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 | 1,973,216,796 | I_kwDOHa8MBc51nOYc | 5,616 | https://github.com/huggingface/diffusers/issues/5616 | https://api.github.com/repos/huggingface/diffusers/issues/5616 | running SDXL model with downloaded safetensors give error | ### Describe the bug
```
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Cell In[15], line 1
----> 1 pipe = StableDiffusionXLPipeline.from_single_file(
2 "model/sdxl-DevlishPhotoRealism.safetens... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2023-11-01T22:44:36Z | 2024-01-19T15:06:13Z | 2024-01-19T15:06:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,973,232,742 | I_kwDOHa8MBc51nSRm | 5,617 | https://github.com/huggingface/diffusers/issues/5617 | https://api.github.com/repos/huggingface/diffusers/issues/5617 | Compiled with dynamo (torch.compile) img2imgControlNet pipeline not able to be loaded after save_pretrained | ### Describe the bug
Compiled img2imgControlNet with dynamo cannot be loaded from pretrained folder after compile/save
maybe also bug: if you compile unet/controlnet together or in reverse order -it doesn't work too.
### Reproduction
```
from diffusers import StableDiffusionControlNetPipeline, StableDiffus... | closed | completed | false | 5 | [
"bug"
] | [
"DN6"
] | 2023-11-01T23:03:09Z | 2024-01-11T17:24:33Z | 2024-01-11T17:24:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Alexadar | 14,125,937 | MDQ6VXNlcjE0MTI1OTM3 | User | false |
huggingface/diffusers | 1,973,396,732 | I_kwDOHa8MBc51n6T8 | 5,619 | https://github.com/huggingface/diffusers/issues/5619 | https://api.github.com/repos/huggingface/diffusers/issues/5619 | AutoencodeTiny doesn't work for LCM img2img when passing an image to encode | ### Describe the bug
> AttributeError: 'AutoencoderTinyOutput' object has no attribute 'latent_dist'
A normal(?) VAE has 'latent_dist'. Tiny has "latents" instead.
The custom extension: latent_consistency_img2img.py does:
> self.vae.encode(image[i : i + 1]).latent_dist.sample(generator[i]) for i in range(batch... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2023-11-02T02:34:31Z | 2023-12-04T02:21:38Z | 2023-12-03T20:22:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aifartist | 116,415,616 | U_kgDOBvBcgA | User | false |
huggingface/diffusers | 1,377,807,676 | I_kwDOHa8MBc5SH608 | 562 | https://github.com/huggingface/diffusers/issues/562 | https://api.github.com/repos/huggingface/diffusers/issues/562 | StableDiffusionInpaintPipeline does not work with LMSDiscreteScheduler | ### Describe the bug
`StableDiffusionInpaintPipeline` does not work with `LMSDiscreteScheduler`
### Reason
- This seems to be because of the absence of special operations for `LMSDiscreteScheduler` as performed in `StableDiffusionImg2ImgPipeline`
https://github.com/huggingface/diffusers/blob/429dace10a356a... | closed | completed | false | 1 | [
"bug"
] | [] | 2022-09-19T11:28:50Z | 2022-09-20T17:10:46Z | 2022-09-20T17:10:46Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shirayu | 963,961 | MDQ6VXNlcjk2Mzk2MQ== | User | false |
huggingface/diffusers | 1,973,704,210 | I_kwDOHa8MBc51pFYS | 5,621 | https://github.com/huggingface/diffusers/issues/5621 | https://api.github.com/repos/huggingface/diffusers/issues/5621 | sdxl controlNet not support encoder_hid_proj model | ### Describe the bug
I was training sdxl+controlNet recently, but I changed sdxl model,when I only used a single text encode to extract prompt_embedding. In the UNet2DConditionModel implementation, feature mapping was performed by using encoder_hid_proj, but in controlNet, this function was not implemented, causing th... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-11-02T07:58:38Z | 2023-12-26T15:09:13Z | 2023-12-26T15:09:13Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | junqiangwu | 32,137,981 | MDQ6VXNlcjMyMTM3OTgx | User | false |
huggingface/diffusers | 1,973,870,553 | I_kwDOHa8MBc51pt_Z | 5,623 | https://github.com/huggingface/diffusers/issues/5623 | https://api.github.com/repos/huggingface/diffusers/issues/5623 | Enhanced safety checker functionality | **Is your feature request related to a problem? Please describe.**
No
**Describe the solution you'd like.**
I request the ability to flexibly configure the safety checker.
**Describe alternatives you've considered.**
[I made a trial.](https://github.com/suzukimain/diffusers_mod.git) However, I am not sure if i... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-11-02T09:44:42Z | 2023-12-27T15:06:09Z | 2023-12-27T15:06:09Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | suzukimain | 131,413,573 | U_kgDOB9U2RQ | User | false |
huggingface/diffusers | 1,974,627,580 | I_kwDOHa8MBc51smz8 | 5,628 | https://github.com/huggingface/diffusers/issues/5628 | https://api.github.com/repos/huggingface/diffusers/issues/5628 | When using LMS scheduler with Karras sigmas, denoising_end and denoising_start seem to get offset for some reason. | ### Describe the bug
I've been playing around with the LMS scheduler trying to understand a different bug (which I'll file a separate issue for) and noticed that denoising_start and denoising_end don't seem to work correctly when using LMS with Karras sigmas.
With `denoising_end=0.9` and `num_inference_steps=50`,... | closed | not_planned | false | 4 | [
"bug"
] | [
"yiyixuxu"
] | 2023-11-02T16:32:19Z | 2023-11-15T01:20:37Z | 2023-11-15T01:20:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nhnt11 | 485,488 | MDQ6VXNlcjQ4NTQ4OA== | User | false |
huggingface/diffusers | 1,974,648,645 | I_kwDOHa8MBc51sr9F | 5,629 | https://github.com/huggingface/diffusers/issues/5629 | https://api.github.com/repos/huggingface/diffusers/issues/5629 | Generated images contain a green background with foreground objects | ### Describe the bug
Hi,
For any input image and any prompt, I am continually getting green backgrounds.
I am using the CompVis/stable-diffusion-v1-4 model to generate an image to image with a prompt similar to:
`Abstract pixel art of gramophone with blue background`
With many different variations of stren... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-11-02T16:43:28Z | 2024-01-19T15:06:11Z | 2024-01-19T15:06:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | baarongit | 8,247,514 | MDQ6VXNlcjgyNDc1MTQ= | User | false |
huggingface/diffusers | 1,974,682,204 | I_kwDOHa8MBc51s0Jc | 5,630 | https://github.com/huggingface/diffusers/issues/5630 | https://api.github.com/repos/huggingface/diffusers/issues/5630 | LMS scheduler leaves a lot of noise leftover in the result image when used with SDXL Img2Img Pipeline | ### Describe the bug
When using the LMS scheduler with SDXL Img2Img pipeline, there is a lot of noise leftover in the image especially when `strength` is closer to `0`. In other words, when the total number of performed steps is "low" (e.g. `num_inference_steps=50` and `strength=0.1`), the result images are unusably... | closed | completed | false | 11 | [
"bug",
"stale",
"scheduler"
] | [
"yiyixuxu"
] | 2023-11-02T17:04:46Z | 2024-01-26T15:07:25Z | 2024-01-26T15:07:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nhnt11 | 485,488 | MDQ6VXNlcjQ4NTQ4OA== | User | false |
huggingface/diffusers | 1,974,720,137 | I_kwDOHa8MBc51s9aJ | 5,631 | https://github.com/huggingface/diffusers/issues/5631 | https://api.github.com/repos/huggingface/diffusers/issues/5631 | MultiControlNet does not perform correct input/output chaining | ### Describe the bug
I may be wrong, but looking at primary for loop in `pipelines/controlnet/multicontrolnet.py`, it seems its using same input for all controlnets in list.
that is a valid use case, but somewhat rare.
far more common use case for **MultiControlNetModel** is to be able to use output of first c... | closed | completed | false | 6 | [
"bug",
"help wanted",
"Good second issue",
"contributions-welcome",
"low-priority"
] | [
"yiyixuxu"
] | 2023-11-02T17:19:33Z | 2025-02-12T16:22:36Z | 2025-02-12T16:22:36Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 1,974,803,419 | I_kwDOHa8MBc51tRvb | 5,633 | https://github.com/huggingface/diffusers/issues/5633 | https://api.github.com/repos/huggingface/diffusers/issues/5633 | A bug in diffusers/models/transformer_temporal.py | ### Describe the bug
Line 174 of diffusers/models/transformer_temporal.py is currently `.reshape(batch_size, height, width, channel, num_frames)`, which is part of the following code block:
# 3. Output
hidden_states = self.proj_out(hidden_states)
hidden_states = (
hidden_state... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-11-02T18:09:20Z | 2023-11-05T13:54:39Z | 2023-11-05T13:54:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | neilfei | 29,170,502 | MDQ6VXNlcjI5MTcwNTAy | User | false |
huggingface/diffusers | 1,976,218,579 | I_kwDOHa8MBc51yrPT | 5,643 | https://github.com/huggingface/diffusers/issues/5643 | https://api.github.com/repos/huggingface/diffusers/issues/5643 | How to use the ip adapter controlnet? | Hi, I can't use this specific controlnet because it's from here: https://huggingface.co/lllyasviel/sd_control_collection/tree/main
and the format doesn't allow from_pretrained. When I use from_single_file, I get:
```
stable_diffusion/convert_from_ckpt.py", line 422, in convert_ldm_unet_checkpoint
new_checkp... | closed | completed | false | 3 | [] | [] | 2023-11-03T13:34:44Z | 2023-11-13T15:12:29Z | 2023-11-13T15:12:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,976,873,323 | I_kwDOHa8MBc511LFr | 5,645 | https://github.com/huggingface/diffusers/issues/5645 | https://api.github.com/repos/huggingface/diffusers/issues/5645 | New LCM support in core seems broken | ### Describe the bug
Trying to load LCM model using standard methods (`DiffusionPipeline` or `AutoPipelineForText2Image`)
fails due to missing `scheduler` component in pipeline.
(note that LCM has its own scheduler, so that should be handled by the pipeline).
Loading same model using old way via `custom_pipelin... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-11-03T20:16:21Z | 2023-11-06T15:00:39Z | 2023-11-06T14:45:28Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 1,977,277,586 | I_kwDOHa8MBc512tyS | 5,646 | https://github.com/huggingface/diffusers/issues/5646 | https://api.github.com/repos/huggingface/diffusers/issues/5646 | DPM/UniPC schedulers (possibly more) seem to have massive stability issues on model KohakuXL vs ComfyUI | ### Describe the bug
The model KohakuXL in diffusers, seems to run into a massive noise/stability issue that doesn't seem to be present within ComfyUI/AUTO1111, on UniPC/DPM schedulers (possibly more) that doesn't go away with high step counts.
At 15S:
![test](https://github.com/huggingface/diffusers/assets/63993670... | closed | completed | false | 14 | [
"bug",
"stale",
"scheduler"
] | [
"yiyixuxu"
] | 2023-11-04T10:46:38Z | 2024-04-29T23:38:04Z | 2024-04-29T23:38:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Xynonners | 63,993,670 | MDQ6VXNlcjYzOTkzNjcw | User | false |
huggingface/diffusers | 1,977,564,548 | I_kwDOHa8MBc513z2E | 5,647 | https://github.com/huggingface/diffusers/issues/5647 | https://api.github.com/repos/huggingface/diffusers/issues/5647 | Bug in StableDiffusionXLInpaintPipeline when strength==1 | ### Describe the bug
When you run inference with `strength==1` (is_max_strength set to True), you fail with the following error:
```
...
...
...
File /usr/local/lib/python3.10/dist-packages/torch/utils/_contextlib.py:115, in context_decorator.<locals>.decorate_context(*args, **kwargs)
112 @functools.wrap... | closed | completed | false | 8 | [
"bug"
] | [] | 2023-11-05T00:09:47Z | 2023-11-20T08:39:53Z | 2023-11-20T08:39:53Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | a-r-r-o-w | 72,266,394 | MDQ6VXNlcjcyMjY2Mzk0 | User | false |
huggingface/diffusers | 1,978,694,146 | I_kwDOHa8MBc518HoC | 5,654 | https://github.com/huggingface/diffusers/issues/5654 | https://api.github.com/repos/huggingface/diffusers/issues/5654 | Min-SNR weight for v-prediction in examples | In #5238, the [Min-SNR](https://openaccess.thecvf.com/content/ICCV2023/papers/Hang_Efficient_Diffusion_Training_via_Min-SNR_Weighting_Strategy_ICCV_2023_paper.pdf) weight for v-prediction is implemented as
$$w_t = \frac {min(\text{SNR}+1, \gamma)} {\text{SNR}+1}$$
but it should be
$$w_t = \frac {min(\text{SNR}, \g... | closed | completed | false | 15 | [] | [] | 2023-11-06T09:35:32Z | 2024-01-27T03:48:10Z | 2024-01-27T03:48:10Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | thuliu-yt16 | 27,289,204 | MDQ6VXNlcjI3Mjg5MjA0 | User | false |
huggingface/diffusers | 1,978,800,163 | I_kwDOHa8MBc518hgj | 5,656 | https://github.com/huggingface/diffusers/issues/5656 | https://api.github.com/repos/huggingface/diffusers/issues/5656 | Add LORA Loader to community pipelines | **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 [...]
**Problem**
When using any community pipeline none of the `LoraLoaderMixin` methods are available and I have to either implement the methods myself or find... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-11-06T10:29:38Z | 2023-12-26T15:09:09Z | 2023-12-26T15:09:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | IancuVerghelet | 23,400,886 | MDQ6VXNlcjIzNDAwODg2 | User | false |
huggingface/diffusers | 1,979,564,926 | I_kwDOHa8MBc51_cN- | 5,661 | https://github.com/huggingface/diffusers/issues/5661 | https://api.github.com/repos/huggingface/diffusers/issues/5661 | name 'PreTrainedTokenizer' is not defined in diffusers 0.22.1 | ### Describe the bug
when i try to loading pretrained model using DDPMPipeline or DiffusionPipeline, i take error `Failed to import diffusers.models.unet_2d because of the following error (look up to see its traceback):
name 'PreTrainedTokenizer' is not defined`
### Reproduction
`
from diffusers import Diffu... | closed | completed | false | 8 | [
"bug"
] | [] | 2023-11-06T16:31:55Z | 2023-11-07T16:38:13Z | 2023-11-07T16:38:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | WiNE-iNEFF | 41,611,046 | MDQ6VXNlcjQxNjExMDQ2 | User | false |
huggingface/diffusers | 1,980,249,428 | I_kwDOHa8MBc52CDVU | 5,666 | https://github.com/huggingface/diffusers/issues/5666 | https://api.github.com/repos/huggingface/diffusers/issues/5666 | Support OpenAI's Consistency Decoder | Source - https://github.com/openai/consistencydecoder
OpenAI open-sourced consistency decoder, that decodes latent like a pro. Makes images more consistent and less disturbing
But. Right now I see no options to p... | closed | completed | false | 4 | [] | [] | 2023-11-06T23:17:51Z | 2023-11-09T21:42:52Z | 2023-11-09T21:42:51Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Pirog17000 | 3,022,852 | MDQ6VXNlcjMwMjI4NTI= | User | false |
huggingface/diffusers | 1,980,367,486 | I_kwDOHa8MBc52CgJ- | 5,667 | https://github.com/huggingface/diffusers/issues/5667 | https://api.github.com/repos/huggingface/diffusers/issues/5667 | [Hub + Examples] Standardize model cards in the training scripts | The [model card](https://github.com/huggingface/diffusers/blob/64603389da01082055a901f2883c4810d1144edb/src/diffusers/utils/hub_utils.py#L125) helper in diffusers does not set either a `finetuned_from` or `base_model` metadata.
If we follow the template and start adding the `finetuned_from` or `base_model` tag in t... | closed | completed | false | 12 | [] | [
"sayakpaul"
] | 2023-11-07T01:30:15Z | 2024-02-09T02:24:32Z | 2024-02-07T09:37:12Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,980,480,838 | I_kwDOHa8MBc52C71G | 5,669 | https://github.com/huggingface/diffusers/issues/5669 | https://api.github.com/repos/huggingface/diffusers/issues/5669 | [Refactor] Decouple the different loaders and create a separate `loaders` module | Currently, we have a big monolith [`loaders.py`](https://github.com/huggingface/diffusers/blob/main/src/diffusers/loaders.py) that has got all kinds of different loaders classes (`LoraLoaderMixin`, `FromSingleFileMixin`, etc.).
I propose to create a separate `loaders` module now that we have a growing number of diff... | closed | completed | false | 5 | [] | [
"sayakpaul"
] | 2023-11-07T03:39:15Z | 2023-11-15T01:49:26Z | 2023-11-14T11:54:30Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,377,973,668 | I_kwDOHa8MBc5SIjWk | 567 | https://github.com/huggingface/diffusers/issues/567 | https://api.github.com/repos/huggingface/diffusers/issues/567 | Save training `dtype` as part of the configuration | It'd be interesting to know the type that was used to train. However, as pointed out by @patrickvonplaten and @patil-suraj:
* It needs to be saved as a string; `jnp.bfloat16` etc. cannot be serialized.
* We need to decide the naming to make it clear what it means.
Context: https://github.com/huggingface/diffuser... | closed | completed | false | 2 | [
"stale"
] | [] | 2022-09-19T13:36:36Z | 2022-10-31T15:03:00Z | 2022-10-31T15:03:00Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pcuenca | 1,177,582 | MDQ6VXNlcjExNzc1ODI= | User | false |
huggingface/diffusers | 1,980,550,772 | I_kwDOHa8MBc52DM50 | 5,670 | https://github.com/huggingface/diffusers/issues/5670 | https://api.github.com/repos/huggingface/diffusers/issues/5670 | PixArtAlphaPipeline raise RuntimeError when try to generate non-square image | ### Describe the bug
when specify non-square height, width to pipeline, raise `RuntimeError`
### Reproduction
```
from diffusers import PixArtAlphaPipeline
import torch
pipe = PixArtAlphaPipeline.from_pretrained("PixArt-alpha/PixArt-XL-2-1024-MS", torch_dtype=torch.bfloat16)
pipe.enable_model_cpu_offload... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-11-07T05:00:29Z | 2023-11-07T17:21:34Z | 2023-11-07T17:21:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | s60912frank | 14,332,188 | MDQ6VXNlcjE0MzMyMTg4 | User | false |
huggingface/diffusers | 1,980,796,151 | I_kwDOHa8MBc52EIz3 | 5,676 | https://github.com/huggingface/diffusers/issues/5676 | https://api.github.com/repos/huggingface/diffusers/issues/5676 | why does it take 30 seconds to load lora and generate images using PEFT? | I upgraded lora's loading method to PEFT. It can load multiple loras normally, but compared to the previous diffusers version, it only takes less than 10 seconds to load a single lora, while the PEFT method takes 30s+. The graphics card I use is 3090.
my code：
adapter_name = weight_name.replace('.safetensors','... | closed | completed | false | 4 | [] | [] | 2023-11-07T08:19:53Z | 2023-11-08T06:43:20Z | 2023-11-08T06:43:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chinatian | 3,233,111 | MDQ6VXNlcjMyMzMxMTE= | User | false |
huggingface/diffusers | 1,980,859,526 | I_kwDOHa8MBc52EYSG | 5,678 | https://github.com/huggingface/diffusers/issues/5678 | https://api.github.com/repos/huggingface/diffusers/issues/5678 | "scaled_dot_product_attention" using in "models/attention_processor.py" may lead to error when using autograd to calculate higher order derivatives | ### Describe the bug
In `models/attention_processor.py` we will use `AttnProcessor2_0()` instead of `AttnProcessor()` when torch 2.x is used which uses `torch.nn.functional.scaled_dot_product_attention` for native Flash/memory_efficient_attention , but this may lead to some error saying `derivative for aten::_scaled_d... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-11-07T08:58:48Z | 2023-12-26T15:09:01Z | 2023-12-26T15:09:01Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhongzero | 34,911,655 | MDQ6VXNlcjM0OTExNjU1 | User | false |
huggingface/diffusers | 1,981,569,905 | I_kwDOHa8MBc52HFtx | 5,684 | https://github.com/huggingface/diffusers/issues/5684 | https://api.github.com/repos/huggingface/diffusers/issues/5684 | Why NCSN series models and pipelines have been deleted？ | null | closed | completed | false | 1 | [] | [] | 2023-11-07T15:00:39Z | 2023-11-13T16:40:36Z | 2023-11-13T16:40:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LYMDLUT | 70,597,027 | MDQ6VXNlcjcwNTk3MDI3 | User | false |
huggingface/diffusers | 1,981,711,118 | I_kwDOHa8MBc52HoMO | 5,685 | https://github.com/huggingface/diffusers/issues/5685 | https://api.github.com/repos/huggingface/diffusers/issues/5685 | Potential scheduler issues | ### Describe the bug
I was adjusting the scheduler configs and I've come across a few potential issues.
