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 | 1,845,076,678 | I_kwDOHa8MBc5t-aLG | 4,560 | https://github.com/huggingface/diffusers/issues/4560 | https://api.github.com/repos/huggingface/diffusers/issues/4560 | support for python 3.11 | Trying to run diffusers with python 3.11 however the module cannot be found when attempting to run from this python version
It would be ideal to enable this as python 3.11 bring significant speed improvements over previous versions
| closed | completed | false | 2 | [] | [
"patrickvonplaten"
] | 2023-08-10T12:21:40Z | 2023-08-25T13:26:39Z | 2023-08-25T13:26:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shott92 | 37,227,619 | MDQ6VXNlcjM3MjI3NjE5 | User | false |
huggingface/diffusers | 1,845,289,560 | I_kwDOHa8MBc5t_OJY | 4,561 | https://github.com/huggingface/diffusers/issues/4561 | https://api.github.com/repos/huggingface/diffusers/issues/4561 | Please add “original_config_file” argument for download_from_original_stable_diffusion_ckpt() | Please add “original_config_file” argument for download_from_original_stable_diffusion_ckpt(),
then can set a local config file.
:
... | closed | completed | false | 2 | [] | [] | 2023-08-10T15:46:51Z | 2023-08-14T12:16:03Z | 2023-08-14T12:16:03Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dai-ichiro | 53,703,225 | MDQ6VXNlcjUzNzAzMjI1 | User | false |
huggingface/diffusers | 1,846,098,048 | I_kwDOHa8MBc5uCTiA | 4,566 | https://github.com/huggingface/diffusers/issues/4566 | https://api.github.com/repos/huggingface/diffusers/issues/4566 | SDXL LoRA training, cannot resume from checkpoint | ### Describe the bug
I am able to train a SDXL Lora no problem. However, when I tried to resume from an existing checkpoint, I'm faced with the error:
```
File "/opt/conda/lib/python3.10/site-packages/accelerate/optimizer.py", line 133, in step
self.scaler.step(self.optimizer, closure)
File "/opt/conda... | closed | completed | false | 3 | [
"bug"
] | [
"sayakpaul"
] | 2023-08-11T02:04:13Z | 2023-12-06T10:30:54Z | 2023-08-23T05:28:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xiankgx | 4,113,258 | MDQ6VXNlcjQxMTMyNTg= | User | false |
huggingface/diffusers | 1,846,121,719 | I_kwDOHa8MBc5uCZT3 | 4,567 | https://github.com/huggingface/diffusers/issues/4567 | https://api.github.com/repos/huggingface/diffusers/issues/4567 | Add Stable Diffusion XL JAX / Flax pipeline support | Seems like StableDiffusion XL will not work with FlaxStableDiffusionPipeline. It would be better to add SDXL Flax pipeline support.
Thanks! | closed | completed | false | 3 | [
"stale"
] | [] | 2023-08-11T02:47:48Z | 2023-10-13T10:37:18Z | 2023-10-13T10:37:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rupeshs | 3,255,994 | MDQ6VXNlcjMyNTU5OTQ= | User | false |
huggingface/diffusers | 1,846,122,647 | I_kwDOHa8MBc5uCZiX | 4,568 | https://github.com/huggingface/diffusers/issues/4568 | https://api.github.com/repos/huggingface/diffusers/issues/4568 | Temporal attention layer should consider position embedding of frames | ### Describe the bug
Hi, I found that in VideoLDM paper, the authors noted that
We use relative sinusoidal positional encodings for our temporal attention layers similar to those used to encode the timesteps in our U-Net backbone
However, in TransformerTemporalModel class, the previous implementation does ... | closed | completed | false | 1 | [
"bug",
"stale"
] | [] | 2023-08-11T02:49:31Z | 2023-10-18T15:13:51Z | 2023-10-18T15:13:51Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | SkyAndCloud | 12,583,017 | MDQ6VXNlcjEyNTgzMDE3 | User | false |
huggingface/diffusers | 1,368,349,336 | I_kwDOHa8MBc5Rj1qY | 457 | https://github.com/huggingface/diffusers/issues/457 | https://api.github.com/repos/huggingface/diffusers/issues/457 | Strange blue artifacts when running on MPS | ### Describe the bug
When running SD on mps, pictures appears with strange blue artifacts.
They appear on any count of inference steps. I tried on 50 and on 100. Blue noise everywhere
... | closed | completed | false | 3 | [
"bug"
] | [] | 2022-09-09T21:33:38Z | 2022-09-10T05:56:11Z | 2022-09-09T22:04:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | abesmon | 16,459,467 | MDQ6VXNlcjE2NDU5NDY3 | User | false |
huggingface/diffusers | 1,846,442,671 | I_kwDOHa8MBc5uDnqv | 4,573 | https://github.com/huggingface/diffusers/issues/4573 | https://api.github.com/repos/huggingface/diffusers/issues/4573 | Running the SD API, but the memory keeps increasing and does not automatically release? | Guys, I'm using this API to generate images, but as I continue to use it, I find that the available memory keeps decreasing and it does not free up automatically.
So, is there any way to resolve this issue? | closed | completed | false | 3 | [] | [] | 2023-08-11T08:46:54Z | 2023-08-16T03:24:51Z | 2023-08-16T03:24:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pengfei-lyu | 49,973,019 | MDQ6VXNlcjQ5OTczMDE5 | User | false |
huggingface/diffusers | 1,846,571,078 | I_kwDOHa8MBc5uEHBG | 4,575 | https://github.com/huggingface/diffusers/issues/4575 | https://api.github.com/repos/huggingface/diffusers/issues/4575 | Adding self-recurrence (time-travel) trick to schedulers | **Is your feature request related to a problem? Please describe.**
In image restoration problems like (image-inpainting, super-resolution, etc) it's common practice to use this trick to improve quality of generated images.
**Describe the solution you'd like**
Adding parameters in schedulers which allow to apply t... | closed | completed | false | 3 | [] | [] | 2023-08-11T10:19:17Z | 2023-08-16T14:30:57Z | 2023-08-16T14:30:57Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | anvilarth | 43,551,010 | MDQ6VXNlcjQzNTUxMDEw | User | false |
huggingface/diffusers | 1,846,701,496 | I_kwDOHa8MBc5uEm24 | 4,578 | https://github.com/huggingface/diffusers/issues/4578 | https://api.github.com/repos/huggingface/diffusers/issues/4578 | vae is not use in the doc | ### Describe the bug
https://huggingface.co/diffusers/controlnet-canny-sdxl-1.0?fbclid=IwAR3CTWQ3hLP1VxTpgFHVP82zhkYm7rnlKJ4HPq32rPAI9ltOWPbIcA2PFys
https://huggingface.co/docs/diffusers/main/en/api/pipelines/... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-08-11T11:59:50Z | 2023-08-11T17:43:41Z | 2023-08-11T17:43:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | elcolie | 18,206,728 | MDQ6VXNlcjE4MjA2NzI4 | User | false |
huggingface/diffusers | 1,368,427,170 | I_kwDOHa8MBc5RkIqi | 458 | https://github.com/huggingface/diffusers/issues/458 | https://api.github.com/repos/huggingface/diffusers/issues/458 | API: move pipeline's `output_type` to its Output object | Now that Pipelines return a well-defined PipelineOutput object, I suggest something along the lines of
```.py
class StableDiffusionPipelineOutput:
def images(self) -> List[PIL.Image.Image]: # […]
def as_array(self) -> np.ndarray: # […]
```
That retains the functionality of having PIL Image support built... | closed | completed | false | 3 | [
"stale"
] | [] | 2022-09-10T00:14:27Z | 2022-10-31T15:03:20Z | 2022-10-31T15:03:20Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | keturn | 83,819 | MDQ6VXNlcjgzODE5 | User | false |
huggingface/diffusers | 1,847,132,610 | I_kwDOHa8MBc5uGQHC | 4,580 | https://github.com/huggingface/diffusers/issues/4580 | https://api.github.com/repos/huggingface/diffusers/issues/4580 | unipc schedule throws exception when use_karras_sigmas=True | ### Describe the bug
unipc scheduler throws exception when use_karras_sigmas=True
File "/usr/local/src/diffusers/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py", line 642, in __call__
self.scheduler.set_timesteps(num_inference_steps, device=device)
File "/usr/local/src/diffusers/s... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-08-11T16:44:22Z | 2023-08-16T04:31:54Z | 2023-08-16T04:31:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | reimager | 20,918,094 | MDQ6VXNlcjIwOTE4MDk0 | User | false |
huggingface/diffusers | 1,847,343,821 | I_kwDOHa8MBc5uHDrN | 4,583 | https://github.com/huggingface/diffusers/issues/4583 | https://api.github.com/repos/huggingface/diffusers/issues/4583 | fp8 | Is it possible to utilize `fp8` of new nvidia ada lovelace GPUs? | closed | completed | false | 6 | [] | [] | 2023-08-11T19:50:21Z | 2023-09-02T08:22:09Z | 2023-08-14T12:18:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | maxpain | 5,875,582 | MDQ6VXNlcjU4NzU1ODI= | User | false |
huggingface/diffusers | 1,847,791,697 | I_kwDOHa8MBc5uIxBR | 4,584 | https://github.com/huggingface/diffusers/issues/4584 | https://api.github.com/repos/huggingface/diffusers/issues/4584 | train_text_to_image_lora_sdxl.py throws error on `--resume_from_checkpoint` | ### Describe the bug
When trying to resume training from checkpoint using `train_text_to_image_lora_sdxl.py` and `--resume_from_checkpoint=latest`, I'm getting error (logs below).
### Reproduction
Command to run:
```bash
accelerate launch diffusers/examples/text_to_image/train_text_to_image_lora_sdxl.py \
... | closed | completed | false | 5 | [
"bug"
] | [
"sayakpaul"
] | 2023-08-12T06:15:55Z | 2023-08-24T18:24:08Z | 2023-08-24T02:54:35Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ra100 | 3,185,410 | MDQ6VXNlcjMxODU0MTA= | User | false |
huggingface/diffusers | 1,848,295,341 | I_kwDOHa8MBc5uKr-t | 4,585 | https://github.com/huggingface/diffusers/issues/4585 | https://api.github.com/repos/huggingface/diffusers/issues/4585 | Inconsistency with Documentation: SDXL MultiControlNet Cannot Be Used | ### Describe the bug
In the documentation, there is an indication that MultiControlNet is supported, as seen in the following description:
"controlnet (ControlNetModel or List[ControlNetModel]) — Provides additional conditioning to the unet during the denoising process. If you set multiple ControlNets as a list, the ... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-12T23:27:14Z | 2023-08-14T14:54:09Z | 2023-08-14T12:14:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yamkz | 57,112,629 | MDQ6VXNlcjU3MTEyNjI5 | User | false |
huggingface/diffusers | 1,848,334,020 | I_kwDOHa8MBc5uK1bE | 4,587 | https://github.com/huggingface/diffusers/issues/4587 | https://api.github.com/repos/huggingface/diffusers/issues/4587 | StableDiffusionImg2ImgPipeline with low strength provides blurry result | ### Describe the bug
I have encountered a strange issue with the img2img pipeline recently
When I run the StableDiffusionImg2ImgPipeline with a low strength value, it will generate blurry images.
<img width="950" alt="image" src="https://github.com/huggingface/diffusers/assets/13037836/2c6ebba6-2dd0-4ae8-ba96-d2d2... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-08-13T01:11:20Z | 2023-08-14T02:44:51Z | 2023-08-14T02:44:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xxmtg | 13,037,836 | MDQ6VXNlcjEzMDM3ODM2 | User | false |
huggingface/diffusers | 1,848,353,979 | I_kwDOHa8MBc5uK6S7 | 4,588 | https://github.com/huggingface/diffusers/issues/4588 | https://api.github.com/repos/huggingface/diffusers/issues/4588 | Implement `CustomDiffusionAttnProcessor2_0` | Default AttnProcessor and LoRA can use torch 2.0 SDP operators, but the custom diffusion processors can not.
I can start a PR if you think it is good. | closed | completed | false | 6 | [
"stale"
] | [] | 2023-08-13T02:07:10Z | 2023-09-18T12:49:02Z | 2023-09-18T12:49:02Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eliphatfs | 23,738,781 | MDQ6VXNlcjIzNzM4Nzgx | User | false |
huggingface/diffusers | 1,848,362,041 | I_kwDOHa8MBc5uK8Q5 | 4,589 | https://github.com/huggingface/diffusers/issues/4589 | https://api.github.com/repos/huggingface/diffusers/issues/4589 | StableDiffusionControlNetImg2ImgPipeline (XL version) and StableDiffusionControlNetInpaintPipeline (XL version) | **Is your feature request related to a problem? Please describe.**
I want to implement it in my application, but it's very frustrating that I cannot use the combination of "SDXL and ControlNet and img2img" and "SDXL and ControlNet and inpaint" that I most want to use. I can do them individually, but.... I really desir... | closed | completed | false | 1 | [
"stale"
] | [] | 2023-08-13T02:30:24Z | 2023-09-14T14:21:50Z | 2023-09-14T14:21:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yamkz | 57,112,629 | MDQ6VXNlcjU3MTEyNjI5 | User | false |
huggingface/diffusers | 1,368,514,655 | I_kwDOHa8MBc5RkeBf | 459 | https://github.com/huggingface/diffusers/issues/459 | https://api.github.com/repos/huggingface/diffusers/issues/459 | Incremental Diffusion | **Is your feature request related to a problem? Please describe.**
An interesting feature for consumers of the `diffusers` API may be *incremental diffusion*, that is, the ability to parse diffusion results at every `n` steps during the pipeline run.
**Describe the solution you'd like**
One way of implementing t... | closed | completed | false | 4 | [
"enhancement"
] | [] | 2022-09-10T05:54:32Z | 2022-10-02T17:56:37Z | 2022-10-02T17:56:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | deepdiffuser | 112,834,445 | U_kgDOBrm3jQ | User | false |
huggingface/diffusers | 1,848,489,492 | I_kwDOHa8MBc5uLbYU | 4,591 | https://github.com/huggingface/diffusers/issues/4591 | https://api.github.com/repos/huggingface/diffusers/issues/4591 | Generator/Deterministic generation not working for Karlo/unCLIP | ### Describe the bug
Regardless of sending a generator or trying to set a torch seed in the environment, each unCLIP generation makes a different image.
### Reproduction
Setup
```py
import torch
from diffusers import UnCLIPPipeline
pipe = UnCLIPPipeline.from_pretrained("kakaobrain/karlo-v1-alpha", torch_... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-13T08:43:46Z | 2023-08-13T09:40:48Z | 2023-08-13T09:40:47Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | apolinario | 788,417 | MDQ6VXNlcjc4ODQxNw== | User | false |
huggingface/diffusers | 1,848,539,887 | I_kwDOHa8MBc5uLnrv | 4,593 | https://github.com/huggingface/diffusers/issues/4593 | https://api.github.com/repos/huggingface/diffusers/issues/4593 | Unable to load Automatic1111 LoRa due rank being higher than "min(in_features, out_features)}" | ### Describe the bug
Hi! I'm unable to load a LoRa as I get the following error.
`ValueError: LoRA rank 384 must be less or equal than 320`
The LoRa works perfectly if the ValueError check is removed from the code. Is it possible to override this error? I don't understand why it must be less or equal. Thanks!
... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-08-13T10:57:30Z | 2023-08-17T04:37:09Z | 2023-08-17T04:37:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sinanhlvc | 66,330,822 | MDQ6VXNlcjY2MzMwODIy | User | false |
huggingface/diffusers | 1,848,628,473 | I_kwDOHa8MBc5uL9T5 | 4,594 | https://github.com/huggingface/diffusers/issues/4594 | https://api.github.com/repos/huggingface/diffusers/issues/4594 | latents.requires_grad is false in my custom pipeline no matter what. | Hi, in my quest to make a flexible pipeline that can easily add new features instead of creating a pipeline for every variation, I made the following:
```
class StableDiffusionRubberPipeline(StableDiffusionPipeline):
call_funcs=[]
def __init__(
self,
vae: AutoencoderKL,
text_enc... | closed | completed | false | 1 | [] | [] | 2023-08-13T15:02:22Z | 2023-08-14T12:11:36Z | 2023-08-14T12:11:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,849,655,321 | I_kwDOHa8MBc5uP4AZ | 4,598 | https://github.com/huggingface/diffusers/issues/4598 | https://api.github.com/repos/huggingface/diffusers/issues/4598 | DiffusionPipeline does not give informative error messages when a dependency is missing | ### Describe the bug
If I use `StableDiffusionXLPipeline`, I get a good error message if `transformers` is missing. If I use `DiffusionPipeline` instead, the error is not transparent.
### Reproduction
Colab https://colab.research.google.com/drive/1fF7r-FLg729PbQNAn03179vqWDTkDrTP?usp=sharing
```python
fro... | closed | completed | false | 3 | [
"bug"
] | [
"DN6"
] | 2023-08-14T12:11:13Z | 2023-08-22T19:27:43Z | 2023-08-22T19:27:43Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | osanseviero | 7,246,357 | MDQ6VXNlcjcyNDYzNTc= | User | false |
huggingface/diffusers | 1,368,538,273 | I_kwDOHa8MBc5Rkjyh | 460 | https://github.com/huggingface/diffusers/issues/460 | https://api.github.com/repos/huggingface/diffusers/issues/460 | Stable Diffusion ONNX requires cpu `onnxruntime` even if CUDA version is installed | ### Describe the bug
The ONNX support doesn't work with `CUDAExecutionProvider` I installed `onnxruntime-gpu`
Running
`import onnxruntime as ort
ort.get_device()`
results
`GPU`
and
`ort.get_available_providers()`
results
`['CPUExecutionProvider', 'TensorrtExecutionProvider', 'CUDAExecutionProvider']`
but d... | closed | completed | false | 8 | [] | [
"anton-l"
] | 2022-09-10T07:34:20Z | 2022-10-14T17:30:00Z | 2022-10-12T10:20:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Vijayabhaskar96 | 17,958,636 | MDQ6VXNlcjE3OTU4NjM2 | User | false |
huggingface/diffusers | 1,850,133,058 | I_kwDOHa8MBc5uRspC | 4,601 | https://github.com/huggingface/diffusers/issues/4601 | https://api.github.com/repos/huggingface/diffusers/issues/4601 | Controlnet-SD-XL-v1-NormalBae/Tile/MLSD checkpoint | Currently, Only Canny controlnet model is released for SD-XL-1.0, I want to know, When You want to release other conternet checkpoints for SD-XL-V1?
| closed | completed | false | 1 | [] | [] | 2023-08-14T16:27:49Z | 2023-08-16T09:33:35Z | 2023-08-16T09:33:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mzeynali | 95,717,811 | U_kgDOBbSJsw | User | false |
huggingface/diffusers | 1,850,136,134 | I_kwDOHa8MBc5uRtZG | 4,602 | https://github.com/huggingface/diffusers/issues/4602 | https://api.github.com/repos/huggingface/diffusers/issues/4602 | Controlnet SD-XL-v1-refiner+base | Currently, I tested controlnet SD-XL-v1-base model, and it's great, but I want to know when you want to release controlnet SD-XL-v1-base+refinner codes? | closed | completed | false | 3 | [] | [] | 2023-08-14T16:29:59Z | 2023-08-16T09:34:07Z | 2023-08-16T09:34:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mzeynali | 95,717,811 | U_kgDOBbSJsw | User | false |
huggingface/diffusers | 1,850,852,504 | I_kwDOHa8MBc5uUcSY | 4,605 | https://github.com/huggingface/diffusers/issues/4605 | https://api.github.com/repos/huggingface/diffusers/issues/4605 | StableDiffusionLongPromptWeightingPipeline with inpaint model report error | ### Describe the bug
I use StableDiffusionInpaintPipeline and dreamshaper v8 inpainting model to do inpainting, it works well. Then I want to make more prompts, so I added the StableDiffusionLongPromptWeightingPipeline. But it reports following error:
File "/data/images/codes/models/sd/lpw_stable_diffusion.py", ... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-08-15T03:26:43Z | 2023-11-08T15:09:26Z | 2023-11-08T15:09:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | junnikokuki | 17,866,964 | MDQ6VXNlcjE3ODY2OTY0 | User | false |
huggingface/diffusers | 1,850,853,705 | I_kwDOHa8MBc5uUclJ | 4,606 | https://github.com/huggingface/diffusers/issues/4606 | https://api.github.com/repos/huggingface/diffusers/issues/4606 | Training a diffusion prior model | Hi, is there something in this repo to allow for training a diffusion prior model, for example from CLIP image embeddings to CLIP text encodings (batch_size, 77)? | closed | completed | false | 1 | [] | [] | 2023-08-15T03:28:47Z | 2023-08-16T09:26:07Z | 2023-08-16T09:26:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xiankgx | 4,113,258 | MDQ6VXNlcjQxMTMyNTg= | User | false |
huggingface/diffusers | 1,850,864,439 | I_kwDOHa8MBc5uUfM3 | 4,609 | https://github.com/huggingface/diffusers/issues/4609 | https://api.github.com/repos/huggingface/diffusers/issues/4609 | Provides t2iadapter_style_sd14v1 model | ### Model/Pipeline/Scheduler description
See https://huggingface.co/TencentARC/T2I-Adapter/blob/main/models/t2iadapter_style_sd14v1.pth
### Open source status
- [X] The model implementation is available
- [X] The model weights are available (Only relevant if addition is not a scheduler).
### Provide useful links fo... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-08-15T03:46:18Z | 2023-11-09T15:06:33Z | 2023-11-09T15:06:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | debugger-zhang | 70,719,739 | MDQ6VXNlcjcwNzE5NzM5 | User | false |
huggingface/diffusers | 1,850,924,187 | I_kwDOHa8MBc5uUtyb | 4,611 | https://github.com/huggingface/diffusers/issues/4611 | https://api.github.com/repos/huggingface/diffusers/issues/4611 | VideoComposer merged with Modelscope | ### Model/Pipeline/Scheduler description
This is very good news, since VideoComposer comes with multiple new features for ex. img2vid and vid2vid. With this merge, maybe open source txt2vid can be shoulder to shoulder with the closed source solutions.
Examples:
https://videocomposer.github.io/
![image](https... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-08-15T05:32:22Z | 2023-11-09T15:06:31Z | 2023-11-09T15:06:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,850,976,564 | I_kwDOHa8MBc5uU6k0 | 4,613 | https://github.com/huggingface/diffusers/issues/4613 | https://api.github.com/repos/huggingface/diffusers/issues/4613 | Please support “class” feature in unet_1d.py | Hello,
I'm writing a project which need class feature.
Could you add `class_embed_type: Optional[str] = None, num_class_embeds: Optional[int] = None` into UNet1DModel?
Thank you so much! | closed | completed | false | 1 | [] | [] | 2023-08-15T06:42:21Z | 2023-08-16T09:20:25Z | 2023-08-16T09:20:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | YChienHung | 41,673,864 | MDQ6VXNlcjQxNjczODY0 | User | false |
huggingface/diffusers | 1,851,199,064 | I_kwDOHa8MBc5uVw5Y | 4,615 | https://github.com/huggingface/diffusers/issues/4615 | https://api.github.com/repos/huggingface/diffusers/issues/4615 | ImportError: cannot import name 'StableDiffusionControlnetPipeline' from 'diffusers' | ### Describe the bug
Traceback (most recent call last):
File "../canny/test.py", line 27, in <module>
from diffusers import StableDiffusionControlnetPipeline, ControlNetModel
ImportError: cannot import name 'StableDiffusionControlnetPipeline' from 'diffusers' (/opt/conda/lib/python3.8/site-packages/diffusers/... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-08-15T10:07:01Z | 2023-08-15T10:10:00Z | 2023-08-15T10:10:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kkwhale7 | 79,788,571 | MDQ6VXNlcjc5Nzg4NTcx | User | false |
huggingface/diffusers | 1,851,219,574 | I_kwDOHa8MBc5uV152 | 4,616 | https://github.com/huggingface/diffusers/issues/4616 | https://api.github.com/repos/huggingface/diffusers/issues/4616 | convert_lora_safetensor_to_diffusers.py for SDXL not working | ### Describe the bug
```
Downloading (…)ch_model.safetensors: 100% 10.3G/10.3G [02:35<00:00, 66.2MB/s]
Fetching 19 files: 100% 19/19 [02:36<00:00, 8.23s/it]
Loading pipeline components...: 100% 5/5 [00:04<00:00, 1.01it/s]
You have disabled the safety checker for <class 'diffusers.pipelines.stable_diffusion.pip... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-08-15T10:26:15Z | 2023-08-16T09:19:33Z | 2023-08-16T09:19:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,851,468,705 | I_kwDOHa8MBc5uWyuh | 4,618 | https://github.com/huggingface/diffusers/issues/4618 | https://api.github.com/repos/huggingface/diffusers/issues/4618 | How to use dreamshaperXL10_alpha2Xl10.safetensors with controlnet-canny-sdxl-1.0 ? | I want to use dreamshaperXL10_alpha2Xl10.safetensors with controlnet-canny-sdxl-1.0
I downloaded dreamshaperXL10_alpha2Xl10.safetensors file and tried to use :
pipe = StableDiffusionXLControlNetPipeline.from_pretrained(
'./dreamshaperXL10_alpha2Xl10.safetensors',
controlnet=controlnet,
use_safetensors=True,
to... | closed | completed | false | 13 | [] | [] | 2023-08-15T13:44:54Z | 2023-08-22T01:31:37Z | 2023-08-22T01:31:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | arnold408 | 1,244,950 | MDQ6VXNlcjEyNDQ5NTA= | User | false |
huggingface/diffusers | 1,851,813,283 | I_kwDOHa8MBc5uYG2j | 4,619 | https://github.com/huggingface/diffusers/issues/4619 | https://api.github.com/repos/huggingface/diffusers/issues/4619 | RuntimeError: Input type (c10::Half) and bias type (float) mismatch in training_text_to_image_lora_sdxl.py | ### Describe the bug
I'm encountering the same error as described in the closed issue #4478.
I'm currently running the train_text_to_image_lora_sdxl.py script, and the VAE give me the following error:
```bash
RuntimeError: Input type (c10::Half) and bias type (float) should be the same
```
See "Reproduction",... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-15T17:07:42Z | 2023-08-23T05:28:56Z | 2023-08-23T05:28:55Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mnslarcher | 24,874,992 | MDQ6VXNlcjI0ODc0OTky | User | false |
huggingface/diffusers | 1,368,561,245 | I_kwDOHa8MBc5RkpZd | 462 | https://github.com/huggingface/diffusers/issues/462 | https://api.github.com/repos/huggingface/diffusers/issues/462 | StableDiffusionImg2ImgPipeline broken in 0.3.0 release for Apple MPS | ### Describe the bug
Using StableDiffusionImg2ImgPipeline before the 0.3.0 release used to generate images correctly on Apple Silicon devices. Now, if you use StableDiffusionImg2ImgPipeline, all you get is a brown image.
I've copied and pasted the sample code from the docs (with a few minor modifications) and tried... | closed | completed | false | 2 | [
"bug"
] | [] | 2022-09-10T09:15:39Z | 2022-09-11T12:04:33Z | 2022-09-11T12:04:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FahimF | 181,110 | MDQ6VXNlcjE4MTExMA== | User | false |
huggingface/diffusers | 1,852,084,318 | I_kwDOHa8MBc5uZJBe | 4,624 | https://github.com/huggingface/diffusers/issues/4624 | https://api.github.com/repos/huggingface/diffusers/issues/4624 | Failure to load SDXL checkpoint on low ram (16gb) system. | ### Describe the bug
I've written what little I know in vlad's repo: https://github.com/vladmandic/automatic/issues/2015
Basically, convert_from_ckpt.download_from_original_stable_diffusion_ckpt when accelerate is enabled loads the entire model to ram, which is right about the edge of the system's capacity. Then,... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-08-15T20:29:59Z | 2023-08-24T12:45:26Z | 2023-08-24T12:43:58Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Symbiomatrix | 41,131,377 | MDQ6VXNlcjQxMTMxMzc3 | User | false |
huggingface/diffusers | 1,368,697,627 | I_kwDOHa8MBc5RlKsb | 463 | https://github.com/huggingface/diffusers/issues/463 | https://api.github.com/repos/huggingface/diffusers/issues/463 | Super Resolution Diffusion Model | Hello.
I am very interested in the unconditional image generation pipelines. Like in this example here:
https://github.com/huggingface/diffusers/tree/main/examples/unconditional_image_generation
I have trained a network that is 128x128 and it gives very good results for what I need. However the resolution is ver... | closed | completed | false | 26 | [
"enhancement",
"New pipeline/model"
] | [] | 2022-09-10T18:24:42Z | 2024-11-25T17:20:29Z | 2024-11-25T17:20:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dvIdol | 113,259,332 | U_kgDOBsAzRA | User | false |
huggingface/diffusers | 1,852,582,603 | I_kwDOHa8MBc5ubCrL | 4,630 | https://github.com/huggingface/diffusers/issues/4630 | https://api.github.com/repos/huggingface/diffusers/issues/4630 | Modify lora weight after loading them | **Is your feature request related to a problem? Please describe.**
In order to be able to put more focus on some loras rather than others under specific conditions is necessary when implementing stuff like twoshot. I have my own version of this, but it is already not up to date.
**Describe the solution you'd like**... | closed | completed | false | 10 | [
"stale"
] | [] | 2023-08-16T06:16:39Z | 2023-11-09T15:06:29Z | 2023-11-09T15:06:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,852,722,189 | I_kwDOHa8MBc5ubkwN | 4,631 | https://github.com/huggingface/diffusers/issues/4631 | https://api.github.com/repos/huggingface/diffusers/issues/4631 | DPMSolverSinglestepScheduler with use_karras_sigmas=True triggers RuntimeError: a Tensor with 2 elements cannot be converted to Scalar | ### Describe the bug
Hello @patrickvonplaten and @williamberman 🤗
Running `StableDiffusionPipeline` with `DPMSolverSinglestepScheduler` and `use_karras_sigmas=True` triggers the error `RuntimeError: a Tensor with 2 elements cannot be converted to Scalar ` when generating images with a high `num_inference_steps` ... | closed | completed | false | 11 | [
"bug"
] | [
"yiyixuxu"
] | 2023-08-16T08:03:46Z | 2023-12-04T19:47:15Z | 2023-09-20T06:13:06Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexisrolland | 13,064,696 | MDQ6VXNlcjEzMDY0Njk2 | User | false |
huggingface/diffusers | 1,853,575,916 | I_kwDOHa8MBc5ue1Ls | 4,634 | https://github.com/huggingface/diffusers/issues/4634 | https://api.github.com/repos/huggingface/diffusers/issues/4634 | Safe Serialization does not work for Custom Diffusion | ### Describe the bug
Ran into this issue when working on: #4235
Diffusers now defaults to saving models via `safetensors`.
In certain cases:
https://github.com/huggingface/diffusers/blob/main/examples/custom_diffusion/train_custom_diffusion.py
saving Custom Unet Attention Processors via `safetensors` is no... | closed | completed | false | 5 | [
"bug",
"stale"
] | [
"DN6",
"sayakpaul",
"patrickvonplaten"
] | 2023-08-16T16:35:39Z | 2023-11-09T15:06:27Z | 2023-11-09T15:06:27Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DN6 | 7,529,846 | MDQ6VXNlcjc1Mjk4NDY= | User | false |
huggingface/diffusers | 1,853,937,402 | I_kwDOHa8MBc5ugNb6 | 4,636 | https://github.com/huggingface/diffusers/issues/4636 | https://api.github.com/repos/huggingface/diffusers/issues/4636 | Training on zero terminal SNR seems to be broken | ### Describe the bug
I've used the Diffusers pix2pix script, adapted for training the SDXL U-net with a focus on text2img to do experiments over the last week.
I've observed a continual contrast issue with SDXL on v_prediction.
Prompt: `white woman in white pants and white shirt standing in front of white backgr... | closed | completed | false | 28 | [
"bug"
] | [] | 2023-08-16T21:36:32Z | 2023-08-18T19:32:37Z | 2023-08-18T14:36:07Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 1,853,967,669 | I_kwDOHa8MBc5ugU01 | 4,637 | https://github.com/huggingface/diffusers/issues/4637 | https://api.github.com/repos/huggingface/diffusers/issues/4637 | torch._dynamo.exc.BackendCompilerFailed while run train_instruct_pix2pix_sdxl.py | ### Describe the bug
# Short Description
Train Instruct Pix2Pix SDXL failed with torch._dynamo.exc.BackendCompilerFailed error.
Which says,
torch._dynamo.exc.BackendCompilerFailed: debug_wrapper raised SyntaxError: invalid syntax. Perhaps you forgot a comma? (<eval_with_key>.11, line 16663)
Any advice will be ... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-08-16T22:09:46Z | 2023-11-09T15:06:25Z | 2023-11-09T15:06:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | frankjiang | 5,550,580 | MDQ6VXNlcjU1NTA1ODA= | User | false |
huggingface/diffusers | 1,853,977,739 | I_kwDOHa8MBc5ugXSL | 4,638 | https://github.com/huggingface/diffusers/issues/4638 | https://api.github.com/repos/huggingface/diffusers/issues/4638 | load_textual_inversion doesn't work on SDXL | ### Describe the bug
According to the 0.19.3 documentation at https://huggingface.co/docs/diffusers/v0.19.3/en/api/pipelines/stable_diffusion/stable_diffusion_xl#diffusers.StableDiffusionXLPipeline , the StableDiffusionXLPipeline supports `load_textual_inversion()`:
> In addition the pipeline inherits the following... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-08-16T22:22:55Z | 2023-09-26T20:48:13Z | 2023-08-17T04:56:29Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JeLuF | 5,852,422 | MDQ6VXNlcjU4NTI0MjI= | User | false |
huggingface/diffusers | 1,853,979,675 | I_kwDOHa8MBc5ugXwb | 4,639 | https://github.com/huggingface/diffusers/issues/4639 | https://api.github.com/repos/huggingface/diffusers/issues/4639 | mutliprocessing error from data loader in train_instruct_pix2pix_sdxl.py | ### Describe the bug
# Short Description
Run the training with ```train_instruct_pix2pix_sdxl.py```
It fails on --dataloader_num_workers
```RuntimeError: Cannot re-initialize CUDA in forked subprocess. To use CUDA with multiprocessing, you must use the 'spawn' start method```
# What have I tried
I found:
1. Se... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-08-16T22:25:34Z | 2023-11-09T15:06:23Z | 2023-11-09T15:06:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | frankjiang | 5,550,580 | MDQ6VXNlcjU1NTA1ODA= | User | false |
huggingface/diffusers | 1,854,078,812 | I_kwDOHa8MBc5ugv9c | 4,641 | https://github.com/huggingface/diffusers/issues/4641 | https://api.github.com/repos/huggingface/diffusers/issues/4641 | Native diffusers LoRA weights not working with `load_lora_weights` on a local file | ### Describe the bug
Doing a `pipe.load_lora_weights("path/to/pytorch_lora_weights.bin")` trained natively with diffusers and in diffusers format, but downloaded locally is failing.
- Doing `pipe.load_lora_weights("path/to/random_community_lora.safetensors")` works
- Doing `pipe.load_lora_weights("hub_user/hub_re... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-08-17T00:57:56Z | 2023-08-17T04:03:50Z | 2023-08-17T04:03:49Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | apolinario | 788,417 | MDQ6VXNlcjc4ODQxNw== | User | false |
huggingface/diffusers | 1,854,079,903 | I_kwDOHa8MBc5ugwOf | 4,642 | https://github.com/huggingface/diffusers/issues/4642 | https://api.github.com/repos/huggingface/diffusers/issues/4642 | LoRA weights baking. | **Is your feature request related to a problem? Please describe.**
LoRA is consuming extra GPU power that is not required in inference. This can cause a noticable amount of overhead especially with high resolutions. For higher LoRA ranks (64/128), this can be up to 30%-50% performance gain.
**Describe the solution ... | closed | completed | false | 2 | [] | [] | 2023-08-17T00:59:45Z | 2023-08-17T01:07:04Z | 2023-08-17T01:07:04Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eliphatfs | 23,738,781 | MDQ6VXNlcjIzNzM4Nzgx | User | false |
huggingface/diffusers | 1,854,154,642 | I_kwDOHa8MBc5uhCeS | 4,643 | https://github.com/huggingface/diffusers/issues/4643 | https://api.github.com/repos/huggingface/diffusers/issues/4643 | when i load a controlnet model,where is the inference code? | I have read the code of con in diffusers/models/controlnet.py.
but when I load a con weight,where is the code?
