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 | 1,289,751,315 | I_kwDOHa8MBc5M4AsT | 49 | https://github.com/huggingface/diffusers/issues/49 | https://api.github.com/repos/huggingface/diffusers/issues/49 | Missing schema: Invalid URL | Throwing `MissingSchema: Invalid URL` on trying to load a sample `DiffusionPipeline` from the huggingface hub. It does download the model files but fails right after. Maybe it's a problem with the model?
I tried to run:
```python
pipeline = DiffusionPipeline.from_pretrained("fusing/glide-base")
```
Error:
`... | closed | completed | false | 2 | [] | [] | 2022-06-30T08:01:42Z | 2022-07-21T19:05:47Z | 2022-07-21T19:05:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | creatorrr | 931,887 | MDQ6VXNlcjkzMTg4Nw== | User | false |
huggingface/diffusers | 1,370,800,458 | I_kwDOHa8MBc5RtMFK | 490 | https://github.com/huggingface/diffusers/issues/490 | https://api.github.com/repos/huggingface/diffusers/issues/490 | Memory Issues: Compvis/Stable-Diffusion works, diffusers does not | ### Describe the bug
Hi, I've been able to run the stable diffusion scripts from [https://github.com/CompVis/stable-diffusion](https://github.com/CompVis/stable-diffusion) perfectly fine.
However, the [basic text to image example given on the readme of the diffusers page ](https://github.com/huggingface/diffusers/blo... | closed | completed | false | 4 | [
"bug"
] | [
"patrickvonplaten"
] | 2022-09-13T03:31:11Z | 2022-09-13T22:35:43Z | 2022-09-13T22:06:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | PhilippeNguyen | 7,034,878 | MDQ6VXNlcjcwMzQ4Nzg= | User | false |
huggingface/diffusers | 1,881,714,559 | I_kwDOHa8MBc5wKK9_ | 4,900 | https://github.com/huggingface/diffusers/issues/4900 | https://api.github.com/repos/huggingface/diffusers/issues/4900 | Inpaint: Choose starting point | In A1111 you can choose what the inpainted region is initially filled with. E.g., original or latent nothing. Is there an example of this with diffusers? | closed | completed | false | 2 | [
"stale"
] | [] | 2023-09-05T10:49:20Z | 2023-11-08T15:08:24Z | 2023-11-08T15:08:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rb-synth | 135,021,519 | U_kgDOCAxDzw | User | false |
huggingface/diffusers | 1,881,864,273 | I_kwDOHa8MBc5wKvhR | 4,902 | https://github.com/huggingface/diffusers/issues/4902 | https://api.github.com/repos/huggingface/diffusers/issues/4902 | Savetensors loading | ```model_path = "Train/stable-diffusion/models/happi.safetensors"
torch_dtype = torch.float16
use_safetensors = True
pipe = StableDiffusionPipeline.from_pretrained(model_path, torch_dtype=torch_dtype, use_safetensors=use_safetensors)
pipe = pipe.to(device)```
i plan to load the savetensors in locally, but the ... | closed | completed | false | 3 | [] | [] | 2023-09-05T12:19:13Z | 2023-09-07T07:48:15Z | 2023-09-07T05:55:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Justinfungi | 79,019,929 | MDQ6VXNlcjc5MDE5OTI5 | User | false |
huggingface/diffusers | 1,882,001,284 | I_kwDOHa8MBc5wLQ-E | 4,904 | https://github.com/huggingface/diffusers/issues/4904 | https://api.github.com/repos/huggingface/diffusers/issues/4904 | Make GLIGEN modular | **Is your feature request related to a problem? Please describe.**
I have been experimenting with the [StableDiffusionGLIGENPipeline](https://github.com/huggingface/diffusers/blob/3105c710ba16fa2cf54d8deb158099a4146da511/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_gligen.py) and finding it effec... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-09-05T13:36:27Z | 2023-11-13T15:06:35Z | 2023-11-13T15:06:35Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chuzhdontcode | 50,682,966 | MDQ6VXNlcjUwNjgyOTY2 | User | false |
huggingface/diffusers | 1,882,402,255 | I_kwDOHa8MBc5wMy3P | 4,905 | https://github.com/huggingface/diffusers/issues/4905 | https://api.github.com/repos/huggingface/diffusers/issues/4905 | How to convert pretrained SDXL .safetensors model to diffusers folder format | As SDXL is gaining adoption, more and more community based models pop up that that are just saved as a .safetensors file. E.g the popular Realistic Vision: https://civitai.com/models/139562?modelVersionId=154590
When running train_dreambooth_lora_sdxl.py, the training script expects the diffusers folder format to ac... | closed | completed | false | 3 | [] | [] | 2023-09-05T17:01:27Z | 2023-09-06T09:55:54Z | 2023-09-06T09:55:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | agcty | 12,101,091 | MDQ6VXNlcjEyMTAxMDkx | User | false |
huggingface/diffusers | 1,882,472,435 | I_kwDOHa8MBc5wND_z | 4,906 | https://github.com/huggingface/diffusers/issues/4906 | https://api.github.com/repos/huggingface/diffusers/issues/4906 | How to check whether the image is flagged as inappropriate automated? | Is there a way to know whether the generated image (without seeing it) was flagged as inappropriate? | closed | completed | false | 3 | [] | [] | 2023-09-05T17:51:07Z | 2023-09-07T05:49:46Z | 2023-09-07T05:49:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sarmientoj24 | 8,830,319 | MDQ6VXNlcjg4MzAzMTk= | User | false |
huggingface/diffusers | 1,882,479,647 | I_kwDOHa8MBc5wNFwf | 4,907 | https://github.com/huggingface/diffusers/issues/4907 | https://api.github.com/repos/huggingface/diffusers/issues/4907 | New inpainting model warning with SDXL? | ### Describe the bug
config attributes {'decay': 0.9999, 'inv_gamma': 1.0, 'min_decay': 0.0, 'optimization_step': 37000, 'power': 0.6666666666666666, 'update_after_step': 0, 'use_ema_warmup': False} were passed to UNet2DConditionModel, but are no
t expected and will be ignored. Please verify your config.json configu... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-09-05T17:56:43Z | 2023-11-09T15:06:04Z | 2023-11-09T15:06:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JohanHuynh0130 | 86,690,620 | MDQ6VXNlcjg2NjkwNjIw | User | false |
huggingface/diffusers | 1,882,840,245 | I_kwDOHa8MBc5wOdy1 | 4,908 | https://github.com/huggingface/diffusers/issues/4908 | https://api.github.com/repos/huggingface/diffusers/issues/4908 | Combined pipeline's `from_pretrained` method: does it make sense to allