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 | 2,096,555,815 | I_kwDOHa8MBc589ucn | 6,685 | https://github.com/huggingface/diffusers/issues/6685 | https://api.github.com/repos/huggingface/diffusers/issues/6685 | [Advanced Dreambooth] Absolute paths for output_dir are not supported | ### Describe the bug
Saving the model files currently prepends the `args.output_dir` to the file name to various files:
`{args.output_dir}/{args.output_dir}_emb.safetensors`
`{args.output_dir}/{args.output_dir}.safetensors`
This breaks when using an absolute directory path for output_dir.
### Reproduction
Run... | closed | completed | false | 7 | [
"bug",
"stale",
"training"
] | [] | 2024-01-23T17:01:51Z | 2025-01-27T02:01:50Z | 2025-01-27T02:01:49Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | levi | 8,493 | MDQ6VXNlcjg0OTM= | User | false |
huggingface/diffusers | 2,097,133,758 | I_kwDOHa8MBc58_7i- | 6,686 | https://github.com/huggingface/diffusers/issues/6686 | https://api.github.com/repos/huggingface/diffusers/issues/6686 | Dynamo compilation of SDXL, SDXL Refiner, SDXL ControlNet have no performance improvements after save/load | ### Describe the bug
Dynamo compilation of SDXL, SDXL Refiner, SDXL ControlNet have no performance improvements if you
1) Compile
2) Save pretrained
3) Load pretrained
### Reproduction
SDXL:
```
import torch
from diffusers import StableDiffusionXLPipeline
txt2img_pipeline = StableDiffusionXLPipeline.fro... | closed | not_planned | false | 9 | [
"bug",
"stale"
] | [] | 2024-01-23T23:07:31Z | 2024-03-01T21:46:28Z | 2024-03-01T21:46:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Alexadar | 14,125,937 | MDQ6VXNlcjE0MTI1OTM3 | User | false |
huggingface/diffusers | 2,097,425,461 | I_kwDOHa8MBc59BCw1 | 6,688 | https://github.com/huggingface/diffusers/issues/6688 | https://api.github.com/repos/huggingface/diffusers/issues/6688 | Add MotionCntrl | ### Model/Pipeline/Scheduler description
The [MotionCntrl](https://github.com/TencentARC/MotionCtrl) model seems like a promising way to control movement in video generation models. According to the project page, it can work with both SVD and AnimateDiff.
It would be nice to get an idea of how we might be able to ... | closed | completed | false | 7 | [
"New pipeline/model",
"contributions-welcome"
] | [] | 2024-01-24T04:52:29Z | 2024-09-17T23:37:34Z | 2024-09-17T23:37:34Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DN6 | 7,529,846 | MDQ6VXNlcjc1Mjk4NDY= | User | false |
huggingface/diffusers | 2,097,555,694 | I_kwDOHa8MBc59Biju | 6,689 | https://github.com/huggingface/diffusers/issues/6689 | https://api.github.com/repos/huggingface/diffusers/issues/6689 | Unable to convert SSD-1B.safetensors → Diffusers | ### Describe the bug
Hi! I'm trying to convert SSD-1B model ([link](https://huggingface.co/segmind/SSD-1B/tree/main)) from SafeTensors to Diffusers
### Reproduction
`python convert_original_stable_diffusion_to_diffusers.py --checkpoint_path ./models/ssd_1b.safetensors --from_safetensors --device cpu --extract... | closed | completed | false | 3 | [
"bug",
"stale",
"conversion script"
] | [
"DN6"
] | 2024-01-24T06:55:29Z | 2024-11-22T01:42:46Z | 2024-11-22T01:42:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nosferatu500 | 9,116,892 | MDQ6VXNlcjkxMTY4OTI= | User | false |
huggingface/diffusers | 1,389,632,907 | I_kwDOHa8MBc5S1B2L | 669 | https://github.com/huggingface/diffusers/issues/669 | https://api.github.com/repos/huggingface/diffusers/issues/669 | Simplify CrossAttention to run on Apple Neural Engine | **Is your feature request related to a problem? Please describe.**
I'm trying to convert portions of unet into CoreML. However, CrossAttention fails to compile to the Apple Neural Engine.
**Describe the solution you'd like**
My best guess after a lot of experimentation on converting CrossAttention in many respects... | closed | completed | false | 6 | [
"stale"
] | [
"pcuenca"
] | 2022-09-28T16:22:01Z | 2022-11-20T15:23:24Z | 2022-11-20T15:23:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MatthewWaller | 5,520,521 | MDQ6VXNlcjU1MjA1MjE= | User | false |
huggingface/diffusers | 2,098,027,343 | I_kwDOHa8MBc59DVtP | 6,695 | https://github.com/huggingface/diffusers/issues/6695 | https://api.github.com/repos/huggingface/diffusers/issues/6695 | "Raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd) " happened when starting studing | ### Describe the bug
I run the train_text_to_image.py with the command described in the instruction. However, when the process is in the VAE encoding, it took a lot of time and raised the error (shown below).
01/24/2024 11:00:18 - INFO - __main__ - ***** Running training *****
01/24/2024 11:00:18 - INFO - __main__... | open | null | false | 7 | [
"bug",
"stale"
] | [] | 2024-01-24T11:08:35Z | 2026-05-21T23:30:27Z | null | NONE | null | 20260521T233511Z | 2026-05-21T23:35:11Z | dahui-y | 93,740,259 | U_kgDOBZZc4w | User | false |
huggingface/diffusers | 2,098,297,420 | I_kwDOHa8MBc59EXpM | 6,697 | https://github.com/huggingface/diffusers/issues/6697 | https://api.github.com/repos/huggingface/diffusers/issues/6697 | CUDA out of memory when training SDXL Lora | ### Describe the bug
SDXL Lora can't be trained on 24 GB gpu
### Reproduction
for reproduction
```
export MODEL_NAME=/home/imgen/models/SDXL/juggernautXL_v8Rundiffusion/
export OUTPUT_DIR=`pwd`/poke11-lora
export DATASET_NAME="lambdalabs/pokemon-blip-captions"
python3 -m pdb examples/text_to_image/tra... | closed | completed | false | 3 | [
"bug"
] | [] | 2024-01-24T13:42:42Z | 2024-01-25T11:09:38Z | 2024-01-25T11:09:38Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | noskill | 733,626 | MDQ6VXNlcjczMzYyNg== | User | false |
huggingface/diffusers | 2,098,586,281 | I_kwDOHa8MBc59FeKp | 6,699 | https://github.com/huggingface/diffusers/issues/6699 | https://api.github.com/repos/huggingface/diffusers/issues/6699 | Type hint for `callback_on_step_end` in pipeline `__call__` is incorrect | ### Describe the bug
As far as I can tell, the type hint `callback_on_step_end` for all pipelines (as seen in [this search](https://github.com/search?q=repo%3Ahuggingface%2Fdiffusers+%22callback_on_step_end%3A+Optional%22&type=code)) is incorrect.
Taking the SDXL pipeline as an example, the type hint in `__call__... | open | null | false | 5 | [
"bug",
"stale"
] | [
"a-r-r-o-w"
] | 2024-01-24T16:00:49Z | 2024-12-13T15:07:54Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | philpax | 707,827 | MDQ6VXNlcjcwNzgyNw== | User | false |
huggingface/diffusers | 2,098,732,543 | I_kwDOHa8MBc59GB3_ | 6,700 | https://github.com/huggingface/diffusers/issues/6700 | https://api.github.com/repos/huggingface/diffusers/issues/6700 | Attention processors that do not expect a `temb` parameter still get passed them and cause TypeErrors | ### Describe the bug
The parameterized attention is always passed a `temb` keyword argument, and many attention processors do not expect it, causing a type error.
A (currently incomplete) fix for this is #6687
### Reproduction
from diffusers import DiffusionPipeline
from diffusers.models.attention_processor imp... | closed | completed | false | 0 | [
"bug"
] | [] | 2024-01-24T17:16:12Z | 2024-02-10T07:56:21Z | 2024-02-10T07:56:21Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lsb | 45,281 | MDQ6VXNlcjQ1Mjgx | User | false |
huggingface/diffusers | 2,099,225,873 | I_kwDOHa8MBc59H6UR | 6,702 | https://github.com/huggingface/diffusers/issues/6702 | https://api.github.com/repos/huggingface/diffusers/issues/6702 | [Diffusion DPO] Loss function and reproducing results in the paper | Hi,
I got a question on the loss function used in both training scripts for diffusion DPO.
In train_diffusion_dpo.py line 863 and train_diffusion_dpo_sdxl.py line 978
` loss = -1 * F.logsigmoid(inside_term.mean())`
From my understanding, this mean is the average taken across the batches as the empirical mean ... | closed | completed | false | 5 | [] | [] | 2024-01-24T22:33:07Z | 2024-01-31T07:16:24Z | 2024-01-31T07:16:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yigu1008 | 119,135,690 | U_kgDOBxndyg | User | false |
huggingface/diffusers | 2,099,267,288 | I_kwDOHa8MBc59IEbY | 6,703 | https://github.com/huggingface/diffusers/issues/6703 | https://api.github.com/repos/huggingface/diffusers/issues/6703 | [Advanced Dreambooth] Textual Inversion token is weakened/ignored by negative prompt | ### Describe the bug
A LoRA + TI token pair trained with the advanced dreambooth script doesn't work with negative prompts.