tks | closed | completed | false | 1 | [] | [] | 2023-08-17T02:50:59Z | 2023-08-17T04:55:28Z | 2023-08-17T04:55:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | henbucuoshanghai | 37,647,094 | MDQ6VXNlcjM3NjQ3MDk0 | User | false |
huggingface/diffusers | 1,854,289,320 | I_kwDOHa8MBc5uhjWo | 4,645 | https://github.com/huggingface/diffusers/issues/4645 | https://api.github.com/repos/huggingface/diffusers/issues/4645 | Implementing Fooocus | Authors of controlnet have released Fooocus, which is essentially tricks to get the best quality images,
https://github.com/lllyasviel/Fooocus
List of hidden tricks
https://github.com/lllyasviel/Fooocus#list-of-hidden-tricks
How can we implement those hidden tricks on diffusers? | closed | completed | false | 5 | [
"stale"
] | [] | 2023-08-17T05:41:52Z | 2024-01-09T01:45:12Z | 2023-10-30T15:08:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | adhikjoshi | 11,740,719 | MDQ6VXNlcjExNzQwNzE5 | User | false |
huggingface/diffusers | 1,854,336,287 | I_kwDOHa8MBc5uhu0f | 4,646 | https://github.com/huggingface/diffusers/issues/4646 | https://api.github.com/repos/huggingface/diffusers/issues/4646 | convert controlnet to onnx failed | ### Describe the bug
when i run convert_stable_diffusion_controlnet_to_onnx.py，while runing convert unet to onnx，have a bug
============= Diagnostic Run torch.onnx.export version 2.0.1+cu118 =============
verbose: False, log level: Level.ERROR
======================= 0 NONE 0 NOTE 0 WARNING 0 ERROR ==============... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2023-08-17T06:26:43Z | 2024-09-27T17:52:48Z | 2023-10-30T14:26:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yuxiaoranyu | 53,110,243 | MDQ6VXNlcjUzMTEwMjQz | User | false |
huggingface/diffusers | 1,854,414,672 | I_kwDOHa8MBc5uiB9Q | 4,647 | https://github.com/huggingface/diffusers/issues/4647 | https://api.github.com/repos/huggingface/diffusers/issues/4647 | UNet2DModel doesn't pass "num_layers" into its mid_block | ### Describe the bug
UNet2DModel doesn't specify the parameter "num_layers" of its mid_block, casuing its midblock to always has 1 ResBlock despite of the "layers_per_block". As is shown in the link below:
https://github.com/huggingface/diffusers/blob/7c3e7fedcdfb702a095861c418964e9d1e2c958a/src/diffusers/models/un... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-17T07:18:58Z | 2023-08-17T07:29:04Z | 2023-08-17T07:29:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | godfather991 | 58,773,907 | MDQ6VXNlcjU4NzczOTA3 | User | false |
huggingface/diffusers | 1,854,501,352 | I_kwDOHa8MBc5uiXHo | 4,648 | https://github.com/huggingface/diffusers/issues/4648 | https://api.github.com/repos/huggingface/diffusers/issues/4648 | raise EnvironmentError( OSError: Error no file named config.json found in directory /workspace/canny/ControlNet-v1-1. | ### Describe the bug
`from diffusers import StableDiffusionControlNetPipeline, ControlNetModel
url="/workspace/canny/ControlNet-v1-1"
controlnet = ControlNetModel.from_pretrained(url)`
is my path false ? there is not config,json in the file
### Reproduction
1
### Logs
_No response_
### System Info
1
### Who ... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2023-08-17T08:16:40Z | 2024-11-13T00:13:44Z | 2023-10-30T15:08:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kkwhale7 | 79,788,571 | MDQ6VXNlcjc5Nzg4NTcx | User | false |
huggingface/diffusers | 1,854,521,314 | I_kwDOHa8MBc5uib_i | 4,649 | https://github.com/huggingface/diffusers/issues/4649 | https://api.github.com/repos/huggingface/diffusers/issues/4649 | if "text_embeds" not in added_cond_kwargs: TypeError: argument of type 'NoneType' is not iterable | ### Describe the bug
```
if "text_embeds" not in added_cond_kwargs:
TypeError: argument of type 'NoneType' is not iterable
```
I think I'm getting this error if I feed an image or a video which differs in resolution with the resolution a model was trained on, but I'm not sure exactly what is causing it. Maybe this... | closed | completed | false | 18 | [
"bug"
] | [] | 2023-08-17T08:30:06Z | 2024-08-16T20:58:37Z | 2023-08-18T06:41:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,368,743,932 | I_kwDOHa8MBc5RlV_8 | 465 | https://github.com/huggingface/diffusers/issues/465 | https://api.github.com/repos/huggingface/diffusers/issues/465 | Discrepancy in scheduler `offset` usage | **What API design would you like to have changed or added to the library? Why?**
@jonatanklosko brought up an issue with some confusing code (why are `alphas` indexed differently across `PNDMScheduler` and `DDIMScheduler`), and some inconsistency in the offset usage.
Two questions:
1. Why does only one shift by ... | closed | completed | false | 13 | [] | [] | 2022-09-10T22:21:32Z | 2022-09-17T12:07:44Z | 2022-09-17T12:07:44Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | natolambert | 10,695,622 | MDQ6VXNlcjEwNjk1NjIy | User | false |
huggingface/diffusers | 1,854,558,167 | I_kwDOHa8MBc5uik_X | 4,651 | https://github.com/huggingface/diffusers/issues/4651 | https://api.github.com/repos/huggingface/diffusers/issues/4651 | Key-Locked Rank One Editing for Text-to-Image Personalization | https://arxiv.org/abs/2305.01644
This new method can enhance the ability of personalized t2i, even when user only provide only one image of their subject.
| closed | completed | false | 3 | [
"community-examples",
"wip",
"contributions-welcome"
] | [] | 2023-08-17T08:53:46Z | 2024-11-21T15:42:14Z | 2024-11-21T15:42:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | blx0102 | 17,761,044 | MDQ6VXNlcjE3NzYxMDQ0 | User | false |
huggingface/diffusers | 1,854,716,523 | I_kwDOHa8MBc5ujLpr | 4,654 | https://github.com/huggingface/diffusers/issues/4654 | https://api.github.com/repos/huggingface/diffusers/issues/4654 | Different results after DDIM inverse. | `
from diffusers import DDIMScheduler, UNet2DModel, DDIMInverseScheduler
import torch
import numpy as np
import cv2
from tqdm import tqdm
def ddpm_forward(noise, model, scheduler, timesteps=500):
scheduler.set_timesteps(timesteps)
... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-08-17T10:23:33Z | 2023-10-30T15:08:42Z | 2023-10-30T15:08:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Sumching | 34,994,554 | MDQ6VXNlcjM0OTk0NTU0 | User | false |
huggingface/diffusers | 1,855,274,971 | I_kwDOHa8MBc5ulT_b | 4,657 | https://github.com/huggingface/diffusers/issues/4657 | https://api.github.com/repos/huggingface/diffusers/issues/4657 | StableDiffusionXLControlNetPipeline is missing denoising_end | StableDiffusionXLPipeline has a denoising_end parameter for the Ensemble of Expert Denoisers approach that is missing when using ControlNet.
As a future step, both denoising_start/end could be useful across all the pipelines (not just SDXL). E.g. run several images partially through latent space, combine the result... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-08-17T15:53:33Z | 2023-10-30T15:08:39Z | 2023-10-30T15:08:39Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pdoane | 1,621,950 | MDQ6VXNlcjE2MjE5NTA= | User | false |
huggingface/diffusers | 1,855,339,537 | I_kwDOHa8MBc5uljwR | 4,658 | https://github.com/huggingface/diffusers/issues/4658 | https://api.github.com/repos/huggingface/diffusers/issues/4658 | Should AudioLDM pipeline use separate unet class | @patrickvonplaten see comment here https://github.com/huggingface/diffusers/pull/4549/files#r1297472583 and the rationale in the corresponding docstring. Would be great to get you to sanity check. Thanks! | closed | completed | false | 3 | [
"stale"
] | [] | 2023-08-17T16:36:59Z | 2023-10-30T15:08:37Z | 2023-10-30T15:08:37Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,855,402,835 | I_kwDOHa8MBc5ulzNT | 4,660 | https://github.com/huggingface/diffusers/issues/4660 | https://api.github.com/repos/huggingface/diffusers/issues/4660 | Add SDXL ControlNet pipelines for img2img and inpaint | Base SD1.5 has StableDiffusionControlNetImg2ImgPipeline and StableDiffusionControlNetInpaintPipeline but there is no corresponding versions for SDXL ControlNet. | closed | completed | false | 2 | [] | [] | 2023-08-17T17:20:24Z | 2023-08-20T02:03:42Z | 2023-08-17T17:21:26Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pdoane | 1,621,950 | MDQ6VXNlcjE2MjE5NTA= | User | false |
huggingface/diffusers | 1,855,744,936 | I_kwDOHa8MBc5unGuo | 4,662 | https://github.com/huggingface/diffusers/issues/4662 | https://api.github.com/repos/huggingface/diffusers/issues/4662 | How to call a different scheduler when training a model from repo | I notice that the settings in train_dreambooth_lora_sdxl.py and the scheduler config from the repo seem to conflict. In the .py the noise scheduler is DDPM but whenever training starts it seems to still indicate that I am using the repo config scheduler, ie. EulerDiscreteScheduler. It used to be you could specify sched... | closed | completed | false | 1 | [] | [] | 2023-08-17T21:40:10Z | 2023-08-18T04:18:11Z | 2023-08-18T04:18:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jmaccall316 | 120,006,038 | U_kgDOBycllg | User | false |
huggingface/diffusers | 1,855,777,501 | I_kwDOHa8MBc5unOrd | 4,663 | https://github.com/huggingface/diffusers/issues/4663 | https://api.github.com/repos/huggingface/diffusers/issues/4663 | Loading StableDiffusionXLControlNetPipeline from single file |
Hi guys, is there a way to load SDXLControlNet from just on single safetensors file?
| closed | completed | false | 3 | [
"stale"
] | [] | 2023-08-17T22:19:09Z | 2023-11-01T20:13:21Z | 2023-10-30T15:08:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiedong-yang | 54,411,100 | MDQ6VXNlcjU0NDExMTAw | User | false |
huggingface/diffusers | 1,855,883,202 | I_kwDOHa8MBc5unofC | 4,664 | https://github.com/huggingface/diffusers/issues/4664 | https://api.github.com/repos/huggingface/diffusers/issues/4664 | Lora loading for SDXL works on StableDiffusionXLPipeline but not on StablediffusionXLPipeline | ### 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 controlnet, 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... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-08-18T00:47:49Z | 2023-08-23T09:13:26Z | 2023-08-22T01:43:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ilisparrow | 4,880,273 | MDQ6VXNlcjQ4ODAyNzM= | User | false |
huggingface/diffusers | 1,855,978,144 | I_kwDOHa8MBc5un_qg | 4,665 | https://github.com/huggingface/diffusers/issues/4665 | https://api.github.com/repos/huggingface/diffusers/issues/4665 | use_safetensors Fails on Windows When There Are Alternative bin Files | ### Describe the bug
When a model contains safetensors and bin weights diffusers will only download the bin weights. An error is raised when `use_safetensors=True`:
```
OSError: Could not found the necessary `safetensors` weights in {'vae/diffusion_pytorch_model.fp16.bin', 'text_encoder/pytorch_model.fp16.bin', 'une... | closed | completed | false | 3 | [
"bug"
] | [
"DN6"
] | 2023-08-18T03:17:52Z | 2023-11-08T10:51:16Z | 2023-11-08T10:51:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | NullSenseStudio | 47,096,043 | MDQ6VXNlcjQ3MDk2MDQz | User | false |
huggingface/diffusers | 1,856,179,572 | I_kwDOHa8MBc5uow10 | 4,667 | https://github.com/huggingface/diffusers/issues/4667 | https://api.github.com/repos/huggingface/diffusers/issues/4667 | In version 0.20.0, StableDiffusionPipeline.from_single_file does not work. | It worked fine in diffusers==0.19.3, but after updating to version 0.20.0, I got the following error.
