/... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-07-22T17:28:56Z | 2024-10-18T16:34:37Z | 2023-10-30T15:09:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ZihaoW123 | 47,708,655 | MDQ6VXNlcjQ3NzA4NjU1 | User | false |
huggingface/diffusers | 1,816,947,711 | I_kwDOHa8MBc5sTGv_ | 4,213 | https://github.com/huggingface/diffusers/issues/4213 | https://api.github.com/repos/huggingface/diffusers/issues/4213 | Running SD-XL with from_single_file() does not work without accelerate | ### Describe the bug
I don't actually know if this is specific to `from_single_file()` or not. This error occurred with a fresh install of `python -m pip install git+https://github.com/huggingface/diffusers`.
```
(sd) root@5fdf95ca5d58:/workspace/sd# python run.py
Traceback (most recent call last):
File "/wo... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-07-23T01:15:40Z | 2023-07-25T16:20:19Z | 2023-07-25T16:20:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | float-trip | 102,226,344 | U_kgDOBhfZqA | User | false |
huggingface/diffusers | 1,817,157,535 | I_kwDOHa8MBc5sT5-f | 4,217 | https://github.com/huggingface/diffusers/issues/4217 | https://api.github.com/repos/huggingface/diffusers/issues/4217 | Slow loading time with safetensors | ### Describe the bug
When use_safetensors is True while loading a pipeline with from_pretrained, the first call is slow, the next one is the same as without safetensors.
### Reproduction
https://huggingface.co/stabilityai/stable-diffusion-2-inpainting
```
from diffusers import StableDiffusionInpaintPipeline
impor... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-07-23T14:48:11Z | 2023-08-30T15:03:50Z | 2023-08-30T15:03:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aviadmx | 95,345,571 | U_kgDOBa7bow | User | false |
huggingface/diffusers | 1,817,165,559 | I_kwDOHa8MBc5sT773 | 4,218 | https://github.com/huggingface/diffusers/issues/4218 | https://api.github.com/repos/huggingface/diffusers/issues/4218 | diffusers.FlaxStableDiffusionPipeline doesnt have negative prompts | as title 🤷♀️🤷♀️
diffusers.FlaxStableDiffusionPipeline doesnt have negative prompts so any model checkpoint cant generate any decent faces | closed | completed | false | 5 | [
"stale"
] | [] | 2023-07-23T15:11:31Z | 2023-08-23T16:47:26Z | 2023-08-23T16:47:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | peternasser99 | 94,254,554 | U_kgDOBZ412g | User | false |
huggingface/diffusers | 1,817,184,721 | I_kwDOHa8MBc5sUAnR | 4,219 | https://github.com/huggingface/diffusers/issues/4219 | https://api.github.com/repos/huggingface/diffusers/issues/4219 | Support to convert between ControlNet 2.1 checkpoints | ### Describe the bug
Hello everyone,
I was trying to download: https://civitai.com/models/38784?modelVersionId=44716
https://github.com/Mikubill/sd-webui-controlnet
```
I did :
!python ./diffusers/scripts/convert_original_controlnet_to_diffusers.py --checkpoint_path ./test_model.safetensors --original_config_... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-07-23T16:03:50Z | 2023-07-24T03:21:31Z | 2023-07-24T03:00:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,817,197,778 | I_kwDOHa8MBc5sUDzS | 4,220 | https://github.com/huggingface/diffusers/issues/4220 | https://api.github.com/repos/huggingface/diffusers/issues/4220 | RAM cache is not cleared when using cpu model offloading | I use stable diffusion inference using code on my local Ubuntu machine to analyze dozens of images. However, when the CPU model offloading option is enabled, the system's RAM gradually increases with each inference run. It appears that the cache of the stable diffusion (SD) modules is not being cleared and is stored in... | closed | completed | false | 3 | [] | [] | 2023-07-23T16:44:12Z | 2023-07-24T13:47:22Z | 2023-07-24T13:47:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mahdip72 | 42,680,708 | MDQ6VXNlcjQyNjgwNzA4 | User | false |
huggingface/diffusers | 1,817,525,138 | I_kwDOHa8MBc5sVTuS | 4,222 | https://github.com/huggingface/diffusers/issues/4222 | https://api.github.com/repos/huggingface/diffusers/issues/4222 | How to train ldm on a low-resolution image dataset (128*128) | **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"
] | [] | 2023-07-24T03:14:20Z | 2023-08-31T15:04:25Z | 2023-08-31T15:04:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | crowningwang | 114,406,126 | U_kgDOBtGy7g | User | false |
huggingface/diffusers | 1,817,591,545 | I_kwDOHa8MBc5sVj75 | 4,224 | https://github.com/huggingface/diffusers/issues/4224 | https://api.github.com/repos/huggingface/diffusers/issues/4224 | 'UNet2DConditionModel' object has no attribute 'control_model' | ### Describe the bug
```
!python ./diffusers/scripts/convert_lora_safetensor_to_diffusers.py --base_model_path runwayml/stable-diffusion-v1-5 --checkpoint_path ./test_model.safetensors --dump_path ./test_model
/opt/conda/lib/python3.10/site-packages/scipy/__init__.py:146: UserWarning: A NumPy version >=1.16.5 and ... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-07-24T04:26:11Z | 2023-09-05T15:03:25Z | 2023-09-05T15:03:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,817,645,610 | I_kwDOHa8MBc5sVxIq | 4,225 | https://github.com/huggingface/diffusers/issues/4225 | https://api.github.com/repos/huggingface/diffusers/issues/4225 | Error(s) in loading state_dict for ControlNetModel | ### Describe the bug
i am trying to edit the image and to it, for example for victorian image i would like to add flower with it: using https://github.com/sail-sg/EditAnything
```
!python ./diffusers/scripts/convert_original_stable_diffusion_to_diffusers.py \
--checkpoint_path ./test_model.safetensors \
--... | closed | completed | false | 11 | [
"bug",
"stale"
] | [] | 2023-07-24T05:22:19Z | 2023-08-31T15:04:22Z | 2023-08-31T15:04:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,817,768,580 | I_kwDOHa8MBc5sWPKE | 4,226 | https://github.com/huggingface/diffusers/issues/4226 | https://api.github.com/repos/huggingface/diffusers/issues/4226 | Stable Diffusion XL, unet convert to onnx failed | ### Describe the bug
I met the following issue when converting unet in Stable Diffusion XL to onnx format.
The code I used in converting is as follows:
def convert_unet_onnx(unet:torch.nn.Module, onnx_path:... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-07-24T06:56:10Z | 2023-08-24T13:33:51Z | 2023-08-24T13:33:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhuo-yoyowz | 35,246,475 | MDQ6VXNlcjM1MjQ2NDc1 | User | false |
huggingface/diffusers | 1,818,131,086 | I_kwDOHa8MBc5sXnqO | 4,229 | https://github.com/huggingface/diffusers/issues/4229 | https://api.github.com/repos/huggingface/diffusers/issues/4229 | Using train_ dreamboth.py encountered an error during accelerated training in a multi GPU environment. | ### Describe the bug
Unknown error in configuring multiple GPU environment acceleration。
### Reproduction
- `Accelerate` version: 0.20.3
- Platform: Linux-5.4.0-153-generic-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.13
- Numpy version: 1.21.6
- PyTorch version (GPU?): 1.13.1+cu117 (True)
- PyTorch XP... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-07-24T10:31:27Z | 2023-08-31T15:04:19Z | 2023-08-31T15:04:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | renyi123456789 | 130,567,668 | U_kgDOB8hN9A | User | false |
huggingface/diffusers | 1,818,440,169 | I_kwDOHa8MBc5sYzHp | 4,232 | https://github.com/huggingface/diffusers/issues/4232 | https://api.github.com/repos/huggingface/diffusers/issues/4232 | Train and Inference code for instruct-pix2pix using controlnet condition | ### Model/Pipeline/Scheduler description
I used [instruct-pix2pix](https://huggingface.co/docs/diffusers/training/instructpix2pix) for training pipeline, but I want to add control-net into instruct-pix2pix for both train code and inference.
### Open source status
- [ ] The model implementation is available
- [ ] Th... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-07-24T13:32:45Z | 2024-09-05T21:26:25Z | 2023-08-31T15:04:15Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mzeynali | 95,717,811 | U_kgDOBbSJsw | User | false |
huggingface/diffusers | 1,818,459,908 | I_kwDOHa8MBc5sY38E | 4,233 | https://github.com/huggingface/diffusers/issues/4233 | https://api.github.com/repos/huggingface/diffusers/issues/4233 | Add Taesd light weight encoder/decoder | https://github.com/madebyollin has build a super cool "preview" VAE (Conv+Relu resblocks and 2x upsample layers) for almost instant encode/decode operations.
This can be used very nicely to preview latents inputs / outputs when denoising
Repo: https://github.com/madebyollin/taesd
Weights for SD: https://huggin... | closed | completed | false | 2 | [] | [
"sayakpaul"
] | 2023-07-24T13:43:25Z | 2023-08-02T18:28:07Z | 2023-08-02T18:28:07Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,818,466,619 | I_kwDOHa8MBc5sY5k7 | 4,234 | https://github.com/huggingface/diffusers/issues/4234 | https://api.github.com/repos/huggingface/diffusers/issues/4234 | How to train instruct-pix2pix with controlnet and inference | Hi guys,
I want to train instruct-pix2pix using controlnet condition. As you know, currently available for [instruct-pix2pix](https://huggingface.co/docs/diffusers/training/instructpix2pix) and [control net](https://huggingface.co/docs/diffusers/training/controlnet) separately.