In my testing, the validations all work perfectly if a negative prompt is not used. When it is included, the subjected trained with the token is often entirely ignored or looks nothing like ... | open | null | false | 9 | [
"bug",
"stale",
"training"
] | [] | 2024-01-24T23:03:33Z | 2025-01-24T15:08:08Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | levi | 8,493 | MDQ6VXNlcjg0OTM= | User | false |
huggingface/diffusers | 2,099,418,096 | I_kwDOHa8MBc59IpPw | 6,705 | https://github.com/huggingface/diffusers/issues/6705 | https://api.github.com/repos/huggingface/diffusers/issues/6705 | accelerate + FSDP + T2I train saving ckpt error | ### Describe the bug
I have used /examples/text_to_image/train_text_to_image_sdxl.py to train a fine tune sdxl. I used accelerate 0.25.0 + FSDP, when I was saving a checkpoint it will stuck and can't save a whole ckpt. And I also tried deepspeed it will stuck too. I didn't change any code in train_text_to_image_sdxl.... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2024-01-25T01:54:21Z | 2024-11-25T13:20:40Z | 2024-11-25T13:20:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Forainest | 85,024,612 | MDQ6VXNlcjg1MDI0NjEy | User | false |
huggingface/diffusers | 2,099,612,115 | I_kwDOHa8MBc59JYnT | 6,706 | https://github.com/huggingface/diffusers/issues/6706 | https://api.github.com/repos/huggingface/diffusers/issues/6706 | cads support | https://arxiv.org/abs/2310.17347 | open | null | false | 9 | [
"stale",
"contributions-welcome"
] | [] | 2024-01-25T05:43:25Z | 2024-12-22T15:08:20Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | loboere | 15,265,895 | MDQ6VXNlcjE1MjY1ODk1 | User | false |
huggingface/diffusers | 2,100,309,550 | I_kwDOHa8MBc59MC4u | 6,709 | https://github.com/huggingface/diffusers/issues/6709 | https://api.github.com/repos/huggingface/diffusers/issues/6709 | Bug in Custom Diffusion training with concept list | ### Describe the bug
**Context**: `train_custom_diffusion.py` script in Custom Diffusion using concept list
**Description**: This script fails if I use concept list and try to synthesize images. In this case it tries to generate synthetic images with wrong or missing prompt. Note, that the primary use of concept li... | closed | completed | false | 0 | [
"bug"
] | [] | 2024-01-25T12:38:47Z | 2024-01-26T03:17:53Z | 2024-01-26T03:17:53Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AIshutin | 30,271,345 | MDQ6VXNlcjMwMjcxMzQ1 | User | false |
huggingface/diffusers | 1,390,246,076 | I_kwDOHa8MBc5S3Xi8 | 671 | https://github.com/huggingface/diffusers/issues/671 | https://api.github.com/repos/huggingface/diffusers/issues/671 | converted models using convert_original_stable_diffusion_to_diffusers.py are not identical | ### Describe the bug
I have some ckpt files and I converted them using this script, while using it, there was a warning:
```
Some weights of the model checkpoint at openai/clip-vit-large-patch14 were not used when initializing CLIPTextModel: ['vision_model.encoder.layers.10.layer_norm1.weight', [..a lot more..] 'v... | closed | completed | false | 4 | [
"stale"
] | [
"patrickvonplaten"
] | 2022-09-29T04:43:24Z | 2022-11-06T15:03:35Z | 2022-11-06T15:03:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | NekliozOld | 57,619,097 | MDQ6VXNlcjU3NjE5MDk3 | User | false |
huggingface/diffusers | 2,100,472,851 | I_kwDOHa8MBc59MqwT | 6,711 | https://github.com/huggingface/diffusers/issues/6711 | https://api.github.com/repos/huggingface/diffusers/issues/6711 | Controlnet weight decay | I am working on creating a pipeline using a1111 or comfyui and implementing it with diffusers. In the controlnet mode of a1111, there is an option called 'my prompt is more important.' According to the internal implementation, this feature slowly decays from 1 by 0.85 at each denoising step. This greatly affects the qu... | closed | completed | false | 8 | [
"stale",
"needs-code-example"
] | [] | 2024-01-25T14:04:44Z | 2024-09-17T21:48:11Z | 2024-09-17T21:48:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | seokjin1013 | 63,408,412 | MDQ6VXNlcjYzNDA4NDEy | User | false |
huggingface/diffusers | 2,101,347,064 | I_kwDOHa8MBc59QAL4 | 6,713 | https://github.com/huggingface/diffusers/issues/6713 | https://api.github.com/repos/huggingface/diffusers/issues/6713 | `scripts/convert_original_stable_diffusion_to_diffusers.py` tells me `HeaderTooLarge`. | ### Describe the bug
I'm trying to download a CivitAI model to use with diffusers. I'm following this tutorial `https://medium.com/mlearning-ai/using-civitai-models-with-diffusers-package-45e0c475a67e`, but it's giving me the following error:
```
Traceback (most recent call last):
File "/home/ubuntu/notebooks/c... | closed | completed | false | 4 | [
"bug"
] | [] | 2024-01-25T23:48:04Z | 2024-11-19T09:24:20Z | 2024-11-19T09:24:20Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | landmann | 12,756,081 | MDQ6VXNlcjEyNzU2MDgx | User | false |
huggingface/diffusers | 1,390,482,008 | I_kwDOHa8MBc5S4RJY | 672 | https://github.com/huggingface/diffusers/issues/672 | https://api.github.com/repos/huggingface/diffusers/issues/672 | Convert `diffusers` model weights to the `original CompVis` ckpt format | There is a conversion script that converts the CompVis ckpt to diffusers available [here](https://github.com/huggingface/diffusers/blob/main/scripts/convert_original_stable_diffusion_to_diffusers.py), but the other way around does not exist yet.
As some Stable Diffusion UIs and utilities are built on top of the Comp... | closed | completed | false | 21 | [
"enhancement",
"good first issue"
] | [] | 2022-09-29T08:38:09Z | 2023-01-01T14:40:30Z | 2022-10-04T20:23:55Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | apolinario | 788,417 | MDQ6VXNlcjc4ODQxNw== | User | false |
huggingface/diffusers | 2,102,082,081 | I_kwDOHa8MBc59Szoh | 6,720 | https://github.com/huggingface/diffusers/issues/6720 | https://api.github.com/repos/huggingface/diffusers/issues/6720 | Maybe low_cpu_mem_usage need to be add as a parameter in load_lora_weight of StableDiffusionXLLoraLoaderMixin | **What API design would you like to have changed or added to the library? Why?**
l guess maybe the param low_cpu_mem_usage need to be added in load_lora_weight method of StableDiffusionXLLoraLoaderMixin.
**What use case would this enable or better enable? Can you give us a code example?**
In my pro. l come t... | closed | completed | false | 14 | [
"stale",
"peft"
] | [] | 2024-01-26T11:42:45Z | 2024-11-28T05:36:53Z | 2024-11-28T05:36:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Meloneat | 22,285,542 | MDQ6VXNlcjIyMjg1NTQy | User | false |
huggingface/diffusers | 2,102,501,237 | I_kwDOHa8MBc59UZ91 | 6,722 | https://github.com/huggingface/diffusers/issues/6722 | https://api.github.com/repos/huggingface/diffusers/issues/6722 | Using SigLIP as text encoder | **Is your feature request related to a problem? Please describe.**
I'm looking into testing SigLIP as an alternative text encoder. Is this supported or even possible to do? My knowledge is limited enough around this that I don't know if this makes sense to do.
There is some mention of this as an option in this note... | closed | completed | false | 2 | [] | [] | 2024-01-26T16:28:03Z | 2024-01-29T13:34:07Z | 2024-01-29T13:34:06Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | holwech | 7,474,295 | MDQ6VXNlcjc0NzQyOTU= | User | false |
huggingface/diffusers | 2,102,588,877 | I_kwDOHa8MBc59UvXN | 6,725 | https://github.com/huggingface/diffusers/issues/6725 | https://api.github.com/repos/huggingface/diffusers/issues/6725 | Generate images with `LDMPipeline` | ### Describe the bug
After following the docstring of `LDMPipeline`, it generated some weird images.
Is there anything I should do before sampling images?
### Reproduction
```py
from diffusers import ... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2024-01-26T17:29:08Z | 2024-02-28T08:51:25Z | 2024-02-26T16:56:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | david20571015 | 51,911,434 | MDQ6VXNlcjUxOTExNDM0 | User | false |
huggingface/diffusers | 2,103,109,422 | I_kwDOHa8MBc59Wucu | 6,726 | https://github.com/huggingface/diffusers/issues/6726 | https://api.github.com/repos/huggingface/diffusers/issues/6726 | CCSR: Improving the Stability of Diffusion Models for Content Consistent Super-Resolution | ### Model/Pipeline/Scheduler description
The generative priors of pre-trained latent diffusion models have demonstrated great potential to enhance the perceptual quality of image super-resolution (SR) results. Unfortunately, the existing diffusion prior-based SR methods encounter a common problem, i.e., they tend to g... | closed | completed | false | 5 | [
"stale",
"contributions-welcome"
] | [] | 2024-01-27T00:57:26Z | 2024-11-28T05:35:52Z | 2024-11-28T05:35:52Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | clarencechen | 8,482,341 | MDQ6VXNlcjg0ODIzNDE= | User | false |
huggingface/diffusers | 2,103,171,355 | I_kwDOHa8MBc59W9kb | 6,727 | https://github.com/huggingface/diffusers/issues/6727 | https://api.github.com/repos/huggingface/diffusers/issues/6727 | ValueError: Query/Key/Value should either all have the same dtype, or (in the quantized case) Key/Value should have dtype torch.int32 | ### Describe the bug
When I enable xformers_memory_efficient_attention, I encounter following issues, note that I use `with torch.autocast("cuda", enabled=True)` and when I disable xformers_memory_efficient_attention everything is fine.
ValueError: Query/Key/Value should either all have the same dtype, or (in the q... | closed | completed | false | 3 | [
"bug",
"stale",
"needs-code-example"
] | [] | 2024-01-27T03:36:11Z | 2024-03-13T15:07:16Z | 2024-03-13T15:07:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luciennnnnnn | 20,135,317 | MDQ6VXNlcjIwMTM1MzE3 | User | false |
huggingface/diffusers | 1,390,784,176 | I_kwDOHa8MBc5S5a6w | 673 | https://github.com/huggingface/diffusers/issues/673 | https://api.github.com/repos/huggingface/diffusers/issues/673 | (Textual Inversion) Initialise Vector For New Token From Multiple Existing Tokens | I'd like to propose an idea analagous to https://github.com/huggingface/diffusers/issues/369.
The current fine tuning script for textual inversion initialises the new `placeholder_token`'s embedding with an existing `initializer_token` (and enforces that the init token is exactly one token).
https://github.com/hu... | closed | completed | false | 6 | [
"stale"
] | [] | 2022-09-29T12:25:51Z | 2023-04-20T10:36:57Z | 2022-11-06T15:03:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rsomani95 | 17,111,474 | MDQ6VXNlcjE3MTExNDc0 | User | false |
huggingface/diffusers | 2,103,637,787 | I_kwDOHa8MBc59Yvcb | 6,731 | https://github.com/huggingface/diffusers/issues/6731 | https://api.github.com/repos/huggingface/diffusers/issues/6731 | ValueError: Unhandled pipeline class: SDXLLongPromptWeightingPipeline | ### Describe the bug
When running such code:
```python
pipeline = SDXLLongPromptWeightingPipeline.from_single_file(pretrained_model,
use_safetensors=True,
load_safety_checker=requ... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2024-01-27T15:37:04Z | 2024-11-22T01:40:17Z | 2024-11-22T01:40:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | calmiLovesAI | 29,011,011 | MDQ6VXNlcjI5MDExMDEx | User | false |
huggingface/diffusers | 2,103,637,816 | I_kwDOHa8MBc59Yvc4 | 6,732 | https://github.com/huggingface/diffusers/issues/6732 | https://api.github.com/repos/huggingface/diffusers/issues/6732 | trained_betas param is not documented in DDPMScheduler class | ### Describe the bug
trained_betas param is not documented in DDPMScheduler. Docstring should be updated here:
https://github.com/huggingface/diffusers/blob/main/src/diffusers/schedulers/scheduling_ddpm.py#L145
### Reproduction
see https://github.com/huggingface/diffusers/blob/main/src/diffusers/schedulers/schedu... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-01-27T15:37:11Z | 2024-01-31T16:53:41Z | 2024-01-31T16:53:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | agramfort | 161,052 | MDQ6VXNlcjE2MTA1Mg== | User | false |
huggingface/diffusers | 2,103,647,431 | I_kwDOHa8MBc59YxzH | 6,733 | https://github.com/huggingface/diffusers/issues/6733 | https://api.github.com/repos/huggingface/diffusers/issues/6733 | AttributeError: 'StableDiffusionPipeline' object has no attribute 'load_ip_adapter' | This error is reported when running the pipeline. Does anyone know how to solve it? | closed | completed | false | 3 | [] | [] | 2024-01-27T16:03:16Z | 2024-02-08T21:05:16Z | 2024-02-08T21:05:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Lhqy1111 | 129,830,835 | U_kgDOB70Psw | User | false |
huggingface/diffusers | 2,103,692,323 | I_kwDOHa8MBc59Y8wj | 6,734 | https://github.com/huggingface/diffusers/issues/6734 | https://api.github.com/repos/huggingface/diffusers/issues/6734 | 'NoneType' object has no attribute 'tokenize' | I try to run sd.15 but it gives an error AttributeError: 'NoneType' object has no attribute 'tokenize'
diffusers Version: 0.25.1
```
from diffusers import DiffusionPipeline, StableDiffusionXLImg2ImgPipeline
import torch
pipe = StableDiffusionXLImg2ImgPipeline.from_pretrained(
'runwayml/stable-diffusio... | closed | completed | false | 2 | [] | [] | 2024-01-27T17:39:01Z | 2024-02-08T21:04:58Z | 2024-02-08T21:04:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | loboere | 15,265,895 | MDQ6VXNlcjE1MjY1ODk1 | User | false |
huggingface/diffusers | 2,104,045,823 | I_kwDOHa8MBc59aTD_ | 6,739 | https://github.com/huggingface/diffusers/issues/6739 | https://api.github.com/repos/huggingface/diffusers/issues/6739 | how to generate images based on the text token embedding outputted from CLIP. token_embedding module? | how to generate images based on the text token embedding outputted from CLIP. token_embedding module? | closed | completed | false | 2 | [
"stale",
"should-move-to-discussion"
] | [] | 2024-01-28T08:51:45Z | 2024-11-19T09:27:00Z | 2024-11-19T09:26:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FlyGreyWolf | 39,620,354 | MDQ6VXNlcjM5NjIwMzU0 | User | false |
huggingface/diffusers | 1,390,808,338 | I_kwDOHa8MBc5S5g0S | 674 | https://github.com/huggingface/diffusers/issues/674 | https://api.github.com/repos/huggingface/diffusers/issues/674 | Can pretrained autoencoder stage be used for any new dataset training? | In the latent diffusion paper authors mention that:
"A notable advantage of this approach is that we need to train the universal autoencoding stage only once and can therefore reuse it for multiple DM trainings or to explore possibly completely different tasks."
Does that mean that we do not have to retrain autoencod... | closed | completed | false | 5 | [
"stale"
] | [] | 2022-09-29T12:43:48Z | 2022-11-10T15:03:41Z | 2022-11-10T15:03:41Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Revist | 26,026,957 | MDQ6VXNlcjI2MDI2OTU3 | User | false |
huggingface/diffusers | 2,104,129,006 | I_kwDOHa8MBc59anXu | 6,741 | https://github.com/huggingface/diffusers/issues/6741 | https://api.github.com/repos/huggingface/diffusers/issues/6741 | [Diffusion DPO] train script: 'Namespace' object has no attribute 'no_flip' | ### Describe the bug
I think there was a typo, because
`no_flip not in args`, only no_hflip
### Reproduction
```
!accelerate launch train_diffusion_dpo_sdxl.py
--pretrained_model_name_or_path="/home/ubuntu/mydrive/auto/stable-diffusion-webui/models/Stable-diffusion/ModelStep3"
--pretrained_vae_model_name... | closed | completed | false | 1 | [
"bug"
] | [] | 2024-01-28T12:00:24Z | 2024-01-28T12:12:36Z | 2024-01-28T12:12:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kodinkod | 69,761,539 | MDQ6VXNlcjY5NzYxNTM5 | User | false |
huggingface/diffusers | 2,104,132,312 | I_kwDOHa8MBc59aoLY | 6,742 | https://github.com/huggingface/diffusers/issues/6742 | https://api.github.com/repos/huggingface/diffusers/issues/6742 | DownBlockMotion performs different with or without checkpointing | ### Describe the bug
DownBlockMotion, UpBlockMotion can be normally called without gradient_checkpointing but cant with it.
something seems weird here:
https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/unets/unet_3d_blocks.py
line 1034~1043
```python
if self.training and self.gra... | closed | completed | false | 4 | [
"bug"
] | [
"DN6",
"yiyixuxu"
] | 2024-01-28T12:08:41Z | 2024-02-05T02:34:03Z | 2024-02-05T02:34:03Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | BaaBaaGoat | 38,155,995 | MDQ6VXNlcjM4MTU1OTk1 | User | false |
huggingface/diffusers | 2,105,447,478 | I_kwDOHa8MBc59fpQ2 | 6,749 | https://github.com/huggingface/diffusers/issues/6749 | https://api.github.com/repos/huggingface/diffusers/issues/6749 | Unable to replicate results using the InstantID community example | ### Describe the bug
I'm attempting to replicate the results from the [InstantID space](https://huggingface.co/spaces/InstantX/InstantID) using the [InstantID pipeline community example](https://github.com/huggingface/diffusers/blob/main/examples/community/pipeline_stable_diffusion_xl_instantid.py). However, the resul... | closed | completed | false | 3 | [
"bug"
] | [] | 2024-01-29T12:59:09Z | 2024-01-30T19:56:44Z | 2024-01-30T19:56:44Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | soumik12345 | 19,887,676 | MDQ6VXNlcjE5ODg3Njc2 | User | false |
huggingface/diffusers | 1,390,902,413 | I_kwDOHa8MBc5S53yN | 675 | https://github.com/huggingface/diffusers/issues/675 | https://api.github.com/repos/huggingface/diffusers/issues/675 | LDM Bert `config.json` path | ### Describe the bug
### Problem
There is a reference to an LDM Bert that 404's
```bash
src/diffusers/pipelines/latent_diffusion/pipeline_latent_diffusion.py: "ldm-bert": "https://huggingface.co/ldm-bert/resolve/main/config.json",
```
I was able to locate a `config.json` at `https://huggingface.co/valhalla/l... | closed | completed | false | 1 | [
"bug"
] | [
"patil-suraj"
] | 2022-09-29T13:44:22Z | 2022-09-30T09:52:16Z | 2022-09-30T09:52:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ryanrussell | 523,300 | MDQ6VXNlcjUyMzMwMA== | User | false |
huggingface/diffusers | 2,105,482,762 | I_kwDOHa8MBc59fx4K | 6,750 | https://github.com/huggingface/diffusers/issues/6750 | https://api.github.com/repos/huggingface/diffusers/issues/6750 | Enable batch processing from directory for multicontrolnets | **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 [...].
No working solution found to processing images from a directory
Error usually "TypeError: For multiple controlnets: `image` must be type `list`"
**De... | closed | completed | false | 2 | [] | [] | 2024-01-29T13:17:48Z | 2024-02-12T11:34:45Z | 2024-02-10T10:49:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | G-force78 | 114,336,644 | U_kgDOBtCjhA | User | false |
huggingface/diffusers | 2,106,179,075 | I_kwDOHa8MBc59ib4D | 6,753 | https://github.com/huggingface/diffusers/issues/6753 | https://api.github.com/repos/huggingface/diffusers/issues/6753 | VAE configuration issue | ### Describe the bug
It seems vae configuration might be broken in recent commits (~7f58a76f). Even though I can verify the correct config contents are read in I still end up with the wrong vae config. Most notably the scale_factor for SDXL models is that of SD1.5 resulting in washed out generations. I suspect the e... | closed | completed | false | 6 | [
"bug"
] | [
"DN6"
] | 2024-01-29T18:38:18Z | 2024-02-08T20:55:15Z | 2024-02-08T20:55:15Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | spezialspezial | 75,758,219 | MDQ6VXNlcjc1NzU4MjE5 | User | false |
huggingface/diffusers | 2,106,445,376 | I_kwDOHa8MBc59jc5A | 6,755 | https://github.com/huggingface/diffusers/issues/6755 | https://api.github.com/repos/huggingface/diffusers/issues/6755 | how to train a lora in inpainting model? | Is there a script to train Lora in SD 1.5 inpainting?