### Logs
~~~
TypeError: StableDiffusionPipeline.__init__() got an unexpected keyword argument 'text_encoder_2'
~~~
### Reproduction
~~~python
from diffusers import StableDiffusionPipeline
import torch
pipe =... | closed | completed | false | 1 | [] | [] | 2023-08-18T07:05:49Z | 2023-08-21T03:44:54Z | 2023-08-21T03:44:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dai-ichiro | 53,703,225 | MDQ6VXNlcjUzNzAzMjI1 | User | false |
huggingface/diffusers | 1,856,330,392 | I_kwDOHa8MBc5upVqY | 4,668 | https://github.com/huggingface/diffusers/issues/4668 | https://api.github.com/repos/huggingface/diffusers/issues/4668 | Issue with Kohya-style LoRA and SD1.5 based model: Weird Images | ### Describe the bug
Weird images are being generated with Kohya-style LoRA and SD1.5 based model (Deliberate 2.0). It works well with A1111. I also tried to run it with kohya_lora_loader (https://gist.github.com/takuma104/e38d683d72b1e448b8d9b3835f7cfa44) and the results are much better.
![download](https://github... | closed | completed | false | 10 | [
"bug",
"stale"
] | [] | 2023-08-18T08:55:32Z | 2023-10-30T15:08:33Z | 2023-10-30T15:08:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | antonioglass | 71,437,038 | MDQ6VXNlcjcxNDM3MDM4 | User | false |
huggingface/diffusers | 1,856,524,802 | I_kwDOHa8MBc5uqFIC | 4,670 | https://github.com/huggingface/diffusers/issues/4670 | https://api.github.com/repos/huggingface/diffusers/issues/4670 | why canny controlnet generation so elusive? | ### Describe the bug
i have tried with some different canny edge detection algorithm like cv2 edges = cv.Canny(img,100,200) and this git clone https://github.com/FienSoP/canny_edge_detector
it looks like fine with canny image:
![car_gist_gray](https://github.com/huggingface/diffusers/assets/130369523/4b9916da-dd2c-4... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-08-18T11:00:55Z | 2023-10-30T15:08:31Z | 2023-10-30T15:08:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | frankchieng | 130,369,523 | U_kgDOB8VH8w | User | false |
huggingface/diffusers | 1,856,589,310 | I_kwDOHa8MBc5uqU3- | 4,671 | https://github.com/huggingface/diffusers/issues/4671 | https://api.github.com/repos/huggingface/diffusers/issues/4671 | Question about accelerator.end_training() | It is ok when I use single gpu to run my code. When I use 4 GPU to run my code, when run `accelerator.end_training()`, I always gets this Error(although it has no influence on my training):
```
Traceback (most recent call last):
File "/home/ma/anaconda3/envs/Geo/bin/accelerate", line 8, in <module>
sys.exit... | closed | completed | false | 1 | [] | [] | 2023-08-18T11:49:19Z | 2023-08-21T03:43:27Z | 2023-08-21T03:43:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | w-yibo | 70,892,688 | MDQ6VXNlcjcwODkyNjg4 | User | false |
huggingface/diffusers | 1,856,767,295 | I_kwDOHa8MBc5urAU_ | 4,672 | https://github.com/huggingface/diffusers/issues/4672 | https://api.github.com/repos/huggingface/diffusers/issues/4672 | Reduce the VRAM requirement for training sdxl_lora | **Is your feature request related to a problem? Please describe.**
Even with lora, sdxl training requires too much VRAM.
**Describe the solution you'd like**
Add a new training script.Can we reduce the VRAM cost by precomputing ```latents,prompt_embeds,pooled_embeds```, so that vae and text_encoder can be remo... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-08-18T13:54:04Z | 2023-11-13T15:14:36Z | 2023-11-13T15:14:36Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aihao2000 | 51,043,929 | MDQ6VXNlcjUxMDQzOTI5 | User | false |
huggingface/diffusers | 1,857,032,328 | I_kwDOHa8MBc5usBCI | 4,674 | https://github.com/huggingface/diffusers/issues/4674 | https://api.github.com/repos/huggingface/diffusers/issues/4674 | Memory Leak in DiffusionPipeline when switching width/height in subsequent calls | ### Describe the bug
Using the stable-diffusion-xl-base-1.0 pipeline on a 15 GB GPU everything works as expected as long as I do not try to change the output width and height in subsequent calls to the same model. Once I do that the machine runs out of memory after just a few successful runs. I tried to fix the prob... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-18T17:09:54Z | 2023-08-21T19:27:07Z | 2023-08-21T19:26:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Quasimondo | 182,098 | MDQ6VXNlcjE4MjA5OA== | User | false |
huggingface/diffusers | 1,857,317,682 | I_kwDOHa8MBc5utGsy | 4,675 | https://github.com/huggingface/diffusers/issues/4675 | https://api.github.com/repos/huggingface/diffusers/issues/4675 | ControlNet Reference Only for SD XL | **Is your feature request related to a problem? Please describe.**
There is currently a Diffusers [example for using "Reference Only" with ControlNet.](https://github.com/huggingface/diffusers/blob/main/examples/community/stable_diffusion_controlnet_reference.py) But there is not yet one for XL.
**Describe the so... | closed | completed | false | 15 | [
"stale"
] | [] | 2023-08-18T21:36:47Z | 2024-05-24T16:42:46Z | 2024-02-04T15:06:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jelling | 1,887,486 | MDQ6VXNlcjE4ODc0ODY= | User | false |
huggingface/diffusers | 1,857,520,292 | I_kwDOHa8MBc5ut4Kk | 4,676 | https://github.com/huggingface/diffusers/issues/4676 | https://api.github.com/repos/huggingface/diffusers/issues/4676 | TAESD-encoded latents are too dark | ### Describe the bug
AutoencodeTiny (TAESD) _decoder_ seems to work fine. _encoding_ on the other hand is producing poor results, and an encode-decode round-trip turns out poorly:
input:
output:
![sad benz](ht... | closed | completed | false | 11 | [
"bug"
] | [] | 2023-08-19T03:24:18Z | 2023-08-23T03:08:39Z | 2023-08-23T03:08:39Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | keturn | 83,819 | MDQ6VXNlcjgzODE5 | User | false |
huggingface/diffusers | 1,857,547,516 | I_kwDOHa8MBc5ut-z8 | 4,677 | https://github.com/huggingface/diffusers/issues/4677 | https://api.github.com/repos/huggingface/diffusers/issues/4677 | script in official Text-guided image-inpainting doc reports an error | ### Describe the bug
Running script in official Text-guided image-inpainting doc(https://huggingface.co/docs/diffusers/using-diffusers/inpaint) reports an error
### Reproduction
```python
import PIL
import requests
import torch
from io import BytesIO
from diffusers import StableDiffusionInpaintPipeline
pip... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-19T04:09:51Z | 2023-08-24T16:57:27Z | 2023-08-24T16:57:27Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | realliujiaxu | 72,550,220 | MDQ6VXNlcjcyNTUwMjIw | User | false |
huggingface/diffusers | 1,857,559,844 | I_kwDOHa8MBc5uuB0k | 4,679 | https://github.com/huggingface/diffusers/issues/4679 | https://api.github.com/repos/huggingface/diffusers/issues/4679 | Support for controlnet lora | There's new control lora release,
https://huggingface.co/stabilityai/control-lora?s=09
How can we use it with diffusers? | open | reopened | false | 28 | [
"New pipeline/model"
] | [
"a-r-r-o-w"
] | 2023-08-19T04:42:33Z | 2025-01-30T03:04:09Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | adhikjoshi | 11,740,719 | MDQ6VXNlcjExNzQwNzE5 | User | false |
huggingface/diffusers | 1,368,862,277 | I_kwDOHa8MBc5Rly5F | 468 | https://github.com/huggingface/diffusers/issues/468 | https://api.github.com/repos/huggingface/diffusers/issues/468 | How can I transfer the self-trained model folder to a single ckpt model file? | I trained a model in local PC following this instruction: diffusers\examples\textual_inversion\README.md
The output of the self-trained model is a folder：
I can use StableDiffusionPipeline.... | closed | completed | false | 16 | [
"stale"
] | [] | 2022-09-11T07:43:15Z | 2023-05-15T11:31:26Z | 2022-12-09T15:04:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | boyjunqiang | 49,113 | MDQ6VXNlcjQ5MTEz | User | false |
huggingface/diffusers | 1,857,592,919 | I_kwDOHa8MBc5uuJ5X | 4,680 | https://github.com/huggingface/diffusers/issues/4680 | https://api.github.com/repos/huggingface/diffusers/issues/4680 | Fine tuning SDXL inpainting | **Is your feature request related to a problem? Please describe.**
Training with LORA of SDXL refiner for inpainting
**Describe the solution you'd like**
Is possibile to modify the solution proposed in: https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/README_sdxl.md
in order to train refine... | closed | completed | false | 19 | [
"stale"
] | [] | 2023-08-19T06:14:07Z | 2024-06-20T08:08:13Z | 2023-10-30T15:08:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | EnricoBeltramo | 22,332,651 | MDQ6VXNlcjIyMzMyNjUx | User | false |
huggingface/diffusers | 1,857,808,667 | I_kwDOHa8MBc5uu-kb | 4,681 | https://github.com/huggingface/diffusers/issues/4681 | https://api.github.com/repos/huggingface/diffusers/issues/4681 | Positional embedding for continuous input in Transformer2DModel | https://github.com/huggingface/diffusers/blob/74d902eb59f873b6156621220937f8e2521dfdc0/src/diffusers/models/transformer_2d.py#L136
I'm trying to understand where positional embeddings are added to the input in Transformer2DModel.