**Edit: the 1. is a duplicate of #5628**
## 1. Different `denoising_start` and `denoising_end` calculation for different schedulers.
Repro:
```python
import torch
from diffusers import DiffusionPipel... | closed | completed | false | 4 | [
"bug",
"stale"
] | [
"yiyixuxu"
] | 2023-11-07T16:05:46Z | 2023-12-26T15:08:59Z | 2023-12-26T15:08:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vakker | 3,268,882 | MDQ6VXNlcjMyNjg4ODI= | User | false |
huggingface/diffusers | 1,981,871,353 | I_kwDOHa8MBc52IPT5 | 5,687 | https://github.com/huggingface/diffusers/issues/5687 | https://api.github.com/repos/huggingface/diffusers/issues/5687 | Duplicate timesteps when using Karras sigmas can break Img2Img with low strength | ### Describe the bug
When using Karras sigmas, at high `num_inference_steps` (e.g. 100) and low `strength` (e.g. 0.1), image to image breaks with an error like `a Tensor with 2 elements cannot be converted to Scalar`.
The root cause seems to be duplicate timesteps. The final few sigmas are so small that they end... | closed | completed | false | 8 | [
"bug",
"stale",
"scheduler"
] | [
"yiyixuxu"
] | 2023-11-07T17:29:49Z | 2024-02-16T17:09:33Z | 2024-02-16T17:09:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nhnt11 | 485,488 | MDQ6VXNlcjQ4NTQ4OA== | User | false |
huggingface/diffusers | 1,981,898,936 | I_kwDOHa8MBc52IWC4 | 5,688 | https://github.com/huggingface/diffusers/issues/5688 | https://api.github.com/repos/huggingface/diffusers/issues/5688 | Can we go Diffusers --> ckpt/st --> merge_with_other_models --> back_to_diffusers? | **Is your feature request related to a problem? Please describe.**
Please help!! I've been dealing with this for several weeks now
I have a trained diffusers local model in "unet, vae, etc" folders..
When I load the model with
DiffusionPipeline.from_pretrained
it works just fine, I can see the trained results... | closed | completed | false | 0 | [] | [] | 2023-11-07T17:44:34Z | 2023-11-07T23:38:22Z | 2023-11-07T23:38:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aaaleeexTLC | 132,088,617 | U_kgDOB9-DKQ | User | false |
huggingface/diffusers | 1,982,225,210 | I_kwDOHa8MBc52Jls6 | 5,689 | https://github.com/huggingface/diffusers/issues/5689 | https://api.github.com/repos/huggingface/diffusers/issues/5689 | DPM++ leaves residual noise in SDXL images, which is unexpected | ### Describe the bug
When using DPM++ with SDXL, there is residual noise in the result.
Upon investigation, my colleague @CodeCorrupt and I found that the final sigma was non-zero which we empirically deemed to be the culprit - i.e. when we set the final sigma to 0 artificially, the images were clean.
### Reproduc... | open | null | false | 13 | [
"bug",
"stale",
"scheduler"
] | [
"yiyixuxu"
] | 2023-11-07T21:04:45Z | 2024-03-19T15:04:32Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nhnt11 | 485,488 | MDQ6VXNlcjQ4NTQ4OA== | User | false |
huggingface/diffusers | 1,982,585,743 | I_kwDOHa8MBc52K9uP | 5,691 | https://github.com/huggingface/diffusers/issues/5691 | https://api.github.com/repos/huggingface/diffusers/issues/5691 | About the inpaint mode: fill, original, latent noise, latent nothing | There is different mode to use inpainting in WebUI, and it can get the different result.
Could this be implemented in diffusers ???
| closed | completed | false | 5 | [] | [
"yiyixuxu"
] | 2023-11-08T02:12:48Z | 2024-03-12T01:13:00Z | 2023-11-10T03:32:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | atlantistin | 8,433,753 | MDQ6VXNlcjg0MzM3NTM= | User | false |
huggingface/diffusers | 1,982,730,529 | I_kwDOHa8MBc52LhEh | 5,692 | https://github.com/huggingface/diffusers/issues/5692 | https://api.github.com/repos/huggingface/diffusers/issues/5692 | pipeline_pixart_alpha negative prompt normalization | Hello,
first great work on implementing pixart-alpha so fast, I am really happy to be able to try it!
**Is your feature request related to a problem? Please describe.**
I have noticed that, when using negative prompt, if the negative prompt is much shorter than the prompt it seems to have too much effect.... | closed | completed | false | 2 | [] | [] | 2023-11-08T04:44:21Z | 2023-11-23T23:16:42Z | 2023-11-23T23:16:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Bigfield77 | 108,305,548 | U_kgDOBnScjA | User | false |
huggingface/diffusers | 1,982,749,472 | I_kwDOHa8MBc52Llsg | 5,693 | https://github.com/huggingface/diffusers/issues/5693 | https://api.github.com/repos/huggingface/diffusers/issues/5693 | load vae with error | ### Describe the bug
I am trying to load the vae model from https://huggingface.co/runwayml/stable-diffusion-inpainting using the code
```
from diffusers import AutoencoderKL
model_path = "huggingface_cache/models--runwayml--stable-diffusion-inpainting"
vae = AutoencoderKL.from_pretrained(model_path, subfolder="... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-11-08T05:04:40Z | 2024-07-31T11:55:48Z | 2023-12-26T15:08:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | garychan22 | 108,175,311 | U_kgDOBnKfzw | User | false |
huggingface/diffusers | 1,983,333,607 | I_kwDOHa8MBc52N0Tn | 5,701 | https://github.com/huggingface/diffusers/issues/5701 | https://api.github.com/repos/huggingface/diffusers/issues/5701 | The output of the StableDiffusionControlNetInpaintPipeline significantly degrades the image quality | When comparing the output images from the StableDiffusionControlNetInpaintPipeline with the input images, a noticeable degradation in image quality is observed, But when viewing the results in the AUTOMATIC1111 UI, there is a certain degree of improvement in image quality
Are there any known issues that might explain ... | closed | completed | false | 5 | [] | [] | 2023-11-08T10:59:44Z | 2024-03-12T10:57:45Z | 2023-11-16T03:51:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mtyuanhangzheng | 138,104,823 | U_kgDOCDtP9w | User | false |
huggingface/diffusers | 1,983,422,372 | I_kwDOHa8MBc52OJ-k | 5,703 | https://github.com/huggingface/diffusers/issues/5703 | https://api.github.com/repos/huggingface/diffusers/issues/5703 | Cannot reconstruct noise using ddim_inverse scheduler. | ### Describe the bug
I'm currently trying to replicate the experiments in DDIM( 5.4 RECONSTRUCTION FROM LATENT SPACE), but the results are quite different from what's reported in the paper. I wanted to ask if you've also conducted experiments on inverting image back into noise, and what were the results?
### Repr... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-11-08T11:51:45Z | 2024-01-04T07:48:25Z | 2023-11-28T07:57:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yanzengatsmu | 43,662,186 | MDQ6VXNlcjQzNjYyMTg2 | User | false |
huggingface/diffusers | 1,983,727,501 | I_kwDOHa8MBc52PUeN | 5,705 | https://github.com/huggingface/diffusers/issues/5705 | https://api.github.com/repos/huggingface/diffusers/issues/5705 | Issue: Error handling of invalid orders | ### Describe the bug
When I attempt to create an XYZ grid using all available Samplers (other than Default) with an SDXL model in the X axis, and [Sampler] solver order: 1-5 in the Y axis, and nothing for Z axis, I get the below error (see relevant log export).
Final error is "UnboundLocalError: local variable 'ord... | closed | completed | false | 20 | [
"bug",
"good first issue",
"contributions-welcome"
] | [] | 2023-11-08T14:22:18Z | 2024-03-14T04:29:43Z | 2024-03-14T04:29:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghost | 10,137 | MDQ6VXNlcjEwMTM3 | User | false |
huggingface/diffusers | 1,983,965,188 | I_kwDOHa8MBc52QOgE | 5,709 | https://github.com/huggingface/diffusers/issues/5709 | https://api.github.com/repos/huggingface/diffusers/issues/5709 | How to run stable diffusion pipeline using multithreading in fastapi ? | Hi.. I have created an stable diffusion API using Fastapi and it is working perfectly fine if sequential request are been made. I have tried to implement multithreading in the api to concurrently run multiple request, but the problem is every request output generation time is dependent on total number of request that a... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-11-08T16:19:45Z | 2024-01-09T15:07:46Z | 2024-01-09T15:07:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | minkvirparia | 44,423,972 | MDQ6VXNlcjQ0NDIzOTcy | User | false |
huggingface/diffusers | 1,378,013,987 | I_kwDOHa8MBc5SItMj | 571 | https://github.com/huggingface/diffusers/issues/571 | https://api.github.com/repos/huggingface/diffusers/issues/571 | Typo checking at regular intervals | **Is your feature request related to a problem? Please describe.**
Currently, the typo check with GitHub Actions runs manually. (#483)
It is nice to run nightly/weekly typo-fixing PRs as @anton-l suggested.
He may already be working towards implementing it, but as just a reminder I make this issue.
**Addi... | closed | completed | false | 2 | [
"stale"
] | [] | 2022-09-19T14:03:43Z | 2022-10-31T15:02:58Z | 2022-10-31T15:02:58Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shirayu | 963,961 | MDQ6VXNlcjk2Mzk2MQ== | User | false |
huggingface/diffusers | 1,984,256,972 | I_kwDOHa8MBc52RVvM | 5,710 | https://github.com/huggingface/diffusers/issues/5710 | https://api.github.com/repos/huggingface/diffusers/issues/5710 | DPMSolverSinglestepScheduler lacks self.sigmas during initialization | ### Describe the bug
`DPMSolverSinglestepScheduler` not initialized with sigmas, but required for add_noise. only set after setting timesteps.
### Reproduction
```
import torch
from diffusers import DPMSolverSinglestepScheduler
foo = DPMSolverSinglestepScheduler()
placeholder_latents = torch.randn(1, 64, 64)
... | closed | completed | false | 3 | [
"bug",
"stale"
] | [
"yiyixuxu"
] | 2023-11-08T19:25:57Z | 2024-01-10T07:29:31Z | 2024-01-09T15:07:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LWprogramming | 13,173,037 | MDQ6VXNlcjEzMTczMDM3 | User | false |
huggingface/diffusers | 1,984,289,043 | I_kwDOHa8MBc52RdkT | 5,711 | https://github.com/huggingface/diffusers/issues/5711 | https://api.github.com/repos/huggingface/diffusers/issues/5711 | LCM - Scheduler Exception Handling | It would be great to handle this exception, when the combination of `strength`, `num_inference_steps` and `lcm_origin_timesteps` is too low this throws an error. @patrickvonplaten I know you're working on tests for LCM, this is a great one to add 😉
https://github.com/huggingface/diffusers/blob/6e68c71503682c8693c... | closed | completed | false | 2 | [] | [] | 2023-11-08T19:48:00Z | 2023-11-14T16:19:22Z | 2023-11-14T11:08:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | radames | 102,277 | MDQ6VXNlcjEwMjI3Nw== | User | false |
huggingface/diffusers | 1,984,328,603 | I_kwDOHa8MBc52RnOb | 5,712 | https://github.com/huggingface/diffusers/issues/5712 | https://api.github.com/repos/huggingface/diffusers/issues/5712 | pip install --upgrade diffusers[torch] fails with Python 3.12 on Windows 11 | ### Describe the bug
`pip install --upgrade diffusers[torch]` with Python 3.12 results in a failed installation, with one the (many) errors being:
`ERROR: Could not find a version that satisfies the requirement torch (from versions: none)`
Workaround:
Use Python 3.11.
Note:
Writing this to hopefully save... | closed | completed | false | 14 | [
"bug",
"stale"
] | [] | 2023-11-08T20:15:31Z | 2024-01-16T21:41:39Z | 2024-01-16T21:41:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | canxerian | 1,138,440 | MDQ6VXNlcjExMzg0NDA= | User | false |
huggingface/diffusers | 1,984,602,062 | I_kwDOHa8MBc52Sp_O | 5,714 | https://github.com/huggingface/diffusers/issues/5714 | https://api.github.com/repos/huggingface/diffusers/issues/5714 | `maybe_raise_or_warn` hasn't support PeftModel | ### Describe the bug
```
import torch
from diffusers import DiffusionPipeline, UNet2DConditionModel
from peft import LoraConfig, get_peft_model
unet = UNet2DConditionModel.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1.0", subfolder='unet', torch_dtype=torch.float16)
unet = get_peft_model(unet, L... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-11-09T00:17:39Z | 2023-12-27T00:25:37Z | 2023-12-27T00:25:37Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | okotaku | 24,734,142 | MDQ6VXNlcjI0NzM0MTQy | User | false |
huggingface/diffusers | 1,984,912,261 | I_kwDOHa8MBc52T1uF | 5,718 | https://github.com/huggingface/diffusers/issues/5718 | https://api.github.com/repos/huggingface/diffusers/issues/5718 | Error in using prompt_2 and negative_prompt_2 in SDXL with Compel | Hi ,
Thanks for sharing this projects, I want to use prompt_2 and negative_prompt_2 in SDXL model with compel, but I get the error.
My code is :
```
from diffusers import DiffusionPipeline
from diffusers.schedulers import DPMSolverMultistepScheduler,EulerDiscreteScheduler
from src.config_manager import Confi... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-11-09T06:17:28Z | 2024-01-09T15:07:42Z | 2024-01-09T15:07:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | saeedkhanehgir | 65,589,645 | MDQ6VXNlcjY1NTg5NjQ1 | User | false |
huggingface/diffusers | 1,985,066,242 | I_kwDOHa8MBc52UbUC | 5,719 | https://github.com/huggingface/diffusers/issues/5719 | https://api.github.com/repos/huggingface/diffusers/issues/5719 | Check for `pytorch_lora_weights.safetensors` or `pytorch_lora_weights.bin` first before checking for multiple files | ### Describe the bug
If a repo contains both `pytorch_lora_weights.safetensors` and some other safetensors file, loading it with
```py
lcm_lora_id = "lcm-sd/lcm-sdxl-base-1.0-lora"
pipe.load_lora_weights(lcm_lora_id, adapter_name="lora", use_auth_token=True)
```
errors out, for example, a repo with:
<img wid... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-11-09T08:16:25Z | 2023-11-09T21:40:54Z | 2023-11-09T21:40:54Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | apolinario | 788,417 | MDQ6VXNlcjc4ODQxNw== | User | false |
huggingface/diffusers | 1,985,691,372 | I_kwDOHa8MBc52Wz7s | 5,726 | https://github.com/huggingface/diffusers/issues/5726 | https://api.github.com/repos/huggingface/diffusers/issues/5726 | [Pipeline] Stable Signature by Meta | ### Model/Pipeline/Scheduler description
**Stable Signature Pipeline!**
I think Stable Signature can be a nice add on to the Diffusers pipeline. Stable Signature creates an invisible watermark that can be used to to distinguish when an image is created by an open source generative AI model or not. It's pretty inter... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-11-09T14:02:04Z | 2024-01-09T15:07:40Z | 2024-01-09T15:07:40Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | charchit7 | 21,178,353 | MDQ6VXNlcjIxMTc4MzUz | User | false |
huggingface/diffusers | 1,986,020,241 | I_kwDOHa8MBc52YEOR | 5,731 | https://github.com/huggingface/diffusers/issues/5731 | https://api.github.com/repos/huggingface/diffusers/issues/5731 | AutoencoderKL decoder implementation is missing the trainable logvar scalar from compvis codebase | ### Describe the bug
This is necessary for training/finetuning of the VAE. Please see the original implementation from compvis over at:
https://github.com/CompVis/latent-diffusion/blob/main/ldm/modules/losses/contperceptual.py#L20
### Reproduction
The trainable logvar scalar does not exist in your implementation.... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-11-09T16:44:32Z | 2024-01-09T15:07:38Z | 2024-01-09T15:07:38Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AmericanPresidentJimmyCarter | 110,263,573 | U_kgDOBpJ9FQ | User | false |
huggingface/diffusers | 1,986,227,167 | I_kwDOHa8MBc52Y2vf | 5,733 | https://github.com/huggingface/diffusers/issues/5733 | https://api.github.com/repos/huggingface/diffusers/issues/5733 | Simplify the `load_textual_inversion` API | With Pivotal Tuning for SDXL it's common to train 2 tokens per text-encoder. The current API for loading it is as follows
```py
pipe.load_textual_inversion(state_dict["text_encoders_0"][0], token=["<s0>"], text_encoder=pipe.text_encoder, tokenizer=pipe.tokenizer)
pipe.load_textual_inversion(state_dict["text_encoders... | closed | completed | false | 9 | [] | [
"yiyixuxu"
] | 2023-11-09T18:59:26Z | 2023-11-28T11:59:15Z | 2023-11-28T11:59:14Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | apolinario | 788,417 | MDQ6VXNlcjc4ODQxNw== | User | false |
huggingface/diffusers | 1,986,804,524 | I_kwDOHa8MBc52bDss | 5,737 | https://github.com/huggingface/diffusers/issues/5737 | https://api.github.com/repos/huggingface/diffusers/issues/5737 | convert_original_stable_diffusion_to_diffusers.py is failed with SDXL model | ### Describe the bug
I tried with converting a safetensor into diffusers readable. Met following issues. Please help to check it.
### Reproduction
python3 convert_original_stable_diffusion_to_diffusers.py --checkpoint_path sdxlFixedvaeFp16Remove_baseFxiedVaeV2Fp16.safetensors --dump_path BaseModel/ --from_safetensor... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-11-10T03:48:25Z | 2023-12-19T10:33:18Z | 2023-11-10T12:25:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jiancong | 701,159 | MDQ6VXNlcjcwMTE1OQ== | User | false |
huggingface/diffusers | 1,987,048,423 | I_kwDOHa8MBc52b_Pn | 5,742 | https://github.com/huggingface/diffusers/issues/5742 | https://api.github.com/repos/huggingface/diffusers/issues/5742 | where is the Parameter Description? | null | closed | completed | false | 1 | [] | [] | 2023-11-10T07:07:03Z | 2023-11-13T18:01:56Z | 2023-11-13T18:01:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MRG-DOT | 95,862,722 | U_kgDOBba_wg | User | false |
huggingface/diffusers | 1,987,073,728 | I_kwDOHa8MBc52cFbA | 5,743 | https://github.com/huggingface/diffusers/issues/5743 | https://api.github.com/repos/huggingface/diffusers/issues/5743 | [Latent Consistency Distillation] training stuck at 0% | ### Describe the bug
When using [LCM-LoRA](https://github.com/huggingface/diffusers/blob/main/examples/consistency_distillation/README_sdxl.md#lcm-lora) training script, the training just stucks at 0%, tired all different parameters, got same error.