**Q1)** Have you plan about this probl... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-07-24T13:47:02Z | 2023-08-31T15:04:14Z | 2023-08-31T15:04:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mzeynali | 95,717,811 | U_kgDOBbSJsw | User | false |
huggingface/diffusers | 1,818,620,102 | I_kwDOHa8MBc5sZfDG | 4,236 | https://github.com/huggingface/diffusers/issues/4236 | https://api.github.com/repos/huggingface/diffusers/issues/4236 | [SDXL Refiner] Refiner doesn't register text encoder? | ## Code:
Taken from https://huggingface.co/docs/diffusers/main/en/api/pipelines/stable_diffusion/stable_diffusion_xl
```python
import torch
from diffusers import StableDiffusionXLImg2ImgPipeline
from diffusers.utils import load_image
pipe = StableDiffusionXLImg2ImgPipeline.from_pretrained(
"stabilityai... | closed | completed | false | 2 | [] | [] | 2023-07-24T15:09:06Z | 2023-07-24T16:37:08Z | 2023-07-24T16:37:08Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,818,983,270 | I_kwDOHa8MBc5sa3tm | 4,241 | https://github.com/huggingface/diffusers/issues/4241 | https://api.github.com/repos/huggingface/diffusers/issues/4241 | When can you support fine-tuning unidiffusers using Lora? | **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"
] | [] | 2023-07-24T19:13:18Z | 2023-09-02T15:03:21Z | 2023-09-02T15:03:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zdxff | 47,980,021 | MDQ6VXNlcjQ3OTgwMDIx | User | false |
huggingface/diffusers | 1,819,107,078 | I_kwDOHa8MBc5sbV8G | 4,243 | https://github.com/huggingface/diffusers/issues/4243 | https://api.github.com/repos/huggingface/diffusers/issues/4243 | Remove `use_safetensors` from `diffusers` codebase | `safetensors` is now a required dependency as can be seen here: https://github.com/huggingface/diffusers/pull/4177
This means we can remove all of the following `is_safetensors_available` codesnippets:
```
scripts/convert_original_audioldm_to_diffusers.py
41:from diffusers.utils import is_omegaconf_available, i... | closed | completed | false | 8 | [] | [
"chiral-carbon"
] | 2023-07-24T20:36:05Z | 2023-11-01T04:20:31Z | 2023-08-11T05:35:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,819,112,674 | I_kwDOHa8MBc5sbXTi | 4,244 | https://github.com/huggingface/diffusers/issues/4244 | https://api.github.com/repos/huggingface/diffusers/issues/4244 | Remove all safetensors import mocking from tests | We have a `use_safetensors` argument in from_pretrained which can fully control whether we load in safetensors or not. We can clean up this dummy import mess a bit here: https://github.com/huggingface/diffusers/blob/95b7de88fd0dffef2533f1cbaf9ffd9d3c6d04c8/tests/pipelines/test_pipelines.py#L399 and just work with `use_... | closed | completed | false | 2 | [] | [
"DN6"
] | 2023-07-24T20:40:21Z | 2023-08-18T07:57:29Z | 2023-08-18T07:57:29Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,819,245,037 | I_kwDOHa8MBc5sb3nt | 4,246 | https://github.com/huggingface/diffusers/issues/4246 | https://api.github.com/repos/huggingface/diffusers/issues/4246 | [Slow tests] Clean up tests | At the moment, we have a couple of problems with our slow tests:
- Too many tests are failing too much. Just today we had 72 test failures, see [here](https://github.com/huggingface/diffusers/actions/runs/5647471380/job/15297645760) which makes it difficult to spot important from non-important test failures
- The slo... | closed | completed | false | 2 | [
"stale"
] | [
"DN6"
] | 2023-07-24T22:33:43Z | 2023-09-02T15:03:20Z | 2023-09-02T15:03:20Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,819,323,825 | I_kwDOHa8MBc5scK2x | 4,247 | https://github.com/huggingface/diffusers/issues/4247 | https://api.github.com/repos/huggingface/diffusers/issues/4247 | Test for khoya style checkpoint conversions | It would be great to add some tests for converting khoya style checkpoints. These don't have to be super in depth but just sanity checks that we can convert the checkpoints correctly :) | closed | completed | false | 6 | [
"good first issue"
] | [
"DN6"
] | 2023-07-25T00:16:41Z | 2023-07-27T17:08:54Z | 2023-07-27T17:08:53Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,819,378,994 | I_kwDOHa8MBc5scYUy | 4,249 | https://github.com/huggingface/diffusers/issues/4249 | https://api.github.com/repos/huggingface/diffusers/issues/4249 | `check_inputs` `StableDiffusionUpscalePipeline` fails when passing `prompt_embeds` | There is an implicit assumption in the below region that `prompt` is always passed, and not prompt embeds, because it only checks `prompt` for the batch size.
https://github.com/huggingface/diffusers/blob/5652c43f8398473702bfdc30b55700cd378b1e87/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_upsc... | closed | completed | false | 2 | [] | [
"yiyixuxu"
] | 2023-07-25T01:19:17Z | 2023-07-29T03:33:35Z | 2023-07-29T03:33:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JulianKnodt | 7,675,847 | MDQ6VXNlcjc2NzU4NDc= | User | false |
huggingface/diffusers | 1,819,434,114 | I_kwDOHa8MBc5sclyC | 4,250 | https://github.com/huggingface/diffusers/issues/4250 | https://api.github.com/repos/huggingface/diffusers/issues/4250 | from_pretrained load model incorrectly | ### Describe the bug
I use pipeline in my code,I confim my path is correct,but occered error:
huggingface_hub.utils._validators.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/root/workspace/stable-diffusion-webui/models/Stable-diffusion/model.ckpt'. Use `repo_type` argument if ... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-07-25T02:19:52Z | 2023-07-25T05:38:11Z | 2023-07-25T05:38:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | misi0202 | 101,965,629 | U_kgDOBhPfPQ | User | false |
huggingface/diffusers | 1,819,492,958 | I_kwDOHa8MBc5sc0Je | 4,252 | https://github.com/huggingface/diffusers/issues/4252 | https://api.github.com/repos/huggingface/diffusers/issues/4252 | StableDiffusionPipeline __call__ parameters format | I am struggling to implement the negative_prompt_embeds parameter to the pipeline for calling.
I tried to convert the pytorch file(.pt) negative_embedding to the torch.FloatTensor using torch.load.
However, I am running into the error
"RuntimeError: Sizes of tensors must match except in dimension 0. Expected siz... | closed | completed | false | 2 | [] | [] | 2023-07-25T03:36:09Z | 2023-07-25T18:33:31Z | 2023-07-25T18:33:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | junhwanlee2316 | 65,512,890 | MDQ6VXNlcjY1NTEyODkw | User | false |
huggingface/diffusers | 1,819,585,297 | I_kwDOHa8MBc5sdKsR | 4,253 | https://github.com/huggingface/diffusers/issues/4253 | https://api.github.com/repos/huggingface/diffusers/issues/4253 | Can I input an image to guide generation using the stable diffusion already fine-tuned with the dreambooth | Can I input an image to guide the stable diffusion already fine-tuned with the dreambooth? It's a bit like stable diffusion image-to-image generation, but with Dreambooth's fine-tuned model to generate. | closed | completed | false | 2 | [
"stale"
] | [] | 2023-07-25T05:36:14Z | 2023-09-02T15:03:18Z | 2023-09-02T15:03:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | liyiyue2021 | 81,152,150 | MDQ6VXNlcjgxMTUyMTUw | User | false |
huggingface/diffusers | 1,819,910,753 | I_kwDOHa8MBc5seaJh | 4,255 | https://github.com/huggingface/diffusers/issues/4255 | https://api.github.com/repos/huggingface/diffusers/issues/4255 | Diffusers results much worse than sd-webui | ### Describe the bug
Hi, I’m faced with a confusing situation and would appreciate any advice from anyone. When I generate images with diffusers, the result quality is much worse than the images generated from sd-webui.
And I also found that usually images generated by diffusers are not as good as those of sd-web... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-07-25T09:28:01Z | 2023-07-26T03:04:46Z | 2023-07-26T03:02:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | volcverse | 72,564,247 | MDQ6VXNlcjcyNTY0MjQ3 | User | false |
huggingface/diffusers | 1,819,916,142 | I_kwDOHa8MBc5sebdu | 4,256 | https://github.com/huggingface/diffusers/issues/4256 | https://api.github.com/repos/huggingface/diffusers/issues/4256 | Got error when train SDXL-Dreambooth | ### Describe the bug
after text_encoder fine-tune branch is merged, I got error in run train_dreambooth_lora_sdxl.py.
### Reproduction
python = "3.10.12"
wandb = "0.15.5"
xformers = "0.0.20"
transformers = "4.28.1"
diffusers = { git = "https://github.com/huggingface/diffusers.git" }
accelerate = "0.21.0"
strea... | closed | completed | false | 8 | [
"bug"
] | [] | 2023-07-25T09:30:37Z | 2023-07-25T19:10:03Z | 2023-07-25T19:10:03Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hwjeon98 | 102,274,521 | U_kgDOBhiV2Q | User | false |
huggingface/diffusers | 1,820,019,213 | I_kwDOHa8MBc5se0oN | 4,257 | https://github.com/huggingface/diffusers/issues/4257 | https://api.github.com/repos/huggingface/diffusers/issues/4257 | Improve support of Reference-only and outpainting for SDXL | Reference-only has shown be a very powerful mechanism for outpainting as well as image variation. We should investigate a bit how we can best support this in a modularized, library-friendly way in `diffusers`. Also I think we should try this out for SDXL.