If the input is vectorized, ImagePositionalEmbeddings() is applied. If the input is ... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-08-19T17:29:12Z | 2024-09-25T06:53:04Z | 2023-10-30T15:08:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | durdur69 | 69,616,264 | MDQ6VXNlcjY5NjE2MjY0 | User | false |
huggingface/diffusers | 1,858,098,323 | I_kwDOHa8MBc5uwFST | 4,685 | https://github.com/huggingface/diffusers/issues/4685 | https://api.github.com/repos/huggingface/diffusers/issues/4685 | Issue in utilizing the Canny image in the `StableDiffusionXLControlNetPipeline` |
Hello. I'm facing an issue with utilizing the `StableDiffusionXLControlNetPipeline` for making inferences. Specifically, I'm trying to provide different types of control images, such as Canny edge or depth images to pipeline.
In the `StableDiffusionControlNetInpaintPipeline`, there's a parameter called control_imag... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-08-20T11:57:26Z | 2023-10-30T15:08:22Z | 2023-10-30T15:08:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | smmms71 | 124,030,590 | U_kgDOB2SOfg | User | false |
huggingface/diffusers | 1,858,099,638 | I_kwDOHa8MBc5uwFm2 | 4,686 | https://github.com/huggingface/diffusers/issues/4686 | https://api.github.com/repos/huggingface/diffusers/issues/4686 | Unable to load locally saved `StableDiffusionUpscalePipeline` using from_pretrained | ### Describe the bug
When attempting to load a previously saved `StableDiffusionUpscalePipeline` using the `from_pretrained` method after using `save_pretrained()` to save it to local storage as cache, the loading process fails to work as expected. Despite trying the option `use_safetensors=True`, the problem persis... | closed | completed | false | 4 | [
"bug"
] | [
"yiyixuxu"
] | 2023-08-20T12:01:36Z | 2023-08-26T14:46:23Z | 2023-08-26T14:45:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Fixe96 | 44,437,070 | MDQ6VXNlcjQ0NDM3MDcw | User | false |
huggingface/diffusers | 1,858,381,917 | I_kwDOHa8MBc5uxKhd | 4,688 | https://github.com/huggingface/diffusers/issues/4688 | https://api.github.com/repos/huggingface/diffusers/issues/4688 | Fast Training of Diffusion Models with Masked Transformers | ### Model/Pipeline/Scheduler description
**Paper Abstract**
We propose an efficient approach to train large diffusion models with masked transformers. While masked transformers have been extensively explored for representation learning, their application to generative learning is less explored in the vision domain.... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-08-21T01:13:44Z | 2024-03-13T11:21:31Z | 2023-10-30T15:08:18Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | clarencechen | 8,482,341 | MDQ6VXNlcjg0ODIzNDE= | User | false |
huggingface/diffusers | 1,858,455,631 | I_kwDOHa8MBc5uxchP | 4,689 | https://github.com/huggingface/diffusers/issues/4689 | https://api.github.com/repos/huggingface/diffusers/issues/4689 | Eliminating Lipschitz Singularities in Diffusion Models via Repeated Sampling at Low Noise Levels | ### Model/Pipeline/Scheduler description
**Abstract**
Diffusion models, which employ stochastic differential equations to sample images through integrals, have emerged as a dominant class of generative models. However, the rationality of the diffusion process itself receives limited attention, leaving the question of... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-08-21T02:46:54Z | 2023-10-30T15:08:16Z | 2023-10-30T15:08:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | clarencechen | 8,482,341 | MDQ6VXNlcjg0ODIzNDE= | User | false |
huggingface/diffusers | 1,858,853,032 | I_kwDOHa8MBc5uy9io | 4,691 | https://github.com/huggingface/diffusers/issues/4691 | https://api.github.com/repos/huggingface/diffusers/issues/4691 | torch.onnx.errors.UnsupportedOperatorError: Exporting the operator 'aten::scaled_dot_product_attention' to ONNX opset version 14 is not supported. | ### Describe the bug
`python scripts/convert_stable_diffusion_checkpoint_to_onnx.py --model_path [/path/to/model] --output_path [/path/to/output]`
The whole error message is as follow:
```
[2023-08-21 16:06:02,711] [INFO] [real_accelerator.py:133:get_accelerator] Setting ds_accelerator to cuda (auto detect)
2023... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-08-21T08:17:01Z | 2024-01-09T15:08:41Z | 2024-01-09T15:08:41Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wangqiang9 | 37,444,407 | MDQ6VXNlcjM3NDQ0NDA3 | User | false |
huggingface/diffusers | 1,859,054,704 | I_kwDOHa8MBc5uzuxw | 4,692 | https://github.com/huggingface/diffusers/issues/4692 | https://api.github.com/repos/huggingface/diffusers/issues/4692 | Issue with UNet2DConditionModel memory? | ### Describe the bug
tried running the UNet2DConditionModel with image input size of [512,512] but it keeps producing OOM error that is abnormally huge.
### Reproduction
```import torch
from diffusers import UNet2DConditionModel
def UNet2DCond (image_size):
model = UNet2DConditionModel(
sampl... | closed | completed | false | 10 | [
"bug"
] | [] | 2023-08-21T10:14:50Z | 2023-08-31T09:14:40Z | 2023-08-25T04:02:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ChristopherOoi | 107,095,832 | U_kgDOBmInGA | User | false |
huggingface/diffusers | 1,859,072,121 | I_kwDOHa8MBc5uzzB5 | 4,693 | https://github.com/huggingface/diffusers/issues/4693 | https://api.github.com/repos/huggingface/diffusers/issues/4693 | Zero123 Checkpoint is without safetensors support | The recently added "zero123" in the diffusers pipeline is without safetensors support.
Is there a version that will be released soon with safetensors support?
```
Instantiating LatentDiffusion...
**cc_projection/diffusion_pytorch_model.safetensors not found**
Loading pipeline components...: 100%|██████████████... | closed | completed | false | 1 | [] | [] | 2023-08-21T10:25:35Z | 2023-08-23T15:18:56Z | 2023-08-23T15:18:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ashish-wezzy | 61,575,330 | MDQ6VXNlcjYxNTc1MzMw | User | false |
huggingface/diffusers | 1,859,743,573 | I_kwDOHa8MBc5u2W9V | 4,697 | https://github.com/huggingface/diffusers/issues/4697 | https://api.github.com/repos/huggingface/diffusers/issues/4697 | Unable to load a specific LoRa after 0.20.0 update | ### Describe the bug
Hi! The new version breaks loading [this](https://civitai.com/models/4219?modelVersionId=10897) LoRa on SD 1.5, with the following error.
```
---------------------------------------------------------------------------
KeyError Traceback (most recent call la... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-08-21T16:35:40Z | 2023-08-22T03:17:11Z | 2023-08-22T03:17:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sinanhlvc | 66,330,822 | MDQ6VXNlcjY2MzMwODIy | User | false |
huggingface/diffusers | 1,859,826,439 | I_kwDOHa8MBc5u2rMH | 4,698 | https://github.com/huggingface/diffusers/issues/4698 | https://api.github.com/repos/huggingface/diffusers/issues/4698 | AudioLDM 2 | ### Describe the bug
Consistency check failed: file should be of size 221586761 but has size 158281580
### Reproduction
Download the music model(cvssp/audioldm2-music) and get the error.
### Logs
```shell
raise EnvironmentError(
OSError: Consistency check failed: file should be of size 221586761 but has size 1582... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-08-21T17:27:50Z | 2023-08-22T15:06:31Z | 2023-08-22T15:06:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,859,934,536 | I_kwDOHa8MBc5u3FlI | 4,699 | https://github.com/huggingface/diffusers/issues/4699 | https://api.github.com/repos/huggingface/diffusers/issues/4699 | Remove all usage of AttnProcsLayers | ### Describe the bug
All training scripts which use AttnProcsLayers will not work properly with accelerate for any accelerate feature that requires calling into the wrapped return class for monkey patching the forward method.
All lora training scripts should instead:
1. the top level module must be directly passed... | closed | completed | false | 9 | [
"bug",
"Good second issue"
] | [] | 2023-08-21T18:40:32Z | 2024-11-21T15:41:41Z | 2024-11-21T15:41:41Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,289,448,462 | I_kwDOHa8MBc5M22wO | 47 | https://github.com/huggingface/diffusers/issues/47 | https://api.github.com/repos/huggingface/diffusers/issues/47 | Glide pipeline generates wrong image dimensions | Hi,
Thank you very much for this amazing repository.
I was running the inference code for the glide pipeline, but the output is not of good dimension
This is the error... | closed | completed | false | 2 | [] | [] | 2022-06-30T00:56:04Z | 2022-06-30T23:14:39Z | 2022-06-30T23:14:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xinmiaolin | 43,487,018 | MDQ6VXNlcjQzNDg3MDE4 | User | false |
huggingface/diffusers | 1,369,058,207 | I_kwDOHa8MBc5Rmiuf | 470 | https://github.com/huggingface/diffusers/issues/470 | https://api.github.com/repos/huggingface/diffusers/issues/470 | Cannot install diffusers through PIP, conflicting dependencies | ### Describe the bug
pip install --upgrade diffusers
Collecting diffusers
Using cached diffusers-0.3.0-py3-none-any.whl (153 kB)
Collecting regex!=2019.12.17
Using cached regex-2022.9.11-cp37-cp37m-win32.whl (250 kB)
Collecting Pillow
Using cached Pillow-9.2.0-cp37-cp37m-win32.whl (2.8 MB)
Requirement alr... | closed | completed | false | 5 | [
"bug"
] | [] | 2022-09-11T20:42:55Z | 2022-09-12T12:40:58Z | 2022-09-12T12:40:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | trapbuilder2 | 15,017,349 | MDQ6VXNlcjE1MDE3MzQ5 | User | false |
huggingface/diffusers | 1,859,958,571 | I_kwDOHa8MBc5u3Lcr | 4,701 | https://github.com/huggingface/diffusers/issues/4701 | https://api.github.com/repos/huggingface/diffusers/issues/4701 | AutoencoderKL decoder throw CUDA error when differentiated at float32 | ### Describe the bug
I want to differentiate through the VAE decoder of StableDiffusion v1.5.