Is there any script to train Lora in SD 1.5 inpainting that works?
try this
https://github.com/huggingface/diffusers/tree/main/examples/research_projects/dreambooth_inpaint
but it gives error
`RuntimeError: element 0 of tensors does not require grad and doe... | closed | completed | false | 6 | [
"stale"
] | [] | 2024-01-29T21:14:57Z | 2024-11-22T01:39:54Z | 2024-11-22T01:39:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | loboere | 15,265,895 | MDQ6VXNlcjE1MjY1ODk1 | User | false |
huggingface/diffusers | 2,106,780,881 | I_kwDOHa8MBc59kuzR | 6,756 | https://github.com/huggingface/diffusers/issues/6756 | https://api.github.com/repos/huggingface/diffusers/issues/6756 | Handling LoRA Scale in the Legacy LoRA Implementation. | TL;DR Follow up to #6718. My suggestion is that for model blocks which use a legacy LoRA scale parameter `scale`, such as `ResnetBlock2D`, callers of these blocks like `CrossAttnDownBlock2D` should only supply the `scale` argument at inference time (e.g. when `not self.training`).
----
I am opening this issue to ... | closed | completed | false | 9 | [] | [] | 2024-01-30T01:36:56Z | 2024-03-31T14:30:04Z | 2024-03-31T14:30:04Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dg845 | 58,458,699 | MDQ6VXNlcjU4NDU4Njk5 | User | false |
huggingface/diffusers | 2,106,827,660 | I_kwDOHa8MBc59k6OM | 6,758 | https://github.com/huggingface/diffusers/issues/6758 | https://api.github.com/repos/huggingface/diffusers/issues/6758 | pipeline_stable_diffusion_xl_instantid does not support the num_images_per_prompt parameter setting | ### Describe the bug
pipeline_stable_diffusion_xl_instantid does not support the num_images_per_prompt parameter setting
### Reproduction
```
# !pip install opencv-python transformers accelerate insightface
import sys
sys.path.insert(0, sys.path[0]+"/../../")
sys.path.append("../..")
import diffusers
from dif... | closed | completed | false | 0 | [
"bug"
] | [] | 2024-01-30T02:30:19Z | 2024-02-08T20:10:14Z | 2024-02-08T20:10:14Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shaoxiaowang | 12,937,581 | MDQ6VXNlcjEyOTM3NTgx | User | false |
huggingface/diffusers | 2,107,326,570 | I_kwDOHa8MBc59m0Bq | 6,761 | https://github.com/huggingface/diffusers/issues/6761 | https://api.github.com/repos/huggingface/diffusers/issues/6761 | VaeImageProcessor.preprocess() got an unexpected keyword | ### Describe the bug
from diffusers.image_processor import PipelineImageInput
init_image = self.image_processor.preprocess(
TypeError: VaeImageProcessor.preprocess() got an unexpected keyword argument 'resize_mode'
### Reproduction
init_image = self.image_processor.preprocess(
TypeError: VaeImageProcessor... | closed | completed | false | 3 | [
"bug"
] | [] | 2024-01-30T09:21:51Z | 2024-01-31T11:12:19Z | 2024-01-31T11:12:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | t00350320 | 41,465,048 | MDQ6VXNlcjQxNDY1MDQ4 | User | false |
huggingface/diffusers | 2,107,429,843 | I_kwDOHa8MBc59nNPT | 6,763 | https://github.com/huggingface/diffusers/issues/6763 | https://api.github.com/repos/huggingface/diffusers/issues/6763 | force `timestep` to be a tensor in UNet | > IMO it could make more sense to force `timestep` to be a tensor here (make sure the pipeline is responsible for the dtype casting and not the unet. This would make the unet less error prone and more efficient
_Originally posted by @patrickvonplaten in https://github.com/huggingface/diffusers/pull/6665#discussion_r... | closed | completed | false | 1 | [] | [
"sayakpaul"
] | 2024-01-30T10:04:11Z | 2024-02-09T07:03:13Z | 2024-02-09T07:03:12Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 2,108,859,046 | I_kwDOHa8MBc59sqKm | 6,774 | https://github.com/huggingface/diffusers/issues/6774 | https://api.github.com/repos/huggingface/diffusers/issues/6774 | AttentionProcessor dtype mismatch in mixed-precision mode (xformers) | ### Describe the bug
I am getting ```ValueError: Query/Key/Value should either all have the same dtype, or (in the quantized case) Key/Value should have dtype torch.int32
query.dtype: torch.float32
key.dtype : torch.float16
value.dtype: torch.float16```
when using the xformers library with fp16. I can mit... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2024-01-30T21:59:49Z | 2024-03-05T18:41:10Z | 2024-03-05T18:41:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eitanporat | 121,024,776 | U_kgDOBzaxCA | User | false |
huggingface/diffusers | 2,108,921,939 | I_kwDOHa8MBc59s5hT | 6,776 | https://github.com/huggingface/diffusers/issues/6776 | https://api.github.com/repos/huggingface/diffusers/issues/6776 | Wrong place please ignore | ### Describe the bug
I trained a Lora using images of a person. During training, it seems like it's working on when it's done with the training. I go to use the image, and it doesn't produce images of the girl.
This is an early training image at about 2700 steps. Not great, but clearly, it is the expected result.
... | closed | completed | false | 0 | [
"bug"
] | [] | 2024-01-30T22:38:00Z | 2024-01-30T23:06:13Z | 2024-01-30T23:06:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | klubkelli | 132,837,626 | U_kgDOB-rw-g | User | false |
huggingface/diffusers | 2,109,173,614 | I_kwDOHa8MBc59t29u | 6,777 | https://github.com/huggingface/diffusers/issues/6777 | https://api.github.com/repos/huggingface/diffusers/issues/6777 | lora key not loaded | The trained lora model can not be loaded by ComfyUI
Since the issue is associated with two repos, I'll put the link of the full issues below:
https://github.com/comfyanonymous/ComfyUI/issues/2689#issue-2109167600 | closed | completed | false | 1 | [] | [] | 2024-01-31T02:45:53Z | 2024-01-31T03:53:55Z | 2024-01-31T03:53:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ChengsongLu | 61,783,323 | MDQ6VXNlcjYxNzgzMzIz | User | false |
huggingface/diffusers | 2,109,556,340 | I_kwDOHa8MBc59vUZ0 | 6,783 | https://github.com/huggingface/diffusers/issues/6783 | https://api.github.com/repos/huggingface/diffusers/issues/6783 | PixArt-XL-2-256x256 generations are messed up | ### Describe the bug
All generations with the newly released PixArt-XL-2-256x256 model look like they are viewed through a ["clear grid glass"](https://www.pinterest.co.kr/pin/bower-studios-bower-material-library-sample-glass-clear-grid-mirrors-furniture--251990541644063730/).
![astronaut_rides_horse](https://githu... | closed | completed | false | 14 | [
"bug",
"stale"
] | [] | 2024-01-31T08:32:51Z | 2024-03-03T05:01:22Z | 2024-03-03T05:01:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | stefan-canva | 90,706,016 | MDQ6VXNlcjkwNzA2MDE2 | User | false |
huggingface/diffusers | 2,109,559,041 | I_kwDOHa8MBc59vVEB | 6,784 | https://github.com/huggingface/diffusers/issues/6784 | https://api.github.com/repos/huggingface/diffusers/issues/6784 | Docstring is missing for the strength parameter in `Controlnet` pipeline | The `strength` parameter is currently not described in the docstring of the `__call__` method of the Controlnet Img2Img / inpainting pipeline and hence also not included in the docs.
https://github.com/huggingface/diffusers/blob/f4d3f913f40713cb41485a7f8aab3a31b59e86e6/src/diffusers/pipelines/controlnet/pipelin... | closed | completed | false | 2 | [] | [] | 2024-01-31T08:33:57Z | 2024-02-18T07:47:03Z | 2024-02-18T07:47:03Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tlpss | 37,955,681 | MDQ6VXNlcjM3OTU1Njgx | User | false |
huggingface/diffusers | 2,109,636,874 | I_kwDOHa8MBc59voEK | 6,785 | https://github.com/huggingface/diffusers/issues/6785 | https://api.github.com/repos/huggingface/diffusers/issues/6785 | How to finetune stable diffusion img2img(like instructpix2pix or controlnet) model with only one input channel? | Hello, experts!
I want to finetune stable diffusion img2img(like instructpix2pix or controlnet) model with only one input channel or greyscale image? I saw official docs says it is ok to increase the input channel from 4 to 9, but I want to know that is this ok to decrease the input channel to be one for finetuning?
... | closed | completed | false | 2 | [] | [] | 2024-01-31T09:17:56Z | 2024-01-31T09:27:43Z | 2024-01-31T09:27:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sapkun | 152,953,941 | U_kgDOCR3kVQ | User | false |
huggingface/diffusers | 2,110,142,301 | I_kwDOHa8MBc59xjdd | 6,790 | https://github.com/huggingface/diffusers/issues/6790 | https://api.github.com/repos/huggingface/diffusers/issues/6790 | LCM example training fails Unexpected key(s) in state_dict: "time_embedding.cond_proj.weight" | ### Describe the bug
Hi!
i'm follwing this guide for LCM training to produce models that can run ineference with 4 steps and LCMScheduler: https://github.com/huggingface/diffusers/blob/main/examples/consistency_distillation/README.md
but the example of training model fails with:
` RuntimeError('Error(s) in lo... | closed | completed | false | 13 | [
"bug",
"stale",
"training"
] | [] | 2024-01-31T13:51:27Z | 2024-04-11T17:52:14Z | 2024-04-11T17:52:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shadowofsoul | 375,696 | MDQ6VXNlcjM3NTY5Ng== | User | false |
huggingface/diffusers | 2,110,178,647 | I_kwDOHa8MBc59xsVX | 6,791 | https://github.com/huggingface/diffusers/issues/6791 | https://api.github.com/repos/huggingface/diffusers/issues/6791 | train_text_to_image_sdxl.py can't run | ### Describe the bug
I am running the training code according to the readme: https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/README_sdxl.md
I set the $MODEL_NAME to the local downloaded path, model was downloade from https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/tree/m... | closed | completed | false | 4 | [
"bug"
] | [] | 2024-01-31T14:08:48Z | 2024-02-02T02:20:41Z | 2024-02-02T02:20:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xxxyyycc | 62,633,179 | MDQ6VXNlcjYyNjMzMTc5 | User | false |
huggingface/diffusers | 2,110,545,286 | I_kwDOHa8MBc59zF2G | 6,793 | https://github.com/huggingface/diffusers/issues/6793 | https://api.github.com/repos/huggingface/diffusers/issues/6793 | Using `hydra` to configure models fail to serialize | ### Describe the bug
When using `hydra` to configure diffusers model instances, it fails to serialize (i.e. `.save_config()`). This is very similar to #3814 but the solution isn't really applicable to me.