Related links:
- https://github.com/Mikubill/sd-webui-contro... | open | reopened | false | 22 | [
"New pipeline/model",
"community-examples",
"wip"
] | [
"DN6"
] | 2023-07-25T10:23:03Z | 2024-11-22T01:32:26Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,820,195,720 | I_kwDOHa8MBc5sffuI | 4,260 | https://github.com/huggingface/diffusers/issues/4260 | https://api.github.com/repos/huggingface/diffusers/issues/4260 | Lazy import | To speed up import and to stop this mess: https://github.com/huggingface/diffusers/pull/4259
we should implement Transfomer's way of doing lazy import: https://github.com/huggingface/transformers/pull/9446. | closed | completed | false | 2 | [
"enhancement"
] | [
"DN6",
"patrickvonplaten"
] | 2023-07-25T12:14:23Z | 2024-11-23T00:08:42Z | 2024-11-23T00:08:42Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,820,460,525 | I_kwDOHa8MBc5sggXt | 4,263 | https://github.com/huggingface/diffusers/issues/4263 | https://api.github.com/repos/huggingface/diffusers/issues/4263 | LoRA unloading is broken | ### Describe the bug
Seems like LoRA unloading stopped working after the commit ebfe343149b3aa501bac50621c33eefee59c15e2.
### Reproduction
Tested this with a Kohya-SS style LoRA
```py
pipe = StableDiffusionPipeline.from_pretrained(
model_name,
torch_dtype=torch.float16,
safety_checker=None,
)
pi... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-07-25T14:21:25Z | 2023-11-22T15:07:23Z | 2023-11-22T15:07:23Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | isidentical | 47,358,913 | MDQ6VXNlcjQ3MzU4OTEz | User | false |
huggingface/diffusers | 1,820,482,923 | I_kwDOHa8MBc5sgl1r | 4,264 | https://github.com/huggingface/diffusers/issues/4264 | https://api.github.com/repos/huggingface/diffusers/issues/4264 | SDXL Ensemble of Experts example images are missing | ### Describe the bug
The two lion example images in the ensemble of experts section are missing.
<img width="760" alt="Screenshot 2023-07-25 at 15 30 54" src="https://github.com/huggingface/diffusers/assets/319055/c16d425e-4080-4d4d-bc1b-57610206f9eb">
https://huggingface.co/docs/diffusers/main/en/api/pipeline... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-07-25T14:32:45Z | 2023-07-25T16:42:13Z | 2023-07-25T16:42:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fofr | 319,055 | MDQ6VXNlcjMxOTA1NQ== | User | false |
huggingface/diffusers | 1,820,522,852 | I_kwDOHa8MBc5sgvlk | 4,265 | https://github.com/huggingface/diffusers/issues/4265 | https://api.github.com/repos/huggingface/diffusers/issues/4265 | ImportError: cannot import name 'StableDiffusionXLControlNetPipeline' from 'diffusers.pipelines.controlnet' | ### Describe the bug
ImportError: cannot import name 'StableDiffusionXLControlNetPipeline' from 'diffusers.pipelines.controlnet'
### Reproduction
import torch
from diffusers import StableDiffusionPipeline
pipeline = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16... | closed | completed | false | 10 | [
"bug"
] | [] | 2023-07-25T14:52:45Z | 2023-07-25T23:46:07Z | 2023-07-25T23:46:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Dai-Wenxun | 58,767,402 | MDQ6VXNlcjU4NzY3NDAy | User | false |
huggingface/diffusers | 1,820,584,849 | I_kwDOHa8MBc5sg-uR | 4,266 | https://github.com/huggingface/diffusers/issues/4266 | https://api.github.com/repos/huggingface/diffusers/issues/4266 | SDXL training on Mac: tensor<1x77x1xf16>' and 'tensor<1xf32>' are not broadcast compatible | ### Describe the bug
Pytorch crashes during dreambooth sdxl training from the examples directory:
loc("varianceEps"("(mpsFileLoc): /AppleInternal/Library/BuildRoots/d9889869-120b-11ee-b796-7a03568b17ac/Library/Caches/com.apple.xbs/Sources/MetalPerformanceShadersGraph/mpsgraph/MetalPerformanceShadersGraph/Core/Fil... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-07-25T15:21:26Z | 2023-09-02T15:03:16Z | 2023-09-02T15:03:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rrva | 887,132 | MDQ6VXNlcjg4NzEzMg== | User | false |
huggingface/diffusers | 1,820,953,590 | I_kwDOHa8MBc5siYv2 | 4,268 | https://github.com/huggingface/diffusers/issues/4268 | https://api.github.com/repos/huggingface/diffusers/issues/4268 | Update _convert_kohya_ss_state_dict to directly output the new naming conventions | `_convert_kohya_ss_state_dict` outputs the legacy lora conventions when the additional layer was stored on the model without a formal separate module. We should update it to directly output the new naming convention with formal lora layers
See previous discussion: https://github.com/huggingface/diffusers/pull/4147#d... | closed | completed | false | 7 | [
"good first issue"
] | [] | 2023-07-25T18:57:25Z | 2024-11-21T15:43:06Z | 2024-11-21T15:43:06Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,820,973,113 | I_kwDOHa8MBc5sidg5 | 4,269 | https://github.com/huggingface/diffusers/issues/4269 | https://api.github.com/repos/huggingface/diffusers/issues/4269 | Default args in `StableDiffusionControlNetPipeline.check_inputs` causes TypeError | ### Describe the bug
Default value of `control_guidance_start` and `control_guidance_end` in `StableDiffusionControlNetPipeline.check_inputs` function causes `TypeError: object of type 'float' has no len()`
### Reproduction
```
from diffusers import (
StableDiffusionControlNetPipeline,
ControlNetModel
)
i... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-07-25T19:11:02Z | 2023-08-03T14:44:50Z | 2023-08-03T14:44:50Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | isNeil | 39,628,660 | MDQ6VXNlcjM5NjI4NjYw | User | false |
huggingface/diffusers | 1,821,416,487 | I_kwDOHa8MBc5skJwn | 4,274 | https://github.com/huggingface/diffusers/issues/4274 | https://api.github.com/repos/huggingface/diffusers/issues/4274 | [Pipelines] Add BLIP diffusion | BLIP diffusion (Salesforce): https://dxli94.github.io/BLIP-Diffusion-website/
BLIP diffusion enables subject-driven zero-shot image generation which is probably its best USP.
* Code with pre-trained weights: https://github.com/salesforce/LAVIS/tree/main/projects/blip-diffusion
* Paper: https://arxiv.org/abs/230... | closed | completed | false | 2 | [
"New pipeline/model",
"contributions-welcome"
] | [] | 2023-07-26T01:24:21Z | 2024-04-30T03:18:58Z | 2024-04-30T03:18:58Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,821,710,488 | I_kwDOHa8MBc5slRiY | 4,279 | https://github.com/huggingface/diffusers/issues/4279 | https://api.github.com/repos/huggingface/diffusers/issues/4279 | SD-XL 0.9 LoRAs from CivitAI cannot be loaded | ### Describe the bug
Many SDXL 0.9 LoRAs from Civit AI don't seem to work. Can we investigate what is going on here? It would be important to support SDXL LoRA out of the box for 1.0.
Try the following links:
https://civitai.com/models/110400
https://civitai.com/models/112904?modelVersionId=121933
https://civi... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-07-26T06:55:01Z | 2023-09-02T15:03:14Z | 2023-09-02T15:03:14Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,821,847,488 | I_kwDOHa8MBc5sly_A | 4,282 | https://github.com/huggingface/diffusers/issues/4282 | https://api.github.com/repos/huggingface/diffusers/issues/4282 | Add Multi-ControlNet pipeline | ### Describe the bug
i was trying to use multiple safetensors as mentioned in the example:
```
from diffusers import StableDiffusionControlNetPipeline, ControlNetModel, UniPCMultistepScheduler
import torch
controlnet = [
ControlNetModel.from_pretrained("lllyasviel/sd-controlnet-openpose", torch_dtype=torch.... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-07-26T08:25:33Z | 2023-09-12T15:04:09Z | 2023-09-12T15:04:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,822,301,405 | I_kwDOHa8MBc5snhzd | 4,286 | https://github.com/huggingface/diffusers/issues/4286 | https://api.github.com/repos/huggingface/diffusers/issues/4286 | load_lora_weights not working in SDXL | ### Describe the bug
When we use load_lora_weights for SDXL lora and model it throws below error
```
File "/workspace/app.py", line 18, in <module>
pipe.load_lora_weights(lora_model)
File "/usr/local/lib/python3.10/dist-packages/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py", li... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-07-26T12:22:20Z | 2023-09-04T15:03:44Z | 2023-09-04T15:03:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | adhikjoshi | 11,740,719 | MDQ6VXNlcjExNzQwNzE5 | User | false |
huggingface/diffusers | 1,822,483,996 | I_kwDOHa8MBc5soOYc | 4,288 | https://github.com/huggingface/diffusers/issues/4288 | https://api.github.com/repos/huggingface/diffusers/issues/4288 | [Question] Combining Image-to-Image and ControlNet | Hello,
In A1111 WebUI, it is possible to generate an image by combining and image input with a ControlNet. One use case I have seen for instance is using the image of character, and combine it with Openpose ControlNet to create variations of the character while preserving the exact same pose.