### Reproduction
Follow the steps on this page,
[[Latent C... | closed | completed | false | 10 | [
"bug"
] | [
"patil-suraj"
] | 2023-11-10T07:28:41Z | 2023-12-06T09:35:38Z | 2023-12-06T09:35:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rdcoder33 | 38,337,897 | MDQ6VXNlcjM4MzM3ODk3 | User | false |
huggingface/diffusers | 1,987,207,860 | I_kwDOHa8MBc52cmK0 | 5,749 | https://github.com/huggingface/diffusers/issues/5749 | https://api.github.com/repos/huggingface/diffusers/issues/5749 | IndexError: index 21 is out of bounds for dimension 0 with size 21 | ### Describe the bug
I use Flask to deploy services，use StableDiffusionXLControlNetPipeline.from_single_file，When there are multiple concurrent requests, an error will be reported, and a single request will not report an error
### Reproduction
The following is the main code
image_device = 'cuda:1'
control_net = Co... | closed | completed | false | 16 | [
"bug"
] | [] | 2023-11-10T09:04:49Z | 2024-10-03T12:08:42Z | 2023-11-18T00:58:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TheHonestBob | 58,240,629 | MDQ6VXNlcjU4MjQwNjI5 | User | false |
huggingface/diffusers | 1,987,300,434 | I_kwDOHa8MBc52c8xS | 5,751 | https://github.com/huggingface/diffusers/issues/5751 | https://api.github.com/repos/huggingface/diffusers/issues/5751 | The argument down_intrablock_additional_residuals should not be changed inside the function UNet2DConditionModel.forward() | I found the argument down_intrablock_additional_residuals is poped inside the UNet2DConditionModel.forward() function, and don't think it's a good design. This argument is the result of the adapter, and is repeatedly passed into the forward function during the denoising process. If the argument is changed by the functi... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-11-10T09:57:42Z | 2024-01-09T15:07:34Z | 2024-01-09T15:07:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wzhgba | 11,329,382 | MDQ6VXNlcjExMzI5Mzgy | User | false |
huggingface/diffusers | 1,987,352,224 | I_kwDOHa8MBc52dJag | 5,753 | https://github.com/huggingface/diffusers/issues/5753 | https://api.github.com/repos/huggingface/diffusers/issues/5753 | StableDiffusionXLControlNetPipeline can't work with load_lora_weights calling. | ### Describe the bug
I've tried with StableDiffusionXLControlNetPipeline to load_lora_weights . But failed.
### Reproduction
Use this link https://colab.research.google.com/drive/1eOL_TgdAzNee5GWdODQ8OjT83C4If954?usp=sharing
### Logs
```shell
------------------------------------------------------------... | closed | completed | false | 9 | [
"bug"
] | [] | 2023-11-10T10:28:36Z | 2023-11-11T02:25:39Z | 2023-11-11T02:25:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jiancong | 701,159 | MDQ6VXNlcjcwMTE1OQ== | User | false |
huggingface/diffusers | 1,987,864,153 | I_kwDOHa8MBc52fGZZ | 5,755 | https://github.com/huggingface/diffusers/issues/5755 | https://api.github.com/repos/huggingface/diffusers/issues/5755 | Arbitrary Guidance Support | **Is your feature request related to a problem? Please describe.**
I would like to be able to guide the diffusion process with a classifier (or any other method) for controllable generation (note: not to be confused with classifier-free guidance). In my case specifically, I can't work out how to guide Stable Diffusion... | closed | completed | false | 2 | [] | [] | 2023-11-10T15:41:30Z | 2023-11-13T20:16:01Z | 2023-11-13T19:14:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zzbuzzard | 40,760,215 | MDQ6VXNlcjQwNzYwMjE1 | User | false |
huggingface/diffusers | 1,987,869,229 | I_kwDOHa8MBc52fHot | 5,756 | https://github.com/huggingface/diffusers/issues/5756 | https://api.github.com/repos/huggingface/diffusers/issues/5756 | How to we generate LCM LoRA of an existing model? | I generated a DreamBooth model from SDXL base 1.0
To get the speed boost of LCM I need to generate a LCM LoRA from this model
How we do it? I don't see documentation | closed | completed | false | 8 | [
"stale"
] | [] | 2023-11-10T15:44:52Z | 2023-12-27T13:28:38Z | 2023-12-26T21:32:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FurkanGozukara | 19,240,467 | MDQ6VXNlcjE5MjQwNDY3 | User | false |
huggingface/diffusers | 1,988,092,817 | I_kwDOHa8MBc52f-OR | 5,757 | https://github.com/huggingface/diffusers/issues/5757 | https://api.github.com/repos/huggingface/diffusers/issues/5757 | train_dreambooth_lora_sdxl.py resume from checkpoint errors. | ### Describe the bug
I get the following issue when trying to resume from checkpoint.
### Reproduction
!accelerate launch train_dreambooth_lora_sdxl.py
--pretrained_model_name_or_path=$MODEL_NAME
--instance_data_dir=$INSTANCE_DIR
--pretrained_vae_model_name_or_path=$VAE_PATH
--output_dir=$OUTPUT_DIR
--mixed_pre... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-11-10T17:56:02Z | 2024-01-29T15:12:33Z | 2024-01-29T15:12:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kosukekurimoto | 27,758,476 | MDQ6VXNlcjI3NzU4NDc2 | User | false |
huggingface/diffusers | 1,988,345,503 | I_kwDOHa8MBc52g76f | 5,758 | https://github.com/huggingface/diffusers/issues/5758 | https://api.github.com/repos/huggingface/diffusers/issues/5758 | how to run huggingface model in replicate | ### Describe the bug
i am trying to run https://medium.com/ai-artistry/streamlining-ai-agent-development-with-autogen-and-llava-b84fb0d25262 code by adding https://huggingface.co/LLaVA-VL/llava_plus_v0_7b instead of replicate code.
My Question is: Challenges running the huggingface model using replicate?
somet... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-11-10T20:31:04Z | 2023-11-11T03:33:51Z | 2023-11-11T03:31:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,378,124,383 | I_kwDOHa8MBc5SJIJf | 576 | https://github.com/huggingface/diffusers/issues/576 | https://api.github.com/repos/huggingface/diffusers/issues/576 | Remove `torch` dependency from Flax implementations | **Is your feature request related to a problem? Please describe.**
The following line fails unless PyTorch is installed:
```
from diffusers import FlaxPNDMScheduler
```
This is because `modeling_utils.py` imports `torch`. Since Flax implementations are separate (schedulers, models and pipelines use their own c... | closed | completed | false | 2 | [] | [
"patrickvonplaten"
] | 2022-09-19T15:16:58Z | 2022-10-05T10:13:04Z | 2022-10-05T10:13:04Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pcuenca | 1,177,582 | MDQ6VXNlcjExNzc1ODI= | User | false |
huggingface/diffusers | 1,988,647,896 | I_kwDOHa8MBc52iFvY | 5,760 | https://github.com/huggingface/diffusers/issues/5760 | https://api.github.com/repos/huggingface/diffusers/issues/5760 | DIFF-FOLEY: Synchronized Video-to-Audio Synthesis with Latent Diffusion Models | ### Model/Pipeline/Scheduler description
Video-to-Audio (V2A) models has recently gained attention for generating audio directly from silent videos, particularly in video/film production. However, previous methods in V2A have limited generation quality in terms of temporal synchronization and audio-visual relevance. T... | closed | completed | false | 3 | [
"New pipeline/model"
] | [] | 2023-11-11T02:00:31Z | 2024-11-25T13:06:46Z | 2024-11-25T13:06:46Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | clarencechen | 8,482,341 | MDQ6VXNlcjg0ODIzNDE= | User | false |
huggingface/diffusers | 1,988,705,181 | I_kwDOHa8MBc52iTud | 5,761 | https://github.com/huggingface/diffusers/issues/5761 | https://api.github.com/repos/huggingface/diffusers/issues/5761 | The cost of consistency decoder | ### Describe the bug
I replace original VAE decoder of a stable diffusion model with Consistency Decoder, then CUDA out of memory occurs. My question is that How large of Consistency Decoder is compared to original VAE decoder.
- `diffusers` version: 0.23.0
- Platform: Linux-5.15.0-60-generic-x86_64-with-glibc2.35... | closed | completed | false | 3 | [
"question",
"stale"
] | [] | 2023-11-11T03:54:20Z | 2024-01-09T15:07:30Z | 2024-01-09T15:07:30Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luciennnnnnn | 20,135,317 | MDQ6VXNlcjIwMTM1MzE3 | User | false |
huggingface/diffusers | 1,988,853,530 | I_kwDOHa8MBc52i38a | 5,762 | https://github.com/huggingface/diffusers/issues/5762 | https://api.github.com/repos/huggingface/diffusers/issues/5762 | TemporalConvLayer would raise error when in_dim is not equal to out_dim | ### Describe the bug
In class `TemporalConvLayer`, `conv2`'s output dimension is `in_dim`, but `conv3`'s input dimension is `out_dim`. When `in_dim` is not equal to `out_dim`, this would raise following error (assume `in_dim=512` and `out_dim=256`):
RuntimeError: Expected weight to be a vector of size equal to the ... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-11-11T09:07:41Z | 2024-01-09T15:07:28Z | 2024-01-09T15:07:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CSU-NXY | 19,389,655 | MDQ6VXNlcjE5Mzg5NjU1 | User | false |
huggingface/diffusers | 1,988,999,456 | I_kwDOHa8MBc52jbkg | 5,763 | https://github.com/huggingface/diffusers/issues/5763 | https://api.github.com/repos/huggingface/diffusers/issues/5763 | pipeline.save_pretrained is not saving as safetensors - safe_serialization=True | I am trying to make a LoRA LCM merger gradio app.
**I guess there is no way to extract LCM LoRA from pre-trained model? Am I correct?**
It saves everything but as diffusers not as single safetensors file how to fix?