*Reason for this behavior:* `diffusers`' s `FrozenDict` apparently doesn't take care of list/tuple arguments ... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-01-31T16:54:25Z | 2024-06-20T03:34:57Z | 2024-02-03T10:16:26Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dasayan05 | 18,719,603 | MDQ6VXNlcjE4NzE5NjAz | User | false |
huggingface/diffusers | 2,110,765,060 | I_kwDOHa8MBc59z7gE | 6,794 | https://github.com/huggingface/diffusers/issues/6794 | https://api.github.com/repos/huggingface/diffusers/issues/6794 | Import of cast_training_params from diffusers.training_utils is giving an error | ### Describe the bug
Getting the following error message
```
/.............../venvs/lib/python3.11/site-packages/diffusers/utils/outputs.py:63: UserWarning: torch.utils._pytree._register_pytree_node is deprecated. Please use torch.utils._pytree.register_pytree_node instead.
torch.utils._pytree._register_pytree... | closed | completed | false | 6 | [
"bug"
] | [] | 2024-01-31T18:58:39Z | 2024-10-15T21:03:24Z | 2024-10-15T21:03:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AvisP | 10,809,866 | MDQ6VXNlcjEwODA5ODY2 | User | false |
huggingface/diffusers | 2,111,227,463 | I_kwDOHa8MBc591sZH | 6,798 | https://github.com/huggingface/diffusers/issues/6798 | https://api.github.com/repos/huggingface/diffusers/issues/6798 | Saving/Loading multiple models of the same type with Accelerator | Suppose I'm training two models of the same UNet2DConditionModel type. Is it possible to save/load their weights using Accelerator's `register_save_state_pre_hook` and `register_load_state_pre_hook` methods?
Both methods provide a list of models to a callback function, and within the callback function, it seems that... | closed | completed | false | 2 | [] | [] | 2024-02-01T00:28:18Z | 2024-11-22T01:39:23Z | 2024-11-22T01:39:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hanweikung | 643,825 | MDQ6VXNlcjY0MzgyNQ== | User | false |
huggingface/diffusers | 2,111,254,542 | I_kwDOHa8MBc591zAO | 6,799 | https://github.com/huggingface/diffusers/issues/6799 | https://api.github.com/repos/huggingface/diffusers/issues/6799 | add a `is_torchvision_available` function to diffusers | null | closed | completed | false | 0 | [] | [] | 2024-02-01T00:55:28Z | 2024-02-01T01:39:25Z | 2024-02-01T01:39:25Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 2,111,377,813 | I_kwDOHa8MBc592RGV | 6,801 | https://github.com/huggingface/diffusers/issues/6801 | https://api.github.com/repos/huggingface/diffusers/issues/6801 | Problem load Scheduler (diffusers ver 0.26.0) | ### Describe the bug
If I install diffuser version **0.25.1 it works fine**
but when I updated to version 0.26.0
### Reproduction
(model=sd 1.5)
from diffusers import AutoPipelineForText2Image
import torch
pipeline =
AutoPipelineForText2Image.from_pretrained(model,requires_safety_checker=False,safety_check... | closed | completed | false | 8 | [
"bug"
] | [] | 2024-02-01T02:45:28Z | 2024-02-08T20:31:50Z | 2024-02-08T20:31:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | josemerinom | 20,940,429 | MDQ6VXNlcjIwOTQwNDI5 | User | false |
huggingface/diffusers | 2,111,499,041 | I_kwDOHa8MBc592ush | 6,802 | https://github.com/huggingface/diffusers/issues/6802 | https://api.github.com/repos/huggingface/diffusers/issues/6802 | [IP-Adapter] Adding IP-adapter masking feature | We just added multi-adapter and multi-image support for IP-Adapter https://github.com/huggingface/diffusers/pull/6573
Anyone interested in adding an IP adapter masking feature? It will allow you to divide the image and assign different ip-adapter to the specified portion of the image.
more information on the feat... | closed | completed | false | 10 | [
"community-examples",
"Good second issue",
"contributions-welcome",
"advanced"
] | [] | 2024-02-01T04:11:06Z | 2024-07-16T09:28:39Z | 2024-02-19T04:06:15Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 2,111,768,343 | I_kwDOHa8MBc593wcX | 6,804 | https://github.com/huggingface/diffusers/issues/6804 | https://api.github.com/repos/huggingface/diffusers/issues/6804 | How to only offload some parts but not whole model into cpu? | Using enable_cpu_offload() will offload the whole model into cpu, which can occupy a large part of cpu memory. How can I just offload a part of model into cpu? | closed | completed | false | 2 | [] | [] | 2024-02-01T07:43:04Z | 2024-02-02T04:59:43Z | 2024-02-02T04:59:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | blx0102 | 17,761,044 | MDQ6VXNlcjE3NzYxMDQ0 | User | false |
huggingface/diffusers | 2,112,044,578 | I_kwDOHa8MBc594z4i | 6,807 | https://github.com/huggingface/diffusers/issues/6807 | https://api.github.com/repos/huggingface/diffusers/issues/6807 | [Contributor Experience] Test Collection fails on MPS | ### Describe the bug
On MPS, the test collection fails. Therefore not all tests are shown e.g. in VS Code, which makes the contributor experience more tedious.
Currently, 4515-4486 = 29 tests are missed.
---
The error is caused because in `utls/testing_utils.py` the variable `torch_device` is defined as a str... | closed | completed | false | 1 | [
"bug"
] | [] | 2024-02-01T09:48:25Z | 2024-02-02T15:29:01Z | 2024-02-02T15:29:01Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | UmerHA | 40,663,591 | MDQ6VXNlcjQwNjYzNTkx | User | false |
huggingface/diffusers | 2,112,251,567 | I_kwDOHa8MBc595mav | 6,809 | https://github.com/huggingface/diffusers/issues/6809 | https://api.github.com/repos/huggingface/diffusers/issues/6809 | Some LoRA adapters cannot load anymore when using PEFT library under the hood | ### Describe the bug
The diffusers lib tries to use PEFT under the hood when available.
Some LoRA adapters cannot be loaded properly when the peft backend is in use
Script to reproduce and stack trace below
workaround: uninstalling the peft lib causes this to switch from true to false and use the second else br... | closed | completed | false | 0 | [
"bug"
] | [] | 2024-02-01T11:21:18Z | 2024-02-01T13:19:59Z | 2024-02-01T12:21:30Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | oOraph | 13,552,058 | MDQ6VXNlcjEzNTUyMDU4 | User | false |
huggingface/diffusers | 2,112,615,244 | I_kwDOHa8MBc596_NM | 6,811 | https://github.com/huggingface/diffusers/issues/6811 | https://api.github.com/repos/huggingface/diffusers/issues/6811 | i2vgen-xl Not Found for url: https://huggingface.co/ali-vilab/i2vgen-xl/resolve/main/model_index.json. | ### Describe the bug
```
import torch
from diffusers import I2VGenXLPipeline
from diffusers.utils import export_to_gif, load_image
repo_id = "ali-vilab/i2vgen-xl"
pipeline = I2VGenXLPipeline.from_pretrained(repo_id, torch_dtype=torch.float16).to("cuda")
pipeline.enable_model_cpu_offload()
image_url = "h... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-02-01T14:08:25Z | 2024-02-07T17:28:34Z | 2024-02-07T17:28:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhaoyun0071 | 35,762,050 | MDQ6VXNlcjM1NzYyMDUw | User | false |
huggingface/diffusers | 2,112,678,490 | I_kwDOHa8MBc597Opa | 6,812 | https://github.com/huggingface/diffusers/issues/6812 | https://api.github.com/repos/huggingface/diffusers/issues/6812 | ValueError: Query/Key/Value should either all have the same dtype training LORA | ### Describe the bug
LORA training don't work with mixed precision enabled:
File "/home/imgen/miniconda3/envs/py32/lib/python3.11/site-packages/xformers/ops/fmha/__init__.py", line 348, in _memory_efficient_at
tention_forward_requires_grad
inp.validate_inputs()
File "/home/imgen/miniconda3/envs/py32/li... | closed | completed | false | 16 | [
"bug",
"stale"
] | [] | 2024-02-01T14:34:55Z | 2024-11-22T01:37:53Z | 2024-11-22T01:37:53Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | noskill | 733,626 | MDQ6VXNlcjczMzYyNg== | User | false |
huggingface/diffusers | 2,113,108,961 | I_kwDOHa8MBc5983vh | 6,814 | https://github.com/huggingface/diffusers/issues/6814 | https://api.github.com/repos/huggingface/diffusers/issues/6814 | [design] adding a `validation` decorator | We try to avoid using the `kwargs` argument at diffusers because it's error prone and makes our code less readable. However, there are some instances where a fixed signature might not be practical. For example, our default attention processor class `AttnProcessor/AttnProcessor2.0` can now be used in conjunction with ot... | open | null | false | 6 | [
"stale"
] | [] | 2024-02-01T18:01:06Z | 2024-12-22T15:08:15Z | null | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 2,113,621,230 | I_kwDOHa8MBc59-0zu | 6,815 | https://github.com/huggingface/diffusers/issues/6815 | https://api.github.com/repos/huggingface/diffusers/issues/6815 | RuntimeWarning: invalid value encountered in cast images = (images * 255).round().astype("uint8") output black image | ### Describe the bug
When running the stable-diffusion-2-1 I get a runtime warning "RuntimeWarning: invalid value encountered in cast
images = (images * 255).round().astype("uint8")" and the image output is black.
![astronaut_rides_horse](https://github.com/huggingface/diffusers/assets/5935949/4d158e10-a5bf-4ec5... | closed | completed | false | 21 | [
"bug"
] | [] | 2024-02-01T22:28:26Z | 2025-08-18T14:50:45Z | 2024-03-14T14:07:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pavankay | 5,935,949 | MDQ6VXNlcjU5MzU5NDk= | User | false |
huggingface/diffusers | 2,113,892,348 | I_kwDOHa8MBc59_2_8 | 6,817 | https://github.com/huggingface/diffusers/issues/6817 | https://api.github.com/repos/huggingface/diffusers/issues/6817 | How to use class_labels in the Unet2DConditionalModel or Unet2DModel when forward? | Hi, I want to know what the shape or format of "class" is if I want to add the class condition to the unet? Just set the **classe_labels** 0, 1, 2, 3?
Unet2DModel: **class_labels** (torch.FloatTensor, optional, defaults to None) — Optional class labels for conditioning. Their embeddings will be summed with the times... | closed | completed | false | 2 | [] | [] | 2024-02-02T02:17:40Z | 2024-02-07T07:31:35Z | 2024-02-07T07:31:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | boqian-li | 88,582,462 | MDQ6VXNlcjg4NTgyNDYy | User | false |
huggingface/diffusers | 2,113,913,004 | I_kwDOHa8MBc59_8Cs | 6,818 | https://github.com/huggingface/diffusers/issues/6818 | https://api.github.com/repos/huggingface/diffusers/issues/6818 | After call IP - Adapter unload method StableDiffusionImg2ImgPipeline(**pipe.components) error | hi!
When I used 1.5 to load the model, I enabled the security check mechanism, uninstalled after using the IP-Adapter, and generated the pipe from the previous pipe, which would throw an error with the security check parameters
https://github.com/huggingface/diffusers/blob/adcbe674a49af31415a1131174143d58c0e68d7... | closed | completed | false | 2 | [] | [] | 2024-02-02T02:41:37Z | 2024-02-08T20:27:04Z | 2024-02-08T20:27:04Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Honey-666 | 76,274,942 | MDQ6VXNlcjc2Mjc0OTQy | User | false |
huggingface/diffusers | 2,113,926,973 | I_kwDOHa8MBc59__c9 | 6,819 | https://github.com/huggingface/diffusers/issues/6819 | https://api.github.com/repos/huggingface/diffusers/issues/6819 | How to let diffusers use local code for pipelineinstead of download it online everytime We use it? | I tried to use the instaflowpipeline from example/community to.run my test However, even after i git cloned the repository to my environment it still Keep trying to Download the latest object of the instaflow pipeline code Unfortunately in my area is hard for the environment to download it directly from rawgithub. ... | closed | completed | false | 7 | [] | [] | 2024-02-02T02:53:48Z | 2024-11-28T05:44:10Z | 2024-11-28T05:44:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Kevin-shihello-world | 32,519,859 | MDQ6VXNlcjMyNTE5ODU5 | User | false |
huggingface/diffusers | 2,113,961,984 | I_kwDOHa8MBc5-AIAA | 6,821 | https://github.com/huggingface/diffusers/issues/6821 | https://api.github.com/repos/huggingface/diffusers/issues/6821 | ValueError: --caption_column' value 'text' needs to be one of: image | ### Describe the bug
when i use load_dataset, it's right.
but use train_text_to_image_lora.py it wrong.
>>> from datasets import load_dataset
>>> dataset = load_dataset( "imagefolder" , data_dir= "dataset/train/image" , split= "train")
Resolving data files: 100%|██████████████████████████████████████████████████... | closed | completed | false | 5 | [
"bug",
"stale",
"training"
] | [] | 2024-02-02T03:21:26Z | 2024-03-13T15:03:58Z | 2024-03-13T15:03:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | linnnff | 41,185,967 | MDQ6VXNlcjQxMTg1OTY3 | User | false |
huggingface/diffusers | 2,114,290,569 | I_kwDOHa8MBc5-BYOJ | 6,823 | https://github.com/huggingface/diffusers/issues/6823 | https://api.github.com/repos/huggingface/diffusers/issues/6823 | Cannot use controlnet | ### Describe the bug
When I ｗant use "ControlNetModel.from_single_file" , I give it correct path , but it doesn't work.
### Reproduction
from diffusers import StableDiffusionControlNetPipeline ,ControlNetModel
import torch
import cv2
img = cv2.imread('/home/leo/blues.png')
controlnet_model_path = "/mnt/d/model... | closed | completed | false | 0 | [
"bug"
] | [] | 2024-02-02T07:26:59Z | 2024-02-02T07:34:54Z | 2024-02-02T07:34:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kcy3f0 | 108,375,009 | U_kgDOBnWr4Q | User | false |
huggingface/diffusers | 2,114,457,114 | I_kwDOHa8MBc5-CA4a | 6,824 | https://github.com/huggingface/diffusers/issues/6824 | https://api.github.com/repos/huggingface/diffusers/issues/6824 | Exposing models elementary blocks to design new models | **Is your feature request related to a problem? Please describe.**
I would like to implement a new model following the `diffusers` API, namely the conditional UNet1D (mentioned [here](https://github.com/huggingface/diffusers/issues/3044), but saying that it will not be implemented at the moment).
To do so, I would li... | closed | completed | false | 4 | [] | [] | 2024-02-02T09:13:33Z | 2024-07-01T19:00:08Z | 2024-03-11T10:50:01Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zaccharieramzi | 6,387,497 | MDQ6VXNlcjYzODc0OTc= | User | false |
huggingface/diffusers | 2,115,320,823 | I_kwDOHa8MBc5-FTv3 | 6,827 | https://github.com/huggingface/diffusers/issues/6827 | https://api.github.com/repos/huggingface/diffusers/issues/6827 | Fix PIAPipeline implementation | ### Describe the bug
New pipeline PIAPipeline states:
> PIA works with a MotionAdapter checkpoint and a Stable Diffusion 1.5 model checkpoint
however, PIAPipeline
1. does not have method `from_single_file`
2. cannot be instantiated from existing loaded model
as a result, i pretty much cannot make it work usi... | closed | completed | false | 6 | [
"bug"
] | [
"DN6"
] | 2024-02-02T16:41:15Z | 2024-02-05T02:20:13Z | 2024-02-05T02:20:13Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 2,115,880,040 | I_kwDOHa8MBc5-HcRo | 6,829 | https://github.com/huggingface/diffusers/issues/6829 | https://api.github.com/repos/huggingface/diffusers/issues/6829 | [test] add a `IPAdapterTexterMixin` | I think we should add a `IPAdapterTexterMIxin` and run fast tests on every pipeline that supports IP-Adapter, so that we will be able to catch breaking changes like this https://github.com/huggingface/diffusers/pull/6828
some thoughts:
* we can add a `IPAdapterTexterMIxin` class to `test_pipelines_common.py` https:... | closed | completed | false | 4 | [
"help wanted",
"Good second issue",
"contributions-welcome"
] | [] | 2024-02-02T22:07:51Z | 2024-02-23T00:25:35Z | 2024-02-23T00:25:34Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 2,116,132,371 | I_kwDOHa8MBc5-IZ4T | 6,832 | https://github.com/huggingface/diffusers/issues/6832 | https://api.github.com/repos/huggingface/diffusers/issues/6832 | IP Adapter Face ID community pipeline fails with AttributeError: 'tuple' object has no attribute 'shape' | ### Describe the bug
Hello Diffusers team and contributors,
Starting from commit 2e8d18e, the [IP Adapter Face ID](https://github.com/huggingface/diffusers/blob/main/examples/community/README.md#ip-adapter-face-id) community pipeline stopped working and inference terminate during first iteration with the following ... | closed | completed | false | 6 | [
"bug"
] | [] | 2024-02-03T02:35:47Z | 2024-02-04T06:49:32Z | 2024-02-03T18:56:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Woolverine94 | 146,110,286 | U_kgDOCLV3Tg | User | false |
huggingface/diffusers | 2,116,633,767 | I_kwDOHa8MBc5-KUSn | 6,838 | https://github.com/huggingface/diffusers/issues/6838 | https://api.github.com/repos/huggingface/diffusers/issues/6838 | Diffusers 0.26 break a lot of existing pipelines and 3rd party modules | ### Describe the bug
diffusers==0.26 via pr #6665 which introduces support for `I2VGenXL`
redefines commonly used params in `BasicTransformerBlock`
for example, `if self.use_ada_layer_norm` now becomes `if norm_type == "ada_norm":` which would mean that every single pipeline that relies on those layers would n... | closed | completed | false | 7 | [
"bug"
] | [] | 2024-02-03T18:31:21Z | 2024-02-04T21:16:45Z | 2024-02-04T21:16:45Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 2,116,777,094 | I_kwDOHa8MBc5-K3SG | 6,839 | https://github.com/huggingface/diffusers/issues/6839 | https://api.github.com/repos/huggingface/diffusers/issues/6839 | StableVideoDiffusionPipeline cannot use from_single_file | ### Describe the bug
The StableVideoDiffusionPipeline cannot load models in any format other than diffusers, which is problematic as the latest StableVideoDiffusion model has only been released in safetensors. See: https://huggingface.co/stabilityai/stable-video-diffusion-img2vid-xt-1-1
### Reproduction
svd_p... | closed | completed | false | 14 | [
"bug",
"stale",
"single_file"
] | [] | 2024-02-04T00:49:05Z | 2024-11-28T05:45:00Z | 2024-11-28T05:45:00Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Meatfucker | 74,834,323 | MDQ6VXNlcjc0ODM0MzIz | User | false |
huggingface/diffusers | 1,392,211,194 | I_kwDOHa8MBc5S-3T6 | 684 | https://github.com/huggingface/diffusers/issues/684 | https://api.github.com/repos/huggingface/diffusers/issues/684 | Training memory optimizations not working on AMD hardware | ### Describe the bug
[Dreambooth training example](https://github.com/huggingface/diffusers/tree/main/examples/dreambooth) has a section about training on 16GB GPU. As Radeon Navi 21 series models all have 16GB available this in theory would increase the amount of hardware that can train models by a really large mar... | closed | completed | false | 18 | [
"stale"
] | [
"patil-suraj"
] | 2022-09-30T10:17:56Z | 2023-01-15T15:03:54Z | 2023-01-15T15:03:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | errnoh | 373,946 | MDQ6VXNlcjM3Mzk0Ng== | User | false |
huggingface/diffusers | 2,116,963,657 | I_kwDOHa8MBc5-Lk1J | 6,842 | https://github.com/huggingface/diffusers/issues/6842 | https://api.github.com/repos/huggingface/diffusers/issues/6842 | ImportError: cannot import name 'Prompt2PromptPipeline' from 'diffusers.pipelines' | ### Describe the bug
I run prompt to prompt according [doc](https://github.com/huggingface/diffusers/blob/13001ee/examples/community/README.md#prompt2prompt-pipeline), while get error
```
ImportError: cannot import name 'Prompt2PromptPipeline' from 'diffusers.pipelines'
```
### Reproduction
```
from diffus... | closed | completed | false | 3 | [
"bug",
"good first issue",
"contributions-welcome"
] | [] | 2024-02-04T08:51:08Z | 2024-02-16T01:30:18Z | 2024-02-16T01:30:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | akk-123 | 98,469,560 | U_kgDOBd6GuA | User | false |
huggingface/diffusers | 1,392,386,483 | I_kwDOHa8MBc5S_iGz | 685 | https://github.com/huggingface/diffusers/issues/685 | https://api.github.com/repos/huggingface/diffusers/issues/685 | Can't import `diffusers` with Flax but without PyTorch | ### Describe the bug
I discovered this when testing #659. It is currently possible to `import diffusers` without any of PyTorch and Flax; the dummy classes are loaded. But when Flax is installed and PyTorch isn't, then importing fails.
This is because of two reasons:
- `modeling_flax_utils.py` imports `load_stat... | closed | completed | false | 0 | [
"bug"
] | [
"pcuenca"
] | 2022-09-30T12:43:50Z | 2022-10-03T14:23:46Z | 2022-10-03T14:23:46Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pcuenca | 1,177,582 | MDQ6VXNlcjExNzc1ODI= | User | false |
huggingface/diffusers | 2,117,783,057 | I_kwDOHa8MBc5-Os4R | 6,852 | https://github.com/huggingface/diffusers/issues/6852 | https://api.github.com/repos/huggingface/diffusers/issues/6852 | No safetensors weights found for model PharMolix/BioMedGPT-LM-7B | ### Model/Pipeline/Scheduler description
I am deploying PharMolix/BioMedGPT-LM-7B (https://huggingface.co/PharMolix/BioMedGPT-LM-7B) on Sagemaker. I am using the below code snippet:
"""
import json
import sagemaker
from sagemaker import get_execution_role
import boto3
from sagemaker.huggingface import Hugging... | closed | completed | false | 0 | [] | [] | 2024-02-05T06:32:27Z | 2024-02-05T08:59:57Z | 2024-02-05T08:59:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | learner-subodh | 62,687,864 | MDQ6VXNlcjYyNjg3ODY0 | User | false |
huggingface/diffusers | 2,117,844,525 | I_kwDOHa8MBc5-O74t | 6,853 | https://github.com/huggingface/diffusers/issues/6853 | https://api.github.com/repos/huggingface/diffusers/issues/6853 | replace training for stable diffusion inpainting model | 11 | closed | completed | false | 0 | [] | [] | 2024-02-05T07:19:19Z | 2024-02-05T07:21:50Z | 2024-02-05T07:21:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | StonyB | 104,622,078 | U_kgDOBjxn_g | User | false |
huggingface/diffusers | 2,117,849,680 | I_kwDOHa8MBc5-O9JQ | 6,854 | https://github.com/huggingface/diffusers/issues/6854 | https://api.github.com/repos/huggingface/diffusers/issues/6854 | replace training for stable diffusion inpainting model #6853 | null | closed | completed | false | 0 | [] | [] | 2024-02-05T07:22:57Z | 2024-02-05T07:29:01Z | 2024-02-05T07:28:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | StonyB | 104,622,078 | U_kgDOBjxn_g | User | false |
huggingface/diffusers | 2,118,402,630 | I_kwDOHa8MBc5-REJG | 6,858 | https://github.com/huggingface/diffusers/issues/6858 | https://api.github.com/repos/huggingface/diffusers/issues/6858 | AttributeError: 'str' object has no attribute 'convert' when training controlnet | ### Describe the bug
when I training controlnet (by **example/controlnet** ), `AttributeError: 'str' object has no attribute 'convert' ` appears
### Reproduction
I use the same bash with example/controlnet/README.md
```
accelerate launch train_controlnet.py \
--pretrained_model_name_or_path=$MODEL_DIR ... | open | reopened | false | 16 | [
"bug"
] | [] | 2024-02-05T12:17:38Z | 2025-07-21T13:56:28Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | XiangjiBU | 45,818,685 | MDQ6VXNlcjQ1ODE4Njg1 | User | false |
huggingface/diffusers | 2,119,182,721 | I_kwDOHa8MBc5-UCmB | 6,863 | https://github.com/huggingface/diffusers/issues/6863 | https://api.github.com/repos/huggingface/diffusers/issues/6863 | from_single_file missing in AutoPipeline API | `from_single_file` is missing for classes such as `AutoPipelineForText2Image` ! | closed | completed | false | 6 | [] | [
"yiyixuxu"
] | 2024-02-05T18:17:02Z | 2024-02-18T06:38:13Z | 2024-02-18T06:38:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | oxysoft | 3,945,277 | MDQ6VXNlcjM5NDUyNzc= | User | false |
huggingface/diffusers | 2,119,754,930 | I_kwDOHa8MBc5-WOSy | 6,865 | https://github.com/huggingface/diffusers/issues/6865 | https://api.github.com/repos/huggingface/diffusers/issues/6865 | Missing check for Image instance for | from diffusers.utils import load_image | ### Describe the bug
check load_image method here, https://huggingface.co/transformers/v4.11.3/_modules/transformers/pipelines/image_classification.html ,
and check this image at path "C:\Users\User\AppData\Local\Programs\Python\Python311\Lib\site-packages\diffusers\utils\loading_utils.py"
![image](https://github... | closed | completed | false | 4 | [
"bug"
] | [] | 2024-02-06T01:20:47Z | 2024-02-14T16:29:36Z | 2024-02-08T20:01:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nikhil-swamix | 54,004,431 | MDQ6VXNlcjU0MDA0NDMx | User | false |
huggingface/diffusers | 2,120,171,832 | I_kwDOHa8MBc5-X0E4 | 6,866 | https://github.com/huggingface/diffusers/issues/6866 | https://api.github.com/repos/huggingface/diffusers/issues/6866 | IP-Adapter support for StableDiffusionXLControlNetInpaintPipeline | Hello 👋
**Is your feature request related to a problem? Please describe.**
`StableDiffusionXLControlNetPipeline` and `StableDiffusionXLControlNetImg2ImgPipeline` support IP-Adapter thanks to https://github.com/huggingface/diffusers/pull/6573. However, it is not added for `StableDiffusionXLControlNetInpaintPipelin... | closed | completed | false | 2 | [
"good first issue",
"contributions-welcome"
] | [] | 2024-02-06T07:55:04Z | 2024-02-19T18:20:25Z | 2024-02-19T18:20:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | enesmsahin | 36,894,859 | MDQ6VXNlcjM2ODk0ODU5 | User | false |
huggingface/diffusers | 2,120,185,851 | I_kwDOHa8MBc5-X3f7 | 6,867 | https://github.com/huggingface/diffusers/issues/6867 | https://api.github.com/repos/huggingface/diffusers/issues/6867 | Installation error | ### Describe the bug
Installation error, when pip installing Diffusers(the latest on git).
### Reproduction
pip install git+https://github.com/huggingface/diffusers.git
### Logs
```shell
Installing: diffusers module
Collecting git+https://github.com/huggingface/diffusers.git
Cloning https://github.c... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2024-02-06T08:04:32Z | 2024-04-27T16:24:55Z | 2024-03-07T16:12:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 2,120,233,543 | I_kwDOHa8MBc5-YDJH | 6,869 | https://github.com/huggingface/diffusers/issues/6869 | https://api.github.com/repos/huggingface/diffusers/issues/6869 | export_to_video causes: ValueError: too many values to unpack (expected 3) | ### Describe the bug
export_to_video causes: ValueError: too many values to unpack (expected 3)
I know that some work has been done here, but the error persists in 0.26.2, even though this patch has been committed some time ago : https://github.com/huggingface/diffusers/pull/6715
### Reproduction
Run the code fr... | closed | completed | false | 5 | [
"bug"
] | [] | 2024-02-06T08:34:59Z | 2024-02-08T20:15:27Z | 2024-02-08T20:15:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 2,120,253,596 | I_kwDOHa8MBc5-YICc | 6,870 | https://github.com/huggingface/diffusers/issues/6870 | https://api.github.com/repos/huggingface/diffusers/issues/6870 | enable_vae_tiling is not found SD image to image and inpaint pipeline | ### Describe the bug
I tried to use vae_tiling in StableDiffusionImg2ImgPipeline and StableDiffusionInpaintPipeline, but the function was not defined. This option is probably needed to save memory when generating large images.
I noticed that it is enabled for StableDiffusionPipeline and ControlNet Pipeline. Is th... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-02-06T08:47:18Z | 2024-06-13T02:25:57Z | 2024-02-08T19:50:02Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MrSyee | 17,582,508 | MDQ6VXNlcjE3NTgyNTA4 | User | false |
huggingface/diffusers | 2,121,808,810 | I_kwDOHa8MBc5-eDuq | 6,879 | https://github.com/huggingface/diffusers/issues/6879 | https://api.github.com/repos/huggingface/diffusers/issues/6879 | Custom Diffusion: RuntimeError: expected mat1 and mat2 to have the same dtype, but got: c10::Half != float | ### Describe the bug
when running custom diffusion on my 20 photos repositories ... I run into this error that is related to data type difference..
### Reproduction
!accelerate launch train_custom_diffusion.py \
--pretrained_model_name_or_path=$MODEL_NAME \
--instance_data_dir=$INSTANCE_DIR \
--output_dir=$OUT... | closed | completed | false | 16 | [
"bug",
"training"
] | [] | 2024-02-06T23:03:07Z | 2024-08-20T01:28:51Z | 2024-08-20T01:28:51Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rezkanas | 64,679,971 | MDQ6VXNlcjY0Njc5OTcx | User | false |
huggingface/diffusers | 2,121,814,262 | I_kwDOHa8MBc5-eFD2 | 6,880 | https://github.com/huggingface/diffusers/issues/6880 | https://api.github.com/repos/huggingface/diffusers/issues/6880 | Custom Diffusion retrieve.py | ### Describe the bug
when running retrieve.py to produce class prior photo using laion datasset,, I run through below errors.
url is bad in the code below... could be the root cause?