I am wondering if su... | closed | completed | false | 2 | [] | [] | 2023-07-26T13:59:09Z | 2023-07-30T07:25:45Z | 2023-07-30T07:25:45Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexisrolland | 13,064,696 | MDQ6VXNlcjEzMDY0Njk2 | User | false |
huggingface/diffusers | 1,822,619,178 | I_kwDOHa8MBc5sovYq | 4,289 | https://github.com/huggingface/diffusers/issues/4289 | https://api.github.com/repos/huggingface/diffusers/issues/4289 | [Tests] Add some tests for the principle of connected pipelines as introduced | connected pipelines were introduced in this PR: https://github.com/huggingface/diffusers/pull/4207 . We have neither added good documentation, nor good tests for this yet.
It would be great if someone could take care of this.
In terms of tests we should both test:
- that dummy pipelines that have the `self._loa... | closed | completed | false | 3 | [] | [
"DN6"
] | 2023-07-26T15:04:41Z | 2023-09-02T19:36:21Z | 2023-09-02T19:36:21Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,822,625,377 | I_kwDOHa8MBc5sow5h | 4,290 | https://github.com/huggingface/diffusers/issues/4290 | https://api.github.com/repos/huggingface/diffusers/issues/4290 | [Kandinsky] Improve useability/visibility of interpolation of Kandinsky | This PR: https://github.com/huggingface/diffusers/pull/4207 introduce a relatively simple UX to run Kandinsky pipelines end-to-end. However doing interpolation with Kandinsky still remains quite difficult:
- We don't have Auto... pipelines for interpolation
- Interpolation is attached to the less visible prior pipe... | closed | completed | false | 6 | [
"stale"
] | [
"yiyixuxu"
] | 2023-07-26T15:08:22Z | 2024-01-10T15:08:03Z | 2024-01-10T15:08:03Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,822,630,286 | I_kwDOHa8MBc5soyGO | 4,291 | https://github.com/huggingface/diffusers/issues/4291 | https://api.github.com/repos/huggingface/diffusers/issues/4291 | [Kandinsky] Correct all masking documentation of Kandinsky and advertise / document the end-to-end pipelines better | #4207 introduced end-to-end Kandinsky pipelines and also changed the masking behavior which arguably can be seen as a breaking change. We need to make sure that:
- 1.) All diffusers docs correctly reflect this breaking change
- 2.) All Kandinksy model cards correctly reflect this breaking change
- 3.) It would be ni... | closed | completed | false | 0 | [] | [
"yiyixuxu"
] | 2023-07-26T15:11:18Z | 2023-08-07T12:38:01Z | 2023-08-07T12:38:01Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,822,761,461 | I_kwDOHa8MBc5spSH1 | 4,292 | https://github.com/huggingface/diffusers/issues/4292 | https://api.github.com/repos/huggingface/diffusers/issues/4292 | inpainting pipeline crashes on M1 | ### Describe the bug
When trying to run stable diffusion on a M1 Macbook, I encountered following problem:
Calling the inpainting pipeline results in a crash. I enabled attention slicing, so it shouldn't be a memory leak.
PS: Thanks for this amazing open source software, keep up the good work!
### Reproduction
... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-07-26T16:21:28Z | 2023-08-03T10:24:19Z | 2023-07-26T16:27:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MOOOOOSER | 92,032,959 | U_kgDOBXxPvw | User | false |
huggingface/diffusers | 1,822,870,728 | I_kwDOHa8MBc5spszI | 4,293 | https://github.com/huggingface/diffusers/issues/4293 | https://api.github.com/repos/huggingface/diffusers/issues/4293 | How can I update the StableDiffusionControlNetPipeline controlnet input? | I don't see in the docs how to update the StableDiffusionControlNetPipeline controlnet input. Curious how this can be achieved. I leave my model loaded in memory and only update the settings that have changed so generation can be smooth and fast.
~~If it helps, I'm currently using diffusers 0.17.1~~
I have updated ... | closed | completed | false | 10 | [
"stale"
] | [] | 2023-07-26T17:36:20Z | 2023-11-27T09:51:28Z | 2023-11-22T15:07:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JemiloII | 5,192,788 | MDQ6VXNlcjUxOTI3ODg= | User | false |
huggingface/diffusers | 1,823,045,831 | I_kwDOHa8MBc5sqXjH | 4,296 | https://github.com/huggingface/diffusers/issues/4296 | https://api.github.com/repos/huggingface/diffusers/issues/4296 | Race condition with init_empty_weights in convert_ckpt / thread-unsafe / corrupts other modules (loading in other threads) while SD is loading | ### Describe the bug
Hi,
The `init_empty_weights()` context manager overrides the global `nn.Module.register_parameter`, which is thread-unsafe - https://github.com/huggingface/accelerate/blob/bd083ae1bfe2ed7217bc495b8ff1b109ee4c5a0a/src/accelerate/big_modeling.py#L147
This corrupts any other module being inst... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-07-26T19:40:47Z | 2023-10-04T16:20:52Z | 2023-09-03T15:03:39Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | cmdr2 | 844,287 | MDQ6VXNlcjg0NDI4Nw== | User | false |
huggingface/diffusers | 1,823,053,849 | I_kwDOHa8MBc5sqZgZ | 4,297 | https://github.com/huggingface/diffusers/issues/4297 | https://api.github.com/repos/huggingface/diffusers/issues/4297 | [SD-XL Img2Img] Batched img2img inputs cause "fishnet" artifacts and visible generative watermarks | ### Describe the bug
I have been scratching my head over this for days, as suddenly the quality of my outputs dropped off a cliff. In efforts to nail down the issue,
* Updated pytorch to nightly, no change
* Updated xformers to git master, no change
* Updated triton, pretty much every other dependency
* Used g... | closed | completed | false | 38 | [
"bug",
"Good Example PR"
] | [] | 2023-07-26T19:44:54Z | 2023-07-28T17:04:25Z | 2023-07-28T17:04:25Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 1,823,208,233 | I_kwDOHa8MBc5sq_Mp | 4,300 | https://github.com/huggingface/diffusers/issues/4300 | https://api.github.com/repos/huggingface/diffusers/issues/4300 | Multiple lora with different weights | How can I add different weights to different Loras added to the pipeline? As per my understanding, the cross_attention_kwargs only takes one argument scale, which is common for all the LoRAs.
`cross_attention_kwargs={"scale":0.6}`
@sayakpaul
| closed | completed | false | 2 | [] | [] | 2023-07-26T21:26:23Z | 2023-07-31T13:28:25Z | 2023-07-27T00:44:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shubhdotai | 6,880,985 | MDQ6VXNlcjY4ODA5ODU= | User | false |
huggingface/diffusers | 1,823,351,027 | I_kwDOHa8MBc5sriDz | 4,302 | https://github.com/huggingface/diffusers/issues/4302 | https://api.github.com/repos/huggingface/diffusers/issues/4302 | load_lora_weight() doesn't work for sdxl lora trained model from https://github.com/TheLastBen/fast-stable-diffusion | ### Describe the bug
Hi i tried using TheLastBen runpod to lora trained a model from SDXL base 0.9. I then test ran that model on ComfyUI and it was able to generate inference just fine
but when i tried to do that via code
```
STABLE_DIFFUSION_SDXL = 'stabilityai/stable-diffusion-xl-base-0.9'
pipe = DiffusionPip... | closed | completed | false | 29 | [
"bug"
] | [] | 2023-07-26T23:27:09Z | 2023-08-28T07:04:46Z | 2023-08-26T21:31:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MaxTran96 | 15,274,946 | MDQ6VXNlcjE1Mjc0OTQ2 | User | false |
huggingface/diffusers | 1,823,437,442 | I_kwDOHa8MBc5sr3KC | 4,303 | https://github.com/huggingface/diffusers/issues/4303 | https://api.github.com/repos/huggingface/diffusers/issues/4303 | `FlaxStableDiffusionPipeline` crashes when trying to load SDXL | ### Describe the bug
When loading "stabilityai/stable-diffusion-xl-base-1.0" with `from_pretrained`, I get the following error:
```
Error no file named diffusion_flax_model.msgpack or diffusion_pytorch_model.bin found in directory /root/.cache/huggingface/hub/models--stabilityai--stable-diffusion-xl-base-1.0/snapsho... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-07-27T01:18:49Z | 2023-10-20T11:11:58Z | 2023-10-20T11:11:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | davisyoshida | 1,377,776 | MDQ6VXNlcjEzNzc3NzY= | User | false |
huggingface/diffusers | 1,823,520,176 | I_kwDOHa8MBc5ssLWw | 4,304 | https://github.com/huggingface/diffusers/issues/4304 | https://api.github.com/repos/huggingface/diffusers/issues/4304 | [Training] Support for SDXL Text to Image LoRA | Currently, the `diffusers` package supports training SDXL via Dreambooth LoRA which works great for single subject/style image generation. It would be great to have a training script similar to [train_text_to_image_lora.py](https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image_lo... | closed | completed | false | 3 | [] | [] | 2023-07-27T02:52:28Z | 2023-07-31T05:53:49Z | 2023-07-31T05:53:49Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wfng92 | 43,742,196 | MDQ6VXNlcjQzNzQyMTk2 | User | false |
huggingface/diffusers | 1,823,569,665 | I_kwDOHa8MBc5ssXcB | 4,305 | https://github.com/huggingface/diffusers/issues/4305 | https://api.github.com/repos/huggingface/diffusers/issues/4305 | Diffusers v0.19.0 cause torch.compile graph break | ### Describe the bug
Diffusers v0.19.0 cause graph breaks when using torch.compile, which have been fixed in v0.18.2 (#3286)
### Reproduction
[sd pipeline from #3286](https://github.com/huggingface/diffusers/pull/3286#issue-1689544016)
### Logs
_No response_
### System Info
- `diffusers` version: 0.19.0
- Platf... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-07-27T04:04:31Z | 2023-07-27T12:23:31Z | 2023-07-27T12:23:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LSC527 | 34,333,110 | MDQ6VXNlcjM0MzMzMTEw | User | false |
huggingface/diffusers | 1,823,571,765 | I_kwDOHa8MBc5ssX81 | 4,306 | https://github.com/huggingface/diffusers/issues/4306 | https://api.github.com/repos/huggingface/diffusers/issues/4306 | StableDiffusionReferencePipeline throws list index out of range error | ### Describe the bug
`StableDiffusionReferencePipeline` throws `list index out of range` error when you have `batch_size` over 1 and `guidance_scale` over 1.
Here's how I'm trying to call the pipeline:
```
generator = [torch.Generator(device="cuda").manual_seed(random.randint(1, 250)) for i in range(4)]
pipe(
... | closed | completed | false | 3 | [
"bug"
] | [
"DN6"
] | 2023-07-27T04:07:28Z | 2023-08-08T13:16:18Z | 2023-08-08T13:16:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ethankongee | 123,821,603 | U_kgDOB2FeIw | User | false |
huggingface/diffusers | 1,823,582,121 | I_kwDOHa8MBc5ssaep | 4,307 | https://github.com/huggingface/diffusers/issues/4307 | https://api.github.com/repos/huggingface/diffusers/issues/4307 | from_pretrained error when local_files_only=True | ### Describe the bug
After all the files have been downloaded locally, changing the value of `local_files_only` to `True` will raise reference before assignment error.
### Reproduction
```python
from diffusers import StableDiffusionXLPipeline
import torch
pipe = StableDiffusionXLPipeline.from_pretrained(
... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-07-27T04:17:12Z | 2023-07-31T02:58:45Z | 2023-07-31T02:58:45Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wfng92 | 43,742,196 | MDQ6VXNlcjQzNzQyMTk2 | User | false |
huggingface/diffusers | 1,823,764,782 | I_kwDOHa8MBc5stHEu | 4,308 | https://github.com/huggingface/diffusers/issues/4308 | https://api.github.com/repos/huggingface/diffusers/issues/4308 | discrepancy between the default value of `guidance_scale` of `StableDiffusionXLPipeline` and its docstring description | The default value seems to be set to 5.0, but the docstring says that it's 7.5.
https://github.com/huggingface/diffusers/blob/20e92586c1fda968ea3343ba0f44f2b21f3c09d2/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py#L557
https://github.com/huggingface/diffusers/blob/20e92586c1fda968ea3343ba... | closed | completed | false | 6 | [] | [] | 2023-07-27T07:15:26Z | 2023-07-28T17:16:12Z | 2023-07-28T17:16:12Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hysts | 25,161,192 | MDQ6VXNlcjI1MTYxMTky | User | false |
huggingface/diffusers | 1,366,789,940 | I_kwDOHa8MBc5Rd480 | 431 | https://github.com/huggingface/diffusers/issues/431 | https://api.github.com/repos/huggingface/diffusers/issues/431 | AttributeError: 'AutoencoderKLOutput' object has no attribute 'sample' | ### Describe the bug
running notebook https://colab.research.google.com/github/patil-suraj/Notebooks/blob/master/image_2_image_using_diffusers.ipynb
when running this cell (after previously run all the preceding cell )
generator = torch.Generator(device=device).manual_seed(1024)
with autocast("cuda"):
imag... | closed | completed | false | 5 | [
"bug"
] | [] | 2022-09-08T18:16:13Z | 2022-09-08T19:53:33Z | 2022-09-08T19:29:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nunezmatias | 88,870,315 | MDQ6VXNlcjg4ODcwMzE1 | User | false |
huggingface/diffusers | 1,823,894,074 | I_kwDOHa8MBc5stmo6 | 4,310 | https://github.com/huggingface/diffusers/issues/4310 | https://api.github.com/repos/huggingface/diffusers/issues/4310 | Staircase/Scanline artifacts with SDXL 1.0 | ### Describe the bug
SDXL creates colorful scanline-like artifacts, especially on areas with high-frequency detail.
SDXL - Diffusers 0.19.0 - 25 Steps of DPMSolverMultistepScheduler with Karras - 20% refine steps - CFG Scale 7
Left: SDXL 1.0 - Right: SDXL 0.9
![image](https://github.com/huggingface/diffuser... | closed | completed | false | 20 | [
"bug",
"stale"
] | [] | 2023-07-27T08:39:54Z | 2024-02-19T15:06:19Z | 2024-02-19T15:06:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | n00mkrad | 61,149,547 | MDQ6VXNlcjYxMTQ5NTQ3 | User | false |
huggingface/diffusers | 1,823,917,633 | I_kwDOHa8MBc5stsZB | 4,311 | https://github.com/huggingface/diffusers/issues/4311 | https://api.github.com/repos/huggingface/diffusers/issues/4311 | local_files_only=True, variable pipeline_class referenced before assignment | https://github.com/huggingface/diffusers/blob/20e92586c1fda968ea3343ba0f44f2b21f3c09d2/src/diffusers/pipelines/pipeline_utils.py#L1474C1-L1483C4 | closed | completed | false | 2 | [] | [] | 2023-07-27T08:54:34Z | 2023-08-03T10:25:42Z | 2023-08-03T10:25:42Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Glaceon-Hyy | 8,824,848 | MDQ6VXNlcjg4MjQ4NDg= | User | false |
huggingface/diffusers | 1,823,933,044 | I_kwDOHa8MBc5stwJ0 | 4,312 | https://github.com/huggingface/diffusers/issues/4312 | https://api.github.com/repos/huggingface/diffusers/issues/4312 | SAG for inpainting and img2img | Hello, SAG is great and produces great results, but there are no pipelines that use it with img2img or inpainting. Can it be done?
this is the pipeline btw:
https://github.com/huggingface/diffusers/blob/8e5921cac14c84a670b4e9d89d717b5356e5b9e9/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_sag.p... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-07-27T09:02:35Z | 2023-10-30T15:09:10Z | 2023-10-30T15:09:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,823,998,400 | I_kwDOHa8MBc5suAHA | 4,313 | https://github.com/huggingface/diffusers/issues/4313 | https://api.github.com/repos/huggingface/diffusers/issues/4313 | Possible misuse of scheduler in function log_validation in train_dreambooth.py | I copy train_dreambooth.py and use it to fully train IF Stage II. During training, I found it is unavoidable to generate images with noise point.I personally think it is due to the mandatory use of DPMSolverMultistepScheduler in function log_validation. When I try to use the origin scheduler DDPMScheduler offered by (h... | closed | completed | false | 5 | [] | [] | 2023-07-27T09:38:24Z | 2023-08-23T20:18:19Z | 2023-08-23T20:18:19Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | YinzhenWang | 43,692,893 | MDQ6VXNlcjQzNjkyODkz | User | false |
huggingface/diffusers | 1,824,032,716 | I_kwDOHa8MBc5suIfM | 4,314 | https://github.com/huggingface/diffusers/issues/4314 | https://api.github.com/repos/huggingface/diffusers/issues/4314 | Files will be WriteProtected if I interrupt download process | ### Describe the bug
Hello, I am run this demo on WSL and the model is installing
```
import torch
from diffusers import StableDiffusionPipeline
import matplotlib.pyplot as plt
pipe = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16)
pipe = pipe.to("cuda")
... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-07-27T09:58:29Z | 2023-09-03T15:03:37Z | 2023-09-03T15:03:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | StudyingLover | 68,844,494 | MDQ6VXNlcjY4ODQ0NDk0 | User | false |
huggingface/diffusers | 1,824,206,982 | I_kwDOHa8MBc5suzCG | 4,316 | https://github.com/huggingface/diffusers/issues/4316 | https://api.github.com/repos/huggingface/diffusers/issues/4316 | Error occurs when using validation option of train_dreambooth.py | ### Describe the bug
Training with validation option on failed for train_dreambooth.py and train_dreambooth_lora.py. After specify `--validation_prompt`, the program exits at validation stage, complaining about missing attribute of Namespace.
Turns out that vars() functions added in #4248 only do shallow copy of at... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-07-27T11:50:45Z | 2023-07-27T16:58:55Z | 2023-07-27T16:58:55Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xinyangli | 16,359,093 | MDQ6VXNlcjE2MzU5MDkz | User | false |
huggingface/diffusers | 1,824,758,689 | I_kwDOHa8MBc5sw5uh | 4,321 | https://github.com/huggingface/diffusers/issues/4321 | https://api.github.com/repos/huggingface/diffusers/issues/4321 | OnnxStableDiffusionPipeline produces black images when using an FP16 model [DirectML] | ### Describe the bug
OnnxStableDiffusionPipeline generates an all-black image when loading an FP16 ONNX model.