```
if huggingface:
print("Loading model from Hugging Face.")
pipeline ... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-11-11T15:25:52Z | 2024-01-09T15:07:27Z | 2024-01-09T15:07:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FurkanGozukara | 19,240,467 | MDQ6VXNlcjE5MjQwNDY3 | User | false |
huggingface/diffusers | 1,989,011,744 | I_kwDOHa8MBc52jekg | 5,764 | https://github.com/huggingface/diffusers/issues/5764 | https://api.github.com/repos/huggingface/diffusers/issues/5764 | . | null | closed | completed | false | 0 | [] | [] | 2023-11-11T15:52:55Z | 2023-11-11T16:28:44Z | 2023-11-11T15:53:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yushan777 | 123,392,471 | U_kgDOB1rR1w | User | false |
huggingface/diffusers | 1,989,104,859 | I_kwDOHa8MBc52j1Tb | 5,766 | https://github.com/huggingface/diffusers/issues/5766 | https://api.github.com/repos/huggingface/diffusers/issues/5766 | Image+Image+Text to Image | Maybe a dumb question but I can't seem to find good ways to have multiple images to image modeling. I looked into Multi-ControlNet but I can't tell how to use it. I'm trying to train a model that takes in 2 images and a prompt:
1. a template base image (e.g. a photo of a room in someone's house with a painting on the... | closed | completed | false | 2 | [
"question",
"stale"
] | [] | 2023-11-11T20:15:27Z | 2024-01-09T15:07:25Z | 2024-01-09T15:07:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tval2 | 28,635,062 | MDQ6VXNlcjI4NjM1MDYy | User | false |
huggingface/diffusers | 1,989,231,895 | I_kwDOHa8MBc52kUUX | 5,769 | https://github.com/huggingface/diffusers/issues/5769 | https://api.github.com/repos/huggingface/diffusers/issues/5769 | Does the training code in diffusers implement correct mixed precision training? | ### Describe the bug
Hi, I notice many of example training code explicitly cast every tensor into float16, other than use `torch.autocast`. For example, this controlnet [code](https://github.com/huggingface/diffusers/blob/main/examples/controlnet/train_controlnet.py) convert every tensors into float16 if we use `--m... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-11-12T04:20:56Z | 2023-11-12T08:02:41Z | 2023-11-12T08:02:41Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luciennnnnnn | 20,135,317 | MDQ6VXNlcjIwMTM1MzE3 | User | false |
huggingface/diffusers | 1,989,267,163 | I_kwDOHa8MBc52kc7b | 5,770 | https://github.com/huggingface/diffusers/issues/5770 | https://api.github.com/repos/huggingface/diffusers/issues/5770 | Can not run LCM distill pipeline, due to dataset access | ### Describe the bug
Can not run LCM distillation example script due to aws dataset access
### Reproduction
runwayml/stable-diffusion-v1-5
PROGRAM="train_lcm_distill_lora_sd_wds.py \
--pretrained_teacher_model=$MODEL_DIR \
--output_dir=$OUTPUT_DIR \
--mixed_precision=fp16 \
--resolution=512 \
... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-11-12T07:02:43Z | 2023-12-06T09:36:59Z | 2023-12-06T09:36:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lonestar1990 | 146,791,234 | U_kgDOCL_bQg | User | false |
huggingface/diffusers | 1,989,307,426 | I_kwDOHa8MBc52kmwi | 5,772 | https://github.com/huggingface/diffusers/issues/5772 | https://api.github.com/repos/huggingface/diffusers/issues/5772 | Does webdataset faster than default huggingface datasets? | ### Describe the bug
Hi, I see there is a large scale training example https://github.com/huggingface/diffusers/blob/controlnet_webdatasets/examples/controlnet/train_controlnet_webdatasets.py using webdatasets, which suggests that webdatasets may have better data loading performance than huggingface datasets that is o... | closed | completed | false | 2 | [
"question",
"stale"
] | [] | 2023-11-12T08:40:22Z | 2024-01-09T15:07:23Z | 2024-01-09T15:07:23Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luciennnnnnn | 20,135,317 | MDQ6VXNlcjIwMTM1MzE3 | User | false |
huggingface/diffusers | 1,989,350,108 | I_kwDOHa8MBc52kxLc | 5,773 | https://github.com/huggingface/diffusers/issues/5773 | https://api.github.com/repos/huggingface/diffusers/issues/5773 | About guidance_scale in LCM-LoRA | This is a replicated issue from https://github.com/luosiallen/latent-consistency-model/issues/40.
As suggested in [latent-consistency/lcm-lora-sdv1-5](https://huggingface.co/latent-consistency/lcm-lora-sdv1-5), it is recommended to disable guidance_scale or use values between 1.0 and 2.0. The effect of guidance_scal... | closed | completed | false | 2 | [] | [] | 2023-11-12T10:55:17Z | 2023-11-18T11:03:36Z | 2023-11-12T15:33:07Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | haofanwang | 18,741,068 | MDQ6VXNlcjE4NzQxMDY4 | User | false |
huggingface/diffusers | 1,989,437,729 | I_kwDOHa8MBc52lGkh | 5,774 | https://github.com/huggingface/diffusers/issues/5774 | https://api.github.com/repos/huggingface/diffusers/issues/5774 | How to fine tune Stable Diffusion on custom dataset {caption, image}? | I need to do the task that fine tuning SD on custom dataset {caption, image} and custom size? Could you please give me a tutorial for this task? | closed | completed | false | 2 | [
"stale"
] | [] | 2023-11-12T14:52:23Z | 2024-01-09T15:07:21Z | 2024-01-09T15:07:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | npk7264 | 90,046,327 | MDQ6VXNlcjkwMDQ2MzI3 | User | false |
huggingface/diffusers | 1,989,544,753 | I_kwDOHa8MBc52lgsx | 5,776 | https://github.com/huggingface/diffusers/issues/5776 | https://api.github.com/repos/huggingface/diffusers/issues/5776 | remove the unneeded lines in get_processor | ### Describe the bug
can't not access `Unet.attn_processor` after calling `load_lora_weights()`, will get an attribute error
```
AttributeError: 'UNet2DConditionModel' object has no attribute 'Attention.attn_processor'. Did you mean: 'attn_processors'?
```
tracked this all the way down to an exception inside... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-11-12T19:56:09Z | 2023-11-17T20:36:32Z | 2023-11-17T20:36:31Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 1,989,782,647 | I_kwDOHa8MBc52max3 | 5,777 | https://github.com/huggingface/diffusers/issues/5777 | https://api.github.com/repos/huggingface/diffusers/issues/5777 | training pixart by dreamblooth error | ### Describe the bug
I use diffusers pipeline training(https://huggingface.co/docs/diffusers/training/dreambooth) and use pixart as the base model(https://huggingface.co/PixArt-alpha/PixArt-XL-2-512x512), but an error will be reported.
image
![image](https://github.com/huggingface/diffusers/assets/54690997/cd99df00-... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-11-13T02:44:32Z | 2024-01-09T15:07:19Z | 2024-01-09T15:07:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wen020 | 54,690,997 | MDQ6VXNlcjU0NjkwOTk3 | User | false |
huggingface/diffusers | 1,990,738,391 | I_kwDOHa8MBc52qEHX | 5,781 | https://github.com/huggingface/diffusers/issues/5781 | https://api.github.com/repos/huggingface/diffusers/issues/5781 | TypeError: download_from_original_stable_diffusion_ckpt() got an unexpected keyword argument 'text_encoder' | ### Describe the bug
It happened when I'm using Stable-Diffusion to draw a picture(I'm using a script [Accelerate with OpenVINO] to make it can use Intel GPU[Xe] to make it faster). But I don't know why it can't work normally, and show me the error info：TypeError: download_from_original_stable_diffusion_ckpt() got an ... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-11-13T14:09:53Z | 2024-01-09T15:07:17Z | 2024-01-09T15:07:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LYR-GH | 140,906,768 | U_kgDOCGYREA | User | false |
huggingface/diffusers | 1,991,143,394 | I_kwDOHa8MBc52rm_i | 5,783 | https://github.com/huggingface/diffusers/issues/5783 | https://api.github.com/repos/huggingface/diffusers/issues/5783 | train_text_to_image_sdxl.py ema not working | ### Describe the bug
I did a few training runs with train_text_to_image_sdxl.py before realizing that the EMA checkpoint never changed.
Looking at train_text_to_image.py - it seems that there is no ema_unet.step() in the sdxl training code so the ema model is never updated.
it's also missing this bit at the s... | closed | completed | false | 9 | [
"bug"
] | [] | 2023-11-13T17:42:54Z | 2024-02-26T20:39:58Z | 2024-02-26T20:39:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jack000 | 2,636,509 | MDQ6VXNlcjI2MzY1MDk= | User | false |
huggingface/diffusers | 1,991,862,946 | I_kwDOHa8MBc52uWqi | 5,786 | https://github.com/huggingface/diffusers/issues/5786 | https://api.github.com/repos/huggingface/diffusers/issues/5786 | How to load a precomputed dataset in the cache folder on a different machine? | **Is your feature request related to a problem? Please describe.**
Some slurm cluster may have a limit on time allocation, so I'd like to precompute the dataset on my local machine then move it to a location on the cluster to directly reuse it.
**Describe the solution you'd like**
I saw load dataset automatica... | closed | completed | false | 3 | [
"question",
"stale"
] | [] | 2023-11-14T02:26:00Z | 2024-01-09T15:07:14Z | 2024-01-09T15:07:14Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | linnanwang | 4,675,611 | MDQ6VXNlcjQ2NzU2MTE= | User | false |
huggingface/diffusers | 1,378,529,000 | I_kwDOHa8MBc5SKq7o | 579 | https://github.com/huggingface/diffusers/issues/579 | https://api.github.com/repos/huggingface/diffusers/issues/579 | Textual Inversion got error of cannot find config.json file | ### Describe the bug
When I follow every step described [here](https://github.com/huggingface/diffusers/tree/main/examples/textual_inversion), I got the following error:
```
OSError: CompVis/stable-diffusion-v1-4 does not appear to have a file named config.json.
```
I can use inference Pipeline with no issue. It... | closed | completed | false | 7 | [
"bug"
] | [
"patil-suraj"
] | 2022-09-19T21:12:52Z | 2023-03-26T05:39:29Z | 2022-09-29T09:16:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | XavierXiao | 14,303,231 | MDQ6VXNlcjE0MzAzMjMx | User | false |
huggingface/diffusers | 1,992,604,348 | I_kwDOHa8MBc52xLq8 | 5,791 | https://github.com/huggingface/diffusers/issues/5791 | https://api.github.com/repos/huggingface/diffusers/issues/5791 | [May be a bug] Train dreambooth with prior reservation and snr gamma two parameters together | ### Describe the bug
[May be a bug] Train dream with prior reservation and snr gamma **two parameters together.**
Check the shape of the Tensor:
```python
loss = F.mse_loss(model_pred.float(), target.float(), reduction="none")
logger.info(f"2: {loss.shape}, {mse_loss_weights.shape}")
logger.info(f"3: {snr.shape... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2023-11-14T12:01:57Z | 2024-11-25T13:06:20Z | 2024-11-25T13:06:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Pursue26 | 49,673,141 | MDQ6VXNlcjQ5NjczMTQx | User | false |
huggingface/diffusers | 1,992,640,458 | I_kwDOHa8MBc52xUfK | 5,793 | https://github.com/huggingface/diffusers/issues/5793 | https://api.github.com/repos/huggingface/diffusers/issues/5793 | Why does unload_lora_weights take 14 seconds in PEFT mode? | my code:
import time
pipe.disable_lora()
print('disable_lora')
ts = time.time()
pipe.unload_lora_weights()
print(f'unload_lora_weights {time.time() - ts:.3f}s')
unload_lora_weights 14.214s
My platform:
- GPU 3090
- diffusers 0.23.0.dev0
- peft 0.6.0.dev
- transformers 4.34.1
- Ubuntu 22.04.3 LT... | closed | completed | false | 9 | [] | [] | 2023-11-14T12:25:17Z | 2023-11-27T07:49:33Z | 2023-11-27T07:49:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chinatian | 3,233,111 | MDQ6VXNlcjMyMzMxMTE= | User | false |
huggingface/diffusers | 1,992,796,671 | I_kwDOHa8MBc52x6n_ | 5,794 | https://github.com/huggingface/diffusers/issues/5794 | https://api.github.com/repos/huggingface/diffusers/issues/5794 | There is no “tokenizer/config.json” on stabilityai/stable-diffusion-xl-base-1.0 | ### Describe the bug
The error log is “There is no “tokenizer/config.json” on stabilityai/stable-diffusion-xl-base-1.0”，but there is no this file in huggingface.( https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0)
### Reproduction
python train_controlnet_sdxl.py --pretrained_model_name_or_path=$MODEL_... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-11-14T13:56:24Z | 2024-01-19T15:05:52Z | 2024-01-19T15:05:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chenyang23333 | 55,976,303 | MDQ6VXNlcjU1OTc2MzAz | User | false |
huggingface/diffusers | 1,993,946,405 | I_kwDOHa8MBc522TUl | 5,797 | https://github.com/huggingface/diffusers/issues/5797 | https://api.github.com/repos/huggingface/diffusers/issues/5797 | training and valid loss dont converge during sdxl text2image lora training | I'm using the train_text_to_image_lora_sdxl.py for lora training on sdxl for specific art style with following args:
accelerate launch $PWD/diffusers/examples/text_to_image/lora_with_valid.py \
--pretrained_model_name_or_path='../sd_models/stabilityai__stable-diffusion-xl-base-1.0/' \
--pretrained_vae_model_na... | closed | not_planned | false | 3 | [] | [] | 2023-11-15T03:06:15Z | 2023-11-24T05:02:00Z | 2023-11-24T05:02:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DoctorTar | 147,801,017 | U_kgDOCM9DuQ | User | false |
huggingface/diffusers | 1,993,960,264 | I_kwDOHa8MBc522WtI | 5,798 | https://github.com/huggingface/diffusers/issues/5798 | https://api.github.com/repos/huggingface/diffusers/issues/5798 | StableDiffusionControlNetPipeline triggers unexpected keyword argument 'callback_on_step_end' | ### Describe the bug
When calling `StableDiffusionControlNetPipeline` with the new `callback_on_step_end`, it triggers the error message:
> TypeError: StableDiffusionControlNetPipeline.__call__() got an unexpected keyword argument 'callback_on_step_end'
### Reproduction
```py
import os
import torch
from ... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-11-15T03:20:13Z | 2024-01-09T15:07:06Z | 2024-01-09T15:07:06Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexisrolland | 13,064,696 | MDQ6VXNlcjEzMDY0Njk2 | User | false |
huggingface/diffusers | 1,994,030,733 | I_kwDOHa8MBc522n6N | 5,799 | https://github.com/huggingface/diffusers/issues/5799 | https://api.github.com/repos/huggingface/diffusers/issues/5799 | keyword argument for "strength" on inpainting | ### Describe the bug
I'm simply playing with the inpainting demo from the colab on the repo and added one line to the pipe keyword args for "strength":
```python
prompt = "a mecha robot sitting on a bench"
guidance_scale=7.5
num_samples = 3
generator = torch.Generator(device="cuda").manual_seed(0) # change ... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-11-15T04:43:15Z | 2024-01-19T15:05:51Z | 2024-01-19T15:05:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tval2 | 28,635,062 | MDQ6VXNlcjI4NjM1MDYy | User | false |
huggingface/diffusers | 1,290,727,496 | I_kwDOHa8MBc5M7vBI | 58 | https://github.com/huggingface/diffusers/issues/58 | https://api.github.com/repos/huggingface/diffusers/issues/58 | glide training generates weird images | Hi,
Thanks for this amazing repository! I am currently training Glide using the script `train_glide_text_to_image.py`. However, the generated images do not make sense and no matter the text prompt, the generated images are all similar to each other.
Below, for the text "a woman walks her dog on the beach"
![Scre... | closed | completed | false | 5 | [
"stale"
] | [
"anton-l"
] | 2022-06-30T23:24:00Z | 2022-09-19T14:39:43Z | 2022-09-19T14:39:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xinmiaolin | 43,487,018 | MDQ6VXNlcjQzNDg3MDE4 | User | false |
huggingface/diffusers | 1,994,050,441 | I_kwDOHa8MBc522suJ | 5,800 | https://github.com/huggingface/diffusers/issues/5800 | https://api.github.com/repos/huggingface/diffusers/issues/5800 | [Control-LoRA] Picking Control-LoRA up again | Stability AI's Control LoRA has been around for a while: https://huggingface.co/stabilityai/control-lora.
We also have had requests for it, e.g., https://github.com/huggingface/diffusers/issues/4679.
I even started working on it: https://github.com/huggingface/diffusers/pull/4899, but I couldn't reproduce the r... | closed | completed | false | 16 | [
"Good second issue",
"contributions-welcome"
] | [] | 2023-11-15T05:00:45Z | 2025-12-21T16:21:01Z | 2025-12-21T16:21:01Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,994,165,203 | I_kwDOHa8MBc523IvT | 5,801 | https://github.com/huggingface/diffusers/issues/5801 | https://api.github.com/repos/huggingface/diffusers/issues/5801 | why I got a black generated image through the official train_controlnet.py? |
when training the official example "train_contorlnet.py", I visualized the output form pipeline, then I got:
I want to know the reason.
| closed | completed | false | 1 | [
"stale"
] | [] | 2023-11-15T06:58:52Z | 2024-01-10T15:06:39Z | 2024-01-10T15:06:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yangzhou321 | 35,005,293 | MDQ6VXNlcjM1MDA1Mjkz | User | false |
huggingface/diffusers | 1,994,299,403 | I_kwDOHa8MBc523pgL | 5,802 | https://github.com/huggingface/diffusers/issues/5802 | https://api.github.com/repos/huggingface/diffusers/issues/5802 | Extending callback functionality to pipeline components. | # Is your feature request related to a problem? Please describe
One implementation is Kohya's "Deep shrink" algorithm for increased resolution with limited cloning/collapsing, which requires modifications to the latents in the unet.
Here is an example with deep shrink:
## Cloning/collapsing on high resolutions
<i... | closed | completed | false | 20 | [
"stale"
] | [] | 2023-11-15T08:36:19Z | 2024-03-10T05:46:24Z | 2024-01-10T15:06:37Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TimothyAlexisVass | 55,708,319 | MDQ6VXNlcjU1NzA4MzE5 | User | false |
huggingface/diffusers | 1,994,472,100 | I_kwDOHa8MBc524Tqk | 5,803 | https://github.com/huggingface/diffusers/issues/5803 | https://api.github.com/repos/huggingface/diffusers/issues/5803 | How should I reproduce the process of generating black images using SD | ### Describe the bug
When I use SD for prediction in a production environment, I occasionally encounter black images. In order to observe which model is causing the problem, I try to record the input image, prompt, specify the generator, and record the corresponding seed, but it is not possible to reproduce the blac... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-11-15T10:18:50Z | 2024-01-10T15:06:34Z | 2024-01-10T15:06:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TheHonestBob | 58,240,629 | MDQ6VXNlcjU4MjQwNjI5 | User | false |
huggingface/diffusers | 1,994,545,582 | I_kwDOHa8MBc524lmu | 5,804 | https://github.com/huggingface/diffusers/issues/5804 | https://api.github.com/repos/huggingface/diffusers/issues/5804 | BlipDiffusionPipeline | ### Describe the bug
When using the blipdiffusion pipe
I downloaded the Salesforce/blipdiffusion file to the local and replaced it to the local path.
But the tonkenizer cannot be loaded accurately
This is no... | closed | completed | false | 9 | [
"bug"
] | [] | 2023-11-15T10:57:31Z | 2023-12-01T01:41:03Z | 2023-11-17T09:15:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fkjkey | 146,184,493 | U_kgDOCLaZLQ | User | false |
huggingface/diffusers | 1,994,661,940 | I_kwDOHa8MBc525CA0 | 5,805 | https://github.com/huggingface/diffusers/issues/5805 | https://api.github.com/repos/huggingface/diffusers/issues/5805 | [LCM-SDXL][XLA] RuntimeError: Input type (float) and bias type (c10::Half) should be the same | ### Describe the bug
Running sample code from https://huggingface.co/latent-consistency/lcm-sdxl with a little bit XLA adaption got error
The error remains the same regardless of using TPU or CPU as backend.
Details are in the colab.
I cannot figure out why the input type is float. Need some light.
### Repr... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-11-15T12:14:01Z | 2024-01-16T15:06:19Z | 2024-01-16T15:06:19Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kevint324 | 8,800,468 | MDQ6VXNlcjg4MDA0Njg= | User | false |
huggingface/diffusers | 1,995,131,768 | I_kwDOHa8MBc5260t4 | 5,808 | https://github.com/huggingface/diffusers/issues/5808 | https://api.github.com/repos/huggingface/diffusers/issues/5808 | Inpainting produces results that are uneven with input image | ### Describe the bug
SD inpainting works fine only if mask is absolutely perfect.
Otherwise, there are always visible seams at the edge of the mask, and uneven colors between inpainted and input image.
I've tried manually assembling latents and using them for `image` and `mask_image` instead of images as wel... | open | reopened | false | 43 | [
"bug",
"stale",
"inpainting"
] | [
"yiyixuxu"
] | 2023-11-15T16:33:02Z | 2024-09-14T15:21:05Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 1,995,369,595 | I_kwDOHa8MBc527ux7 | 5,811 | https://github.com/huggingface/diffusers/issues/5811 | https://api.github.com/repos/huggingface/diffusers/issues/5811 | [Callbacks] adding `callback_on_step_end` arguments to all pipelines | we recently introduced `callback_on_step_end` to our pipelines that provides more flexibility for user to customize the pipelines.
anyone interested in adding this argument to all the other pipelines that does not have this argument yet? e.g. contorlnet pipelines are pretty important and still do not have this arg... | closed | completed | false | 3 | [
"good first issue",
"contributions-welcome"
] | [] | 2023-11-15T19:11:03Z | 2023-11-21T14:22:21Z | 2023-11-21T14:22:21Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 1,996,137,525 | I_kwDOHa8MBc52-qQ1 | 5,815 | https://github.com/huggingface/diffusers/issues/5815 | https://api.github.com/repos/huggingface/diffusers/issues/5815 | LCM Scheduler isn't perfect in uniformly spreading the timesteps across the full range | ### Describe the bug
Timesteps for nSteps=26 using the current scheduler
> [999, 979, 959, 939, 919, 899, 879, 859, 839, 819, 799, 779, 759,
739, 719, 699, 679, 659, 639, 619, 599, 579, 559, 539, 519, 499]
Notice that it comes up short of coming near zero.
My modification is still linear but across the ful... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-11-16T06:13:29Z | 2023-11-20T14:46:12Z | 2023-11-20T14:46:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aifartist | 116,415,616 | U_kgDOBvBcgA | User | false |
huggingface/diffusers | 1,996,228,949 | I_kwDOHa8MBc52_AlV | 5,816 | https://github.com/huggingface/diffusers/issues/5816 | https://api.github.com/repos/huggingface/diffusers/issues/5816 | low attention to prompt in SDXL | Hi,
One of the difference between DALLE3 and SDXL is that SDXL pay less attention to prompt,
Is there a way to solve this problem? I don't Know. for example changing the text encoder to other can help to solve this problem ?
Thanks
| closed | completed | false | 4 | [
"question",
"stale"
] | [] | 2023-11-16T07:24:15Z | 2024-01-09T15:06:55Z | 2024-01-09T15:06:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | saeedkhanehgir | 65,589,645 | MDQ6VXNlcjY1NTg5NjQ1 | User | false |
huggingface/diffusers | 1,996,241,740 | I_kwDOHa8MBc52_DtM | 5,817 | https://github.com/huggingface/diffusers/issues/5817 | https://api.github.com/repos/huggingface/diffusers/issues/5817 | Activation function and Fourier process | ### Model/Pipeline/Scheduler description
Hi, i tried to add an activation function to GELU, before i even started , i keep getting import use_PEFT_Backend error, even though i already have utils installed.
it says attempted relative import with no known parent package.
If i change the activation function, the d... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-11-16T07:31:45Z | 2024-01-09T15:06:53Z | 2024-01-09T15:06:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pow9791 | 58,098,952 | MDQ6VXNlcjU4MDk4OTUy | User | false |
huggingface/diffusers | 1,379,025,954 | I_kwDOHa8MBc5SMkQi | 582 | https://github.com/huggingface/diffusers/issues/582 | https://api.github.com/repos/huggingface/diffusers/issues/582 | LMSDiscreteScheduler support for Inpaint Pipeline | Currently there is no support for the LMSDiscreteScheduler in the StableDiffusionInpaintPipeline | closed | completed | false | 1 | [] | [] | 2022-09-20T08:31:00Z | 2022-09-20T17:10:47Z | 2022-09-20T17:10:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CristoJV | 44,086,220 | MDQ6VXNlcjQ0MDg2MjIw | User | false |
huggingface/diffusers | 1,996,718,539 | I_kwDOHa8MBc53A4HL | 5,821 | https://github.com/huggingface/diffusers/issues/5821 | https://api.github.com/repos/huggingface/diffusers/issues/5821 | Can't train in cuda by pytorch's `backward`. | ### Describe the bug
To transfer the pre-trained model to my personal research project that uses the diffusion model to fusion emoji, I used `UNet2DConditionModel` as my denoising model. I didn't use accelerate due to complexity, and I used pytorch to train the model.
My code works well on my Mac with mps, but the ... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-11-16T12:09:42Z | 2024-04-20T19:44:15Z | 2023-11-19T08:25:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ChengAoShen | 68,673,559 | MDQ6VXNlcjY4NjczNTU5 | User | false |
huggingface/diffusers | 1,996,945,569 | I_kwDOHa8MBc53Bvih | 5,824 | https://github.com/huggingface/diffusers/issues/5824 | https://api.github.com/repos/huggingface/diffusers/issues/5824 | Problematic default value for `text_encoder_projection_dim` in `get_add_time_ids()` method of `StableDiffusionXLPipeline` | ### Describe the bug
The `get_add_time_ids()` method of `StableDiffusionXLPipline` currently has the following defaults defined in line 573 of `stable_diffusion_xl/pipeline_stable_diffusion_xl.py`:
```
def _get_add_time_ids(
self, original_size, crops_coords_top_left, target_size, dtype, text_encoder_projec... | closed | completed | false | 5 | [
"bug",
"good first issue",
"contributions-welcome"
] | [] | 2023-11-16T14:09:35Z | 2024-03-11T01:07:35Z | 2024-03-11T01:07:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tyler-tomita | 9,794,449 | MDQ6VXNlcjk3OTQ0NDk= | User | false |
huggingface/diffusers | 1,997,726,327 | I_kwDOHa8MBc53EuJ3 | 5,829 | https://github.com/huggingface/diffusers/issues/5829 | https://api.github.com/repos/huggingface/diffusers/issues/5829 | LCM train scripts crash due to missing `unet_time_cond_proj_dim` argument | ### Describe the bug
There appears to be a set of issues related to `unet_time_cond_proj_dim` argument when running the LCM training scripts in `examples/consistency_distillation/`
**Running the script as-is**
First: the argument is not defined in `parse_args` and raises an error when you run the script with exa... | closed | completed | false | 5 | [
"bug"
] | [
"patil-suraj"
] | 2023-11-16T20:33:15Z | 2023-11-27T12:00:42Z | 2023-11-27T12:00:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | justindujardin | 101,493 | MDQ6VXNlcjEwMTQ5Mw== | User | false |
huggingface/diffusers | 1,997,832,667 | I_kwDOHa8MBc53FIHb | 5,830 | https://github.com/huggingface/diffusers/issues/5830 | https://api.github.com/repos/huggingface/diffusers/issues/5830 | Probably bug with center crop / resize | ### Describe the bug
Diffusers apply resize before center crop and this lead to loss of most part of the picture. So if you train on not square dataset you lost parts of images during distillation
Affected code
Center crop/resize bug in:
wuerstchen/text_to_image
https://github.com/huggingface/diffusers/blo... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-11-16T21:37:44Z | 2023-11-16T22:11:07Z | 2023-11-16T22:11:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | recoilme | 417,177 | MDQ6VXNlcjQxNzE3Nw== | User | false |
huggingface/diffusers | 1,998,227,739 | I_kwDOHa8MBc53Gokb | 5,835 | https://github.com/huggingface/diffusers/issues/5835 | https://api.github.com/repos/huggingface/diffusers/issues/5835 | Using pretrained MusicLDM pipeline | ### Describe the bug
Hello, in diffusers docs, MusicLDM is used as follow:
```python
from diffusers import MusicLDMPipeline
import torch
import scipy
repo_id = "cvssp/audioldm-s-full-v2"
pipe = MusicLDMPipeline.from_pretrained(repo_id, torch_dtype=torch.float16)
pipe = pipe.to("cuda")
prompt = "Techno musi... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-11-17T03:54:17Z | 2023-11-21T14:27:43Z | 2023-11-21T14:27:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hoang1007 | 78,253,878 | MDQ6VXNlcjc4MjUzODc4 | User | false |
huggingface/diffusers | 1,998,422,275 | I_kwDOHa8MBc53HYED | 5,840 | https://github.com/huggingface/diffusers/issues/5840 | https://api.github.com/repos/huggingface/diffusers/issues/5840 | train_dreambooth_lora_sdxl.py cannot resume training from checkpoint ! ! model freezed ! ! | ### Describe the bug
When resume training from a middle lora checkpoint, it stops update the model( i.e. checkpionts remain the same as the middle checkpoint).
For reproducing the bug, just turn on the `--resume_from_checkpoint` flag.
All experimental settings are based on default configurations, using the latest v... | closed | completed | false | 11 | [
"bug",
"stale"
] | [] | 2023-11-17T07:10:35Z | 2024-01-03T15:06:37Z | 2024-01-03T15:06:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yuxu915 | 94,051,337 | U_kgDOBZscCQ | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.