```
client = ClipClient(
url="https://knn.laion.ai/knn-service", indice_name="laion5B-L-14", num_images=num_images, aesthet... | closed | not_planned | false | 18 | [
"bug",
"stale"
] | [] | 2024-02-06T23:07:29Z | 2024-11-28T05:45:25Z | 2024-11-28T05:45:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rezkanas | 64,679,971 | MDQ6VXNlcjY0Njc5OTcx | User | false |
huggingface/diffusers | 2,121,963,871 | I_kwDOHa8MBc5-eplf | 6,882 | https://github.com/huggingface/diffusers/issues/6882 | https://api.github.com/repos/huggingface/diffusers/issues/6882 | Slow SDXL inference with JAX on Cloud TPU v5e for sizes other than 1024x1024 | ### Describe the bug
Following the blog post on Accelerating Stable Diffusion XL Inference with JAX on Cloud TPU v5e. This worked magically until I tried to generate an image in a different size. At 1024x1024 we get inference latency of ~3s per image (as compared to ~8s on the NVIDIA A10G). But change the resolution... | open | null | false | 7 | [
"bug",
"stale",
"jax/flax"
] | [] | 2024-02-07T01:10:13Z | 2024-10-09T15:08:43Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zstiggz | 704,532 | MDQ6VXNlcjcwNDUzMg== | User | false |
huggingface/diffusers | 2,122,087,767 | I_kwDOHa8MBc5-fH1X | 6,883 | https://github.com/huggingface/diffusers/issues/6883 | https://api.github.com/repos/huggingface/diffusers/issues/6883 | =。= | null | closed | completed | false | 0 | [
"bug"
] | [] | 2024-02-07T03:03:00Z | 2024-02-07T07:56:46Z | 2024-02-07T07:50:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiarenyf | 12,405,578 | MDQ6VXNlcjEyNDA1NTc4 | User | false |
huggingface/diffusers | 2,122,533,753 | I_kwDOHa8MBc5-g0t5 | 6,891 | https://github.com/huggingface/diffusers/issues/6891 | https://api.github.com/repos/huggingface/diffusers/issues/6891 | [Tracker] use the new model card utilities for saving model cards from the training script | With https://github.com/huggingface/diffusers/pull/6729, we shifted to a more resilient and robust way to generate model cards from the DreamBooth training scripts.
So, we're keen for the community to help us do this for the other training scripts that save model cards:
- [x] [T2I](https://github.com/huggingface... | closed | completed | false | 28 | [
"contributions-welcome"
] | [] | 2024-02-07T09:11:39Z | 2024-10-17T20:43:56Z | 2024-10-16T02:06:38Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 2,122,902,426 | I_kwDOHa8MBc5-iOua | 6,892 | https://github.com/huggingface/diffusers/issues/6892 | https://api.github.com/repos/huggingface/diffusers/issues/6892 | [Discussion] allow new merging methods from `peft` | We rely on `peft` to supercharge our support doing all things LoRA: https://huggingface.co/docs/diffusers/main/en/tutorials/using_peft_for_inference.
Currently, it's possible to do multi-LoRA inference seamlessly using `peft`:
https://huggingface.co/docs/diffusers/main/en/tutorials/using_peft_for_inference#combine... | closed | completed | false | 25 | [
"stale",
"peft"
] | [] | 2024-02-07T12:15:35Z | 2024-03-13T15:08:33Z | 2024-03-13T15:08:33Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 2,123,144,387 | I_kwDOHa8MBc5-jJzD | 6,894 | https://github.com/huggingface/diffusers/issues/6894 | https://api.github.com/repos/huggingface/diffusers/issues/6894 | LoRA from train_dreambooth_lora_sdxl.py is not working in A1111 anymore | ### Describe the bug
I have been using [train_dreambooth_lora_sdxl.py](https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth_lora_sdxl.py) and [convert_diffusers_sdxl_lora_to_webui.py](https://github.com/huggingface/diffusers/blob/main/scripts/convert_diffusers_sdxl_lora_to_webui.p... | closed | completed | false | 32 | [
"bug",
"stale",
"conversion script"
] | [
"sayakpaul"
] | 2024-02-07T14:20:01Z | 2024-04-22T21:55:39Z | 2024-04-22T21:55:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patryk-bartkowiak-nitid | 159,049,122 | U_kgDOCXrlog | User | false |
huggingface/diffusers | 2,124,092,053 | I_kwDOHa8MBc5-mxKV | 6,899 | https://github.com/huggingface/diffusers/issues/6899 | https://api.github.com/repos/huggingface/diffusers/issues/6899 | Bug in `convert_original_stable_diffusion_to_diffusers.py` | ### Describe the bug
See the steps to reproduce below.
This is caused by [this code change](https://github.com/huggingface/diffusers/commit/cb4b3f0b784db83309b3954bd2118c5dc9f30e8f#diff-10656ff1d173475012d0818b48f05a092001418ac70745e803b97fbb493dfb6aL10). `yaml.safe_load()` takes a stream object not a file path na... | closed | completed | false | 7 | [
"bug"
] | [] | 2024-02-07T23:36:04Z | 2024-04-05T03:39:17Z | 2024-02-08T13:22:13Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ehsan | 168,508 | MDQ6VXNlcjE2ODUwOA== | User | false |
huggingface/diffusers | 1,292,414,167 | I_kwDOHa8MBc5NCKzX | 69 | https://github.com/huggingface/diffusers/issues/69 | https://api.github.com/repos/huggingface/diffusers/issues/69 | 1D Upsampler / Downsample | We should also write tests for 1D Upsampler and 1D downsampler layers (for `unet_rl.py`)
Once all layers are nicely consolidated and we have finished the first round of refactoring, up- and downsampling layers should probably be split into layers that have learnable parameters and those that don't.
cc @patil-sura... | closed | completed | false | 0 | [] | [
"patil-suraj"
] | 2022-07-03T20:33:32Z | 2022-07-21T19:05:57Z | 2022-07-21T19:05:57Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 2,124,180,081 | I_kwDOHa8MBc5-nGpx | 6,901 | https://github.com/huggingface/diffusers/issues/6901 | https://api.github.com/repos/huggingface/diffusers/issues/6901 | FlaxDDIMScheduler is missing clip_sample implementation | ### Describe the bug
Setting `clip_sample=True` can sometimes help w/ numerical stability during DDIM inference. This is an important and useful feature that is actually promised in the FlaxDDIMScheduler [docstring](https://github.com/huggingface/diffusers/blob/97d004b9b4283d194a4d2c410be2d441527e0800/src/diffusers/sc... | closed | completed | false | 5 | [
"bug",
"stale",
"contributions-welcome"
] | [] | 2024-02-08T01:06:40Z | 2024-03-10T15:38:30Z | 2024-03-10T15:38:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | SudeepDasari | 16,604,347 | MDQ6VXNlcjE2NjA0MzQ3 | User | false |
huggingface/diffusers | 2,125,206,968 | I_kwDOHa8MBc5-rBW4 | 6,906 | https://github.com/huggingface/diffusers/issues/6906 | https://api.github.com/repos/huggingface/diffusers/issues/6906 | LoRA dreambooth example is broken | ### Describe the bug
Following the instructions from https://github.com/huggingface/diffusers/tree/main/examples/dreambooth do not yield personalised image.
I noticed the problem in my own script, but decided t... | closed | completed | false | 6 | [
"bug"
] | [] | 2024-02-08T13:35:28Z | 2024-02-09T01:40:25Z | 2024-02-09T01:40:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AlexUmnov | 17,798,528 | MDQ6VXNlcjE3Nzk4NTI4 | User | false |
huggingface/diffusers | 2,125,441,513 | I_kwDOHa8MBc5-r6np | 6,911 | https://github.com/huggingface/diffusers/issues/6911 | https://api.github.com/repos/huggingface/diffusers/issues/6911 | UNetConditional3DModel doesn't use class labels | ### Describe the bug
In `forward` `UNet3DConditionModel` takes an argument `class_labels`, however it isn't used anywhere inside of `forward`. It is rather useful to be able to condition on classes, so please, add some form of conditioning on class labels.
### Reproduction
Bug in the code.
### Logs
_No r... | closed | not_planned | false | 1 | [
"bug"
] | [] | 2024-02-08T15:23:23Z | 2024-02-29T19:08:40Z | 2024-02-29T19:08:40Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MK-2012 | 70,899,588 | MDQ6VXNlcjcwODk5NTg4 | User | false |
huggingface/diffusers | 2,126,200,770 | I_kwDOHa8MBc5-uz_C | 6,914 | https://github.com/huggingface/diffusers/issues/6914 | https://api.github.com/repos/huggingface/diffusers/issues/6914 | 'tuple' object has no attribute 'shape' | enable_model_cpu_offload() breaks IP-Adapter-FaceID: AttributeError | ### Describe the bug
The community pipeline example breaks if `pipeline.enable_model_cpu_offload()` is added.
Error:
`'tuple' object has no attribute 'shape'`
Seems related to #6832
Bypasses the decoupling of encoder_hidden_states.
@fabiorigano Could you please look into this? Your mind is fresh on the... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-02-08T22:55:43Z | 2024-02-09T12:40:52Z | 2024-02-09T12:40:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | davoodwadi | 62,418,145 | MDQ6VXNlcjYyNDE4MTQ1 | User | false |
huggingface/diffusers | 2,126,410,465 | I_kwDOHa8MBc5-vnLh | 6,916 | https://github.com/huggingface/diffusers/issues/6916 | https://api.github.com/repos/huggingface/diffusers/issues/6916 | the hub_token parameter value will be exposed via report_to=wandb in training examples | hello, we noticed that the full cmdline is printed to wandb's overview page and then there's other Artifacts where it is later visible. this has likely resulted in a few leaks of user info in the past, but it's probably no big deal.
anyway, i resolved this on my end with:
```py
if args.report_to == "wandb" a... | closed | completed | false | 5 | [] | [] | 2024-02-09T03:31:12Z | 2024-02-09T05:51:30Z | 2024-02-09T04:38:28Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.