### Reproduction
- Convert SD 1.5 (Diffusers) to ONNX using `convert_stable_diffusion_checkpoint_to_onnx.py --fp16`
- Run the script below, replace `model_path` with the path to your converted SD 1.5... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-07-27T16:31:40Z | 2023-12-12T09:34:18Z | 2023-09-04T15:03:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | n00mkrad | 61,149,547 | MDQ6VXNlcjYxMTQ5NTQ3 | User | false |
huggingface/diffusers | 1,825,030,412 | I_kwDOHa8MBc5sx8EM | 4,324 | https://github.com/huggingface/diffusers/issues/4324 | https://api.github.com/repos/huggingface/diffusers/issues/4324 | TypeError: 'NoneType' not iterable in forward() of unet_2d_condition.py | ### Describe the bug
I tried to run train_dreambooth.py using the following parameters:
```
--pretrained_model_name_or_path="stabilityai/stable-diffusion-xl-base-1.0" \
--instance_data_dir="./photos" \
--class_data_dir="./class_man" \
--output_dir="./results" \
--with_prior_preservation --prior_loss... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-07-27T19:19:26Z | 2023-08-16T07:05:43Z | 2023-08-04T02:58:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | StefosGeo | 29,986,091 | MDQ6VXNlcjI5OTg2MDkx | User | false |
huggingface/diffusers | 1,825,031,417 | I_kwDOHa8MBc5sx8T5 | 4,325 | https://github.com/huggingface/diffusers/issues/4325 | https://api.github.com/repos/huggingface/diffusers/issues/4325 | Black image on SDXL 1.0 + RuntimeWarning: invalid value encountered in cast images = (images * 255).round().astype("uint8") | ### Describe the bug
I'm trying to run SDXL in a container environment (Debian). I tried both diffusers[torch]==0.18.0 and diffusers[torch]==0.19.0 with Python 3.10 on A10g and A100 but I get a black image back and RuntimeWarning: invalid value encountered in cast
images = (images * 255).round().astype("uint8") whe... | closed | completed | false | 17 | [
"bug",
"stale"
] | [] | 2023-07-27T19:19:48Z | 2024-03-23T14:25:01Z | 2023-09-06T15:03:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | belfortf | 24,974,712 | MDQ6VXNlcjI0OTc0NzEy | User | false |
huggingface/diffusers | 1,825,578,912 | I_kwDOHa8MBc5s0B-g | 4,328 | https://github.com/huggingface/diffusers/issues/4328 | https://api.github.com/repos/huggingface/diffusers/issues/4328 | 19.0 Breaks loading and unloading LoRAs with kohya_lora_loader | ### Describe the bug
So in 18.2 I was using the kohya_lora_loader made by @takuma104 to load and unload multiple LoRAs. For most LoRA that wasn't LyCoris or KoHya, it worked quite well. I could have as many LoRAs as I wanted and was able to unload them and get the same images generated.
In version 19, it breaks. It... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-07-28T02:23:24Z | 2023-10-19T14:09:27Z | 2023-09-04T15:03:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JemiloII | 5,192,788 | MDQ6VXNlcjUxOTI3ODg= | User | false |
huggingface/diffusers | 1,825,712,284 | I_kwDOHa8MBc5s0iic | 4,331 | https://github.com/huggingface/diffusers/issues/4331 | https://api.github.com/repos/huggingface/diffusers/issues/4331 | Different results when using train_dreambooth.py and train_dreambooth_lora.py | **I am using exact same parameters for both train_dreambooth.py and train_dreambooth_lora.py while training, here is the script**
```
!accelerate launch train_dreambooth_lora.py \
--pretrained_model_name_or_path="SG161222/Realistic_Vision_V4.0" \
--instance_data_dir="training_images" \
--output_dir="result"... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-07-28T05:07:34Z | 2023-09-06T15:03:27Z | 2023-09-06T15:03:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nayan-dhabarde | 24,761,808 | MDQ6VXNlcjI0NzYxODA4 | User | false |
huggingface/diffusers | 1,825,779,924 | I_kwDOHa8MBc5s0zDU | 4,332 | https://github.com/huggingface/diffusers/issues/4332 | https://api.github.com/repos/huggingface/diffusers/issues/4332 | Fail to use xformers for kandinsky2.2 pipeline | ### Describe the bug
Got error `NotImplementedError: No operator found for this attention`. Does the pipeline support xformers yet considering the fact that it is a fairly new pipeline?
### Reproduction
```python
from diffusers import AutoPipelineForText2Image
import torch
pipe = AutoPipelineForText2Image... | closed | completed | false | 3 | [
"bug"
] | [
"yiyixuxu"
] | 2023-07-28T06:15:12Z | 2023-12-14T00:24:29Z | 2023-08-04T01:11:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eeyrw | 2,088,817 | MDQ6VXNlcjIwODg4MTc= | User | false |
huggingface/diffusers | 1,825,873,787 | I_kwDOHa8MBc5s1J97 | 4,334 | https://github.com/huggingface/diffusers/issues/4334 | https://api.github.com/repos/huggingface/diffusers/issues/4334 | [SDXL 1.0]Slow compilation caused by torch.nonzero in scheduling_euler_discrete.py | **Is your feature request related to a problem? Please describe.**
Yes. The problem is slow compilation caused by torch.nonzero.
I'm running the diffuser with a TPU-like device with a XLA backend.
It works great with the huggingface diffusers on models like stable-diffusion 1/2.
Today I was trying the recently ... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-07-28T07:32:49Z | 2023-09-04T15:03:36Z | 2023-09-04T15:03:36Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kevint324 | 8,800,468 | MDQ6VXNlcjg4MDA0Njg= | User | false |
huggingface/diffusers | 1,826,065,413 | I_kwDOHa8MBc5s14wF | 4,336 | https://github.com/huggingface/diffusers/issues/4336 | https://api.github.com/repos/huggingface/diffusers/issues/4336 | Feature extraction from Unet of Stable Diffusion | Thanks to the HF community for this wonderful library. I'm currently working on a project where we are interested in looking into the intermediate features(feature maps from different layers) of the Unet in Stable diffusion model.
The `transformers` library has the option of `output_hidden_states` to extract the int... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-07-28T09:25:07Z | 2023-11-08T15:10:12Z | 2023-11-08T15:10:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | danushkv | 64,687,472 | MDQ6VXNlcjY0Njg3NDcy | User | false |
huggingface/diffusers | 1,826,114,468 | I_kwDOHa8MBc5s2Euk | 4,337 | https://github.com/huggingface/diffusers/issues/4337 | https://api.github.com/repos/huggingface/diffusers/issues/4337 | torch.compile fails with Community ControlNetReference | ### Describe the bug
Running torch.compile on unet for stable_diffusion_controlnet_reference errors out. I'll note that VAE will compile. Error below. Also to note, I'm running other StabeDiffusion pipelines with the dependencies listed below and they compile just fine.
```
torch._dynamo.utils: [WARNING] Uns... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-07-28T09:51:50Z | 2023-09-04T15:03:34Z | 2023-09-04T15:03:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | djj0s3 | 1,120,027 | MDQ6VXNlcjExMjAwMjc= | User | false |
huggingface/diffusers | 1,366,976,952 | I_kwDOHa8MBc5Remm4 | 434 | https://github.com/huggingface/diffusers/issues/434 | https://api.github.com/repos/huggingface/diffusers/issues/434 | Training notebook does not seem to work with 0.3.0 | ### Describe the bug
The training notebook doesn't seem to work with 0.3.0
### Reproduction
Running this google colab: https://colab.research.google.com/drive/1wAOAo9LCjWukVp6QBwpw6EwoVhsdyuNj?usp=sharing
### Logs
```shell
AttributeError Traceback (most recent call last)
<ipython-input-... | closed | completed | false | 2 | [
"bug"
] | [
"anton-l"
] | 2022-09-08T20:56:18Z | 2022-09-19T15:33:38Z | 2022-09-19T15:33:38Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,826,494,900 | I_kwDOHa8MBc5s3hm0 | 4,340 | https://github.com/huggingface/diffusers/issues/4340 | https://api.github.com/repos/huggingface/diffusers/issues/4340 | SDXL LoRA training error with train_text_encoder | ### Describe the bug
wrt train_dreambooth_lora_sdxl.py
SDXL unet is conditioned on the following from the text_encoders:
* hidden_states of the penultimate layer from encoder one
* hidden_states of the penultimate layer from encoder two
* pooled hidden states from encoder two
When creating the LoRA adapter ... | closed | completed | false | 12 | [
"bug",
"stale"
] | [] | 2023-07-28T14:07:19Z | 2023-09-12T15:04:02Z | 2023-09-12T15:04:02Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jorgemcgomes | 3,987,574 | MDQ6VXNlcjM5ODc1NzQ= | User | false |
huggingface/diffusers | 1,826,536,640 | I_kwDOHa8MBc5s3rzA | 4,341 | https://github.com/huggingface/diffusers/issues/4341 | https://api.github.com/repos/huggingface/diffusers/issues/4341 | Reconsider pooled_prompt_embeds for SDXL | The current API with Compel uses both the `prompt_embeds` and `pooled_prompt_embeds` parameters which is different than other diffusion pipelines. The new Auto pipelines might be interesting to use, but currently I don't see that I can use them because of the API divergence.
If `prompt_embeds` for SDXL pipeline ... | closed | completed | false | 16 | [
"stale"
] | [] | 2023-07-28T14:32:10Z | 2023-10-30T15:09:06Z | 2023-10-30T15:09:06Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pdoane | 1,621,950 | MDQ6VXNlcjE2MjE5NTA= | User | false |
huggingface/diffusers | 1,826,929,962 | I_kwDOHa8MBc5s5L0q | 4,348 | https://github.com/huggingface/diffusers/issues/4348 | https://api.github.com/repos/huggingface/diffusers/issues/4348 | Improved support for working with Kohya-style LoRAs in diffusers 🤗 | A seamless interoperability between the [Kohya-styled](https://github.com/Linaqruf/kohya-trainer) LoRAs and Diffusers has been one of the most requested features from the community in the last months.