When using `float16` with CUDA everything works fines.
When using `float32` with CUDA I get `CUDA error: invalid argument`.
When using `float32` with CPU everything works fines.
Unfortunately, I don't have a clue wh... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-21T18:59:20Z | 2023-08-26T10:09:37Z | 2023-08-26T10:09:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | man-sean | 8,748,837 | MDQ6VXNlcjg3NDg4Mzc= | User | false |
huggingface/diffusers | 1,860,171,535 | I_kwDOHa8MBc5u3_cP | 4,702 | https://github.com/huggingface/diffusers/issues/4702 | https://api.github.com/repos/huggingface/diffusers/issues/4702 | Example script to train a VQ-VAE | **Describe the solution you'd like**
Example script to train a VQ-VAE
**Describe alternatives you've considered**
This example [train_unconditional.py](https://github.com/huggingface/diffusers/blob/main/examples/unconditional_image_generation/train_unconditional.py) is close but requires some edits such as another... | closed | completed | false | 5 | [
"community-examples",
"stale",
"contributions-welcome"
] | [] | 2023-08-21T21:21:30Z | 2023-11-26T15:07:21Z | 2023-11-26T15:07:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jbdel | 17,854,096 | MDQ6VXNlcjE3ODU0MDk2 | User | false |
huggingface/diffusers | 1,860,283,777 | I_kwDOHa8MBc5u4a2B | 4,704 | https://github.com/huggingface/diffusers/issues/4704 | https://api.github.com/repos/huggingface/diffusers/issues/4704 | Custom Diffusion : run train_custom_diffusion.py fails | ### Describe the bug
Hey guys
I've been doing the tutorials on Hugging Face, love the content! I encountered an error during training while following this documentation on [CustomDiffusion](https://huggingface.co/docs/diffusers/training/custom_diffusion). I am using the provided cat dataset, and everything ... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-08-21T22:51:22Z | 2023-10-30T15:08:10Z | 2023-10-30T15:08:10Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | canberk17 | 33,362,633 | MDQ6VXNlcjMzMzYyNjMz | User | false |
huggingface/diffusers | 1,860,335,884 | I_kwDOHa8MBc5u4nkM | 4,705 | https://github.com/huggingface/diffusers/issues/4705 | https://api.github.com/repos/huggingface/diffusers/issues/4705 | Modifying train_controlnet.py for Inpainting Question | Hi, I was trying to modify the train_controlnet.py code for stable diffusion inpainting, but after training, all of the inpainted images look noisy and like intermediate decoding steps (as shown below). I modified the script based on how latent_model_input is created in https://github.com/huggingface/diffusers/blob/mai... | closed | completed | false | 1 | [] | [] | 2023-08-21T23:58:28Z | 2023-08-26T06:32:05Z | 2023-08-22T20:01:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nkpkg23 | 141,064,371 | U_kgDOCGh4sw | User | false |
huggingface/diffusers | 1,860,822,937 | I_kwDOHa8MBc5u6eeZ | 4,708 | https://github.com/huggingface/diffusers/issues/4708 | https://api.github.com/repos/huggingface/diffusers/issues/4708 | import diffusers.utils error | ### Describe the bug
Traceback (most recent call last):
File "demo_controlnet.py", line 9, in <module>
from diffusers.utils import load_image
File "/usr/local/lib/python3.8/dist-packages/diffusers/__init__.py", line 30, in <module>
from .pipelines import OnnxRuntimeModel
File "/usr/local/lib/python3... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-08-22T07:47:50Z | 2023-10-30T15:08:06Z | 2023-10-30T15:08:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kkwhale7 | 79,788,571 | MDQ6VXNlcjc5Nzg4NTcx | User | false |
huggingface/diffusers | 1,860,887,828 | I_kwDOHa8MBc5u6uUU | 4,709 | https://github.com/huggingface/diffusers/issues/4709 | https://api.github.com/repos/huggingface/diffusers/issues/4709 | StableDiffusionXLControlNetPipeline NOT Support Guess Mode | ### Describe the bug
when use StableDiffusionXLControlNetPipeline and guess_mode is True, the pipeline not work and the error is :
RuntimeError: Expected size for first two dimensions of batch2 tensor to be: [20, 64] but got: [10, 64].
### Reproduction
controlnet_model = ControlNetModel.from_pretrained ("")
s... | closed | completed | false | 1 | [
"bug"
] | [
"yiyixuxu"
] | 2023-08-22T08:16:59Z | 2023-08-28T16:51:19Z | 2023-08-28T16:51:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chuxing | 4,728,277 | MDQ6VXNlcjQ3MjgyNzc= | User | false |
huggingface/diffusers | 1,861,541,051 | I_kwDOHa8MBc5u9Ny7 | 4,711 | https://github.com/huggingface/diffusers/issues/4711 | https://api.github.com/repos/huggingface/diffusers/issues/4711 | want to fit controlnet inference in 12 gb Ram( 2080 ti GPU ) | i have trained Controlnet on 2080 0ti with the given script in Readme. but inferece is not fitting. can someone help me with it.
i have tried
#pipe.enable_model_cpu_offload()
pipe.enable_sequential_cpu_offload()
add no grad grd accummulation
pipe.enable_vae_slicing()
pipe.enable_vae_tiling()
pipe.enable_xform... | closed | completed | false | 1 | [] | [] | 2023-08-22T14:04:57Z | 2023-09-13T12:41:37Z | 2023-08-22T15:23:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AparnaAgrawal02 | 52,852,194 | MDQ6VXNlcjUyODUyMTk0 | User | false |
huggingface/diffusers | 1,861,817,878 | I_kwDOHa8MBc5u-RYW | 4,716 | https://github.com/huggingface/diffusers/issues/4716 | https://api.github.com/repos/huggingface/diffusers/issues/4716 | How to handle SDXL long prompt | ### Describe the bug
I am unable to use embeds prompt in order to handle prompt that is longer than 77 tokens.
### Reproduction
```python
import itertools
import os.path
import random
import string
import time
import typing as typ
import torch
from diffusers import StableDiffusionXLPipeline
from tqdm impo... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-08-22T16:28:25Z | 2023-08-27T02:46:18Z | 2023-08-27T02:46:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | elcolie | 18,206,728 | MDQ6VXNlcjE4MjA2NzI4 | User | false |
huggingface/diffusers | 1,861,853,770 | I_kwDOHa8MBc5u-aJK | 4,718 | https://github.com/huggingface/diffusers/issues/4718 | https://api.github.com/repos/huggingface/diffusers/issues/4718 | Add support for Outpainting with SDXL | SDXL for Outpainting functionality can be improved. Attached are some attempts at Outpainting with `StableDiffusionXLInpaintPipeline` and `stabilityai/stable-diffusion-xl-base-1.0` checkpoint.
Given this image/mask combination ((black pixels ignored and white pixels are to be filled)
<p float="left">
<img src=... | closed | completed | false | 8 | [
"stale"
] | [] | 2023-08-22T16:54:00Z | 2024-11-21T16:34:08Z | 2024-11-21T16:34:07Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DN6 | 7,529,846 | MDQ6VXNlcjc1Mjk4NDY= | User | false |
huggingface/diffusers | 1,861,918,965 | I_kwDOHa8MBc5u-qD1 | 4,720 | https://github.com/huggingface/diffusers/issues/4720 | https://api.github.com/repos/huggingface/diffusers/issues/4720 | AutoencoderTiny is incompatible with ImageToImage, Inpainting Pipelines | ### Describe the bug
Current Pipelines are implemented to expect VAE to be `AutoencoderKL` which encodes image **into distribution and not latents**. Such distribution is further sampled inside of a pipe method `_encode_vae_image`:
```
image_latents = self.vae.encode(image).latent_dist.sample(generator=generator)
... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-08-22T17:43:44Z | 2023-10-30T15:07:59Z | 2023-10-30T15:07:59Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | slep0v | 37,597,789 | MDQ6VXNlcjM3NTk3Nzg5 | User | false |
huggingface/diffusers | 1,862,269,933 | I_kwDOHa8MBc5u__vt | 4,723 | https://github.com/huggingface/diffusers/issues/4723 | https://api.github.com/repos/huggingface/diffusers/issues/4723 | Flax StableDiffusion models give wrong from_config deprecation warning | ### Describe the bug
loading models using `from_pretrained` causes a warning that `from_config` is deprecated. but i'm not calling `from_config`. happens with VAE, Unet, and Scheduler.
suggestion on the warning - should indicate what to replace the supposed `from_config` call with, to be future compatible. i.e., di... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-08-22T22:07:30Z | 2023-11-09T15:06:13Z | 2023-11-09T15:06:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | krahnikblis | 84,637,076 | MDQ6VXNlcjg0NjM3MDc2 | User | false |
huggingface/diffusers | 1,862,296,009 | I_kwDOHa8MBc5vAGHJ | 4,724 | https://github.com/huggingface/diffusers/issues/4724 | https://api.github.com/repos/huggingface/diffusers/issues/4724 | StableDiffusionXLImg2ImgPipeline not working as expected when using with stable-diffusion-xl-base-1.0 checkpoint | ### Describe the bug
I can't get the classic img2img example to work with sdxl 1.0
this is the output I get using the script I attached in reproduction section
as a reference, here is outpu... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-08-22T22:40:03Z | 2023-08-24T16:06:27Z | 2023-08-24T08:40:20Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.