We are making promising progress in this regard.
With #4287, this support should be quite improved. We also hav... | closed | completed | false | 34 | [] | [] | 2023-07-28T18:46:04Z | 2023-10-17T05:52:05Z | 2023-10-17T05:52:04Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,826,940,118 | I_kwDOHa8MBc5s5OTW | 4,349 | https://github.com/huggingface/diffusers/issues/4349 | https://api.github.com/repos/huggingface/diffusers/issues/4349 | StableDiffusionUpscalePipeline fails without a text_encoder despite prompt_embeds being already encoded | ### Describe the bug
When we've already constructed `prompt_embeds`, StableDiffusionUpscalePipeline shouldn't require a tokenizer or text encoder, since there's no more tokenizing or text encoding to do. But it fails, as it references the encoder anyway.
Similar use case to #4249.
I think there are a **lot** o... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-07-28T18:53:11Z | 2023-08-03T04:25:40Z | 2023-08-03T04:25:40Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | keturn | 83,819 | MDQ6VXNlcjgzODE5 | User | false |
huggingface/diffusers | 1,366,985,616 | I_kwDOHa8MBc5ReouQ | 435 | https://github.com/huggingface/diffusers/issues/435 | https://api.github.com/repos/huggingface/diffusers/issues/435 | Textual inversion training notebook, 'AutoencoderKLOutput' object has no attribute 'sample' | ### Describe the bug
Been getting this error  was working last night... I've tried restarting/refreshing/New notebook, etc no luck. How do I fix this? thanks!
### Reproduction
_No resp... | closed | completed | false | 3 | [
"bug"
] | [] | 2022-09-08T21:03:02Z | 2022-09-13T16:10:17Z | 2022-09-13T16:10:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | crypticsymmetry | 29,524,706 | MDQ6VXNlcjI5NTI0NzA2 | User | false |
huggingface/diffusers | 1,827,309,137 | I_kwDOHa8MBc5s6oZR | 4,353 | https://github.com/huggingface/diffusers/issues/4353 | https://api.github.com/repos/huggingface/diffusers/issues/4353 | SDXL image generation bug, has poor quality. | ### Describe the bug
All images generate some green-red blobs
... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-07-29T04:17:50Z | 2023-08-03T18:06:18Z | 2023-07-29T05:13:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | conansherry | 2,148,224 | MDQ6VXNlcjIxNDgyMjQ= | User | false |
huggingface/diffusers | 1,827,331,708 | I_kwDOHa8MBc5s6t58 | 4,354 | https://github.com/huggingface/diffusers/issues/4354 | https://api.github.com/repos/huggingface/diffusers/issues/4354 | StableDiffusionXLInpaintPipeline.from_single_file not returning Pipeline of type StableDiffusionXLInpaintPipeline | ### Describe the bug
The documentation says we can use pipelines of SDXL by giving from_single_file, but when SDXL Inpaint pipeline is instantiated using from_single_file, its returning the base SDXL pipeline type.
### Reproduction
pipe = StableDiffusionXLInpaintPipeline.from_single_file(f"{inference_provider_metada... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-07-29T05:42:20Z | 2023-09-06T15:03:22Z | 2023-09-06T15:03:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | polavishnu4444 | 2,263,240 | MDQ6VXNlcjIyNjMyNDA= | User | false |
huggingface/diffusers | 1,827,352,491 | I_kwDOHa8MBc5s6y-r | 4,356 | https://github.com/huggingface/diffusers/issues/4356 | https://api.github.com/repos/huggingface/diffusers/issues/4356 | 0.19.2 ModuleNotFoundError: No module named 'diffusers.utils.dummy_torch_and_transformers_and_objects' | ### Describe the bug
ModuleNotFoundError: No module named 'diffusers.utils.dummy_torch_and_transformers_and_objects'
### Reproduction
install 0.19.2 version of diffusers with python 3.10
### Logs
_No response_
### System Info
- `diffusers` version: 0.18.2
- Platform: Linux-5.15.90.1-microsoft-standard-WSL2-x86_... | closed | completed | false | 9 | [
"bug",
"stale"
] | [
"DN6"
] | 2023-07-29T06:56:18Z | 2024-04-07T01:24:30Z | 2023-10-30T15:09:03Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | npuichigo | 11,533,479 | MDQ6VXNlcjExNTMzNDc5 | User | false |
huggingface/diffusers | 1,827,457,149 | I_kwDOHa8MBc5s7Mh9 | 4,359 | https://github.com/huggingface/diffusers/issues/4359 | https://api.github.com/repos/huggingface/diffusers/issues/4359 | launch issue of Stable Diffusion UI with Python error | After doing a git pull for the lat latest update , Stable Diffusion will not loadthe UI and i get the folling error(s)
error: Your local changes to the following files would be overwritten by merge:
launch.py
modules/styles.py
Please commit your changes or stash them before you merge.
Aborting
U... | closed | completed | false | 1 | [] | [] | 2023-07-29T12:04:14Z | 2023-08-02T09:30:31Z | 2023-08-02T09:30:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | gotosoda | 116,220,725 | U_kgDOBu1jNQ | User | false |
huggingface/diffusers | 1,367,023,137 | I_kwDOHa8MBc5Rex4h | 436 | https://github.com/huggingface/diffusers/issues/436 | https://api.github.com/repos/huggingface/diffusers/issues/436 | Init Image strenght issue | Hi there!
From the pipeline:
`strength (`float`, *optional*, defaults to 0.8):
Conceptually, indicates how much to transform the reference `init_image`. Must be between 0 and 1.
`init_image` will be used as a starting point, adding more noise to it the larger the `strength`. The
... | closed | completed | false | 4 | [
"stale"
] | [] | 2022-09-08T21:47:55Z | 2022-10-31T15:03:23Z | 2022-10-31T15:03:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | brurpo | 24,902,022 | MDQ6VXNlcjI0OTAyMDIy | User | false |
huggingface/diffusers | 1,827,491,489 | I_kwDOHa8MBc5s7U6h | 4,360 | https://github.com/huggingface/diffusers/issues/4360 | https://api.github.com/repos/huggingface/diffusers/issues/4360 | how can i run lora training with sdxl1.0 on google colab? | git clone https://github.com/huggingface/diffusers
cd diffusers
pip install -e .
after i run the above code on colab and finish lora training,then execute the following python code:
from huggingface_hub.repocard import RepoCard
from diffusers import DiffusionPipeline
import torch
lora_model_id = "lora-trai... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-07-29T13:43:52Z | 2023-09-06T15:03:20Z | 2023-09-06T15:03:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | frankchieng | 130,369,523 | U_kgDOB8VH8w | User | false |
huggingface/diffusers | 1,827,525,120 | I_kwDOHa8MBc5s7dIA | 4,361 | https://github.com/huggingface/diffusers/issues/4361 | https://api.github.com/repos/huggingface/diffusers/issues/4361 | The shared tensors saving issue caused by safetensors comes to diffusers after updating diffusers and related libs | ### Describe the bug
It is about https://github.com/huggingface/safetensors/issues/202 and https://github.com/huggingface/transformers/pull/22437. For certain use case, the issue finally comes to diffusers. According my observation, the shared tensors are introduced by `ZeroRedundancyOptimizer`😏 and then comes to `... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-07-29T15:37:08Z | 2023-10-18T15:15:00Z | 2023-10-18T15:15:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eeyrw | 2,088,817 | MDQ6VXNlcjIwODg4MTc= | User | false |
huggingface/diffusers | 1,827,578,004 | I_kwDOHa8MBc5s7qCU | 4,362 | https://github.com/huggingface/diffusers/issues/4362 | https://api.github.com/repos/huggingface/diffusers/issues/4362 | Changing the dtype of pipeline while it has been offloaded will result in exception due to warning message | If a pipeline is set up, offloaded, and at a later time, `.to(torch_dtype=$type)` is called, the linked line will result in panic as `torch_device` is None.
https://github.com/huggingface/diffusers/blob/34b5b63bb893e05c9f8cdefa513ddc02bbd94b51/src/diffusers/pipelines/pipeline_utils.py#L680 | closed | completed | false | 4 | [] | [] | 2023-07-29T18:01:29Z | 2023-08-25T09:13:33Z | 2023-08-25T09:13:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Kubuxu | 2,259,282 | MDQ6VXNlcjIyNTkyODI= | User | false |
huggingface/diffusers | 1,827,668,196 | I_kwDOHa8MBc5s8ADk | 4,363 | https://github.com/huggingface/diffusers/issues/4363 | https://api.github.com/repos/huggingface/diffusers/issues/4363 | how to properly load sd_xl_base_1.0_0.9vae.safetensors | ### Describe the bug
hi, how should i load sd_xl_base_1.0_0.9vae.safetensors given the namespace is the same as 1.0 one?
### Reproduction
N/A
### Logs
_No response_
### System Info
ec2
### Who can help?
@sayakpaul @patrick | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-07-29T21:16:34Z | 2023-10-18T15:14:58Z | 2023-10-18T15:14:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MaxTran96 | 15,274,946 | MDQ6VXNlcjE1Mjc0OTQ2 | User | false |
huggingface/diffusers | 1,827,673,251 | I_kwDOHa8MBc5s8BSj | 4,364 | https://github.com/huggingface/diffusers/issues/4364 | https://api.github.com/repos/huggingface/diffusers/issues/4364 | SDXL Inpaint Refiner failing for non-square images. | ### Describe the bug
Using the ensemble of diffusers method on a non-square image SDXL Inpaint is failing to perform erroring out saying ` The size of tensor a (96) must match the size of tensor b (128) at non-singleton dimension 2`
Input dimensions: 1024 X 768
lib/python3.9/site-packages/diffusers/schedulers/sc... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-07-29T21:26:04Z | 2023-07-30T03:01:08Z | 2023-07-30T03:01:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | polavishnu4444 | 2,263,240 | MDQ6VXNlcjIyNjMyNDA= | User | false |
huggingface/diffusers | 1,827,807,984 | I_kwDOHa8MBc5s8iLw | 4,366 | https://github.com/huggingface/diffusers/issues/4366 | https://api.github.com/repos/huggingface/diffusers/issues/4366 | [SD-XL] Fine-tuning text-to-image script for full U-net | **Is your feature request related to a problem? Please describe.**
The current options available for fine-tuning SDXL are currently inadequate for training a new noise schedule into the base U-net.
**Describe the solution you'd like**
I would like a replica of the Stable Diffusion 1.5/2.1 text-to-image scripts... | closed | completed | false | 17 | [
"contributions-welcome"
] | [] | 2023-07-30T05:17:18Z | 2023-08-16T03:32:51Z | 2023-08-16T03:32:51Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 1,827,854,369 | I_kwDOHa8MBc5s8tgh | 4,367 | https://github.com/huggingface/diffusers/issues/4367 | https://api.github.com/repos/huggingface/diffusers/issues/4367 | Please support from_single_file() for AutoPipeline | Please support from_single_file() for AutoPipeline!!! | closed | completed | false | 4 | [] | [] | 2023-07-30T08:37:46Z | 2023-08-09T08:47:08Z | 2023-08-09T08:47:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wodsoe | 40,934,627 | MDQ6VXNlcjQwOTM0NjI3 | User | false |
huggingface/diffusers | 1,827,870,979 | I_kwDOHa8MBc5s8xkD | 4,368 | https://github.com/huggingface/diffusers/issues/4368 | https://api.github.com/repos/huggingface/diffusers/issues/4368 | LoRA training for sdxl on diffusers CUDA out of memory? | ### Describe the bug
when i train lora thr Zero-2 stage of deepspeed and offload optimizer states and parameters to CPU, torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 26.00 MiB (GPU 0; 14.75 GiB total capacity; 14.06 GiB already allocated; 8.81 MiB free; 14.38 GiB reserved in total by PyTorch)
wh... | closed | completed | false | 8 | [
"bug"
] | [] | 2023-07-30T09:38:07Z | 2023-08-09T08:59:42Z | 2023-08-09T08:59:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | frankchieng | 130,369,523 | U_kgDOB8VH8w | User | false |
huggingface/diffusers | 1,367,161,387 | I_kwDOHa8MBc5RfTor | 437 | https://github.com/huggingface/diffusers/issues/437 | https://api.github.com/repos/huggingface/diffusers/issues/437 | Explanation of the 0.18215 factor in textual_inversion? | https://github.com/huggingface/diffusers/blob/b2b3b1a8ab83b020ecaf32f45de3ef23644331cf/examples/textual_inversion/textual_inversion.py#L501
Hi, just a small question about the quoted script above which is bothering me: where does this `0.18215` number come from? What computation is being done? Is it from some paper... | closed | completed | false | 10 | [] | [] | 2022-09-09T01:21:39Z | 2025-02-28T21:13:51Z | 2022-09-09T13:07:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | garrett361 | 44,747,910 | MDQ6VXNlcjQ0NzQ3OTEw | User | false |
huggingface/diffusers | 1,827,895,152 | I_kwDOHa8MBc5s83dw | 4,372 | https://github.com/huggingface/diffusers/issues/4372 | https://api.github.com/repos/huggingface/diffusers/issues/4372 | original_size, and target_size have no impact for StableDiffusionXL | ### Describe the bug
height and width get set to defaults despite use of original_size or target_size. I think the order of the pairs of lines need to be swapped.
### Reproduction
```
from diffusers import StableDiffusionXLPipeline
import torch
pipe = StableDiffusionXLPipeline.from_pretrained(
"stab... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-07-30T11:07:50Z | 2024-02-07T05:52:23Z | 2023-08-09T08:54:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | GeorgePearse | 47,161,914 | MDQ6VXNlcjQ3MTYxOTE0 | User | false |
huggingface/diffusers | 1,827,906,619 | I_kwDOHa8MBc5s86Q7 | 4,374 | https://github.com/huggingface/diffusers/issues/4374 | https://api.github.com/repos/huggingface/diffusers/issues/4374 | Issue in StableDiffusionControlNetInpaintPipeline | ### Describe the bug
<img width="784" alt="Screenshot 2023-07-30 at 5 18 28 PM" src="https://github.com/huggingface/diffusers/assets/46415298/4fdcc8df-460b-44d6-b7b9-4d64377dfc58">
there is one small mistake in the "StableDiffusionControlnetInpaintPipeline",
in line 1268
`control_model_input` shoud have 9 dime... | closed | completed | false | 14 | [
"bug",
"stale"
] | [
"yiyixuxu"
] | 2023-07-30T11:49:53Z | 2023-10-17T12:39:58Z | 2023-09-30T04:23:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Onkarsus13 | 46,415,298 | MDQ6VXNlcjQ2NDE1Mjk4 | User | false |
huggingface/diffusers | 1,828,088,888 | I_kwDOHa8MBc5s9mw4 | 4,375 | https://github.com/huggingface/diffusers/issues/4375 | https://api.github.com/repos/huggingface/diffusers/issues/4375 | Torch 2 compile causes crash during inference | ### Describe the bug
I was able to create this using StableDiffusionXL and StableDiffusion pipelines. The crash happens during inference.
File "test.py", line 11, in <module>
image = pipe(prompt=prompt).images[0]
File "/opt/venv/lib/python3.8/site-packages/torch/utils/_contextlib.py", line 115, in decor... | closed | completed | false | 19 | [
"bug",
"stale"
] | [] | 2023-07-30T21:11:59Z | 2023-11-29T15:06:43Z | 2023-11-29T15:06:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | frankjoshua | 230,666 | MDQ6VXNlcjIzMDY2Ng== | User | false |
huggingface/diffusers | 1,828,089,972 | I_kwDOHa8MBc5s9nB0 | 4,376 | https://github.com/huggingface/diffusers/issues/4376 | https://api.github.com/repos/huggingface/diffusers/issues/4376 | StableDiffusionXLPipeline crashes when calling load_textual_inversion | ### Describe the bug
The documentation shows that SDXL supports textual inversion.
Traceback (most recent call last):
File "test.py", line 7, in <module>
pipe.load_textual_inversion("./embedding.pt")
File "/opt/venv/lib/python3.8/site-packages/diffusers/configuration_utils.py", line 137, in __getattr__
... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-07-30T21:16:19Z | 2023-08-02T09:59:48Z | 2023-08-02T09:59:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | frankjoshua | 230,666 | MDQ6VXNlcjIzMDY2Ng== | User | false |
huggingface/diffusers | 1,828,225,344 | I_kwDOHa8MBc5s-IFA | 4,377 | https://github.com/huggingface/diffusers/issues/4377 | https://api.github.com/repos/huggingface/diffusers/issues/4377 | Gradient checkpointing not applied to UNet mid_block | The mid block of the SDXL is huge, so fixing it significantly reduces VRAM usage.
I have tested the following changes and have seen great results.
https://github.com/huggingface/diffusers/compare/main...laksjdjf:diffusers:mid_block_gradient_checkpointing
However, there seem to be several variations of midblock, ... | closed | completed | false | 2 | [] | [] | 2023-07-31T01:59:52Z | 2023-08-04T18:07:38Z | 2023-08-04T18:07:38Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | laksjdjf | 22,386,664 | MDQ6VXNlcjIyMzg2NjY0 | User | false |
huggingface/diffusers | 1,828,295,290 | I_kwDOHa8MBc5s-ZJ6 | 4,379 | https://github.com/huggingface/diffusers/issues/4379 | https://api.github.com/repos/huggingface/diffusers/issues/4379 | Lora extension with weight | First of all. Thank you for your effort.
I have some question when loading lora weight in sd.
1. Is it the right way that putting sd model in this way? I mean it doesn't have to be converted with like "convert_models_diffuser_to_diffusers.py" in diffusers?
`model_base = "/hdd/jinnnn/anylora.safetensors"`
`pip... | closed | completed | false | 2 | [] | [] | 2023-07-31T02:58:08Z | 2023-08-02T09:27:37Z | 2023-08-02T09:27:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Gynjn | 106,458,325 | U_kgDOBlhs1Q | User | false |
huggingface/diffusers | 1,367,289,302 | I_kwDOHa8MBc5Rfy3W | 438 | https://github.com/huggingface/diffusers/issues/438 | https://api.github.com/repos/huggingface/diffusers/issues/438 | Out painting | Is there outpainting example? | closed | completed | false | 12 | [
"stale"
] | [] | 2022-09-09T04:59:45Z | 2022-11-28T15:03:56Z | 2022-11-28T15:03:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wodsoe | 40,934,627 | MDQ6VXNlcjQwOTM0NjI3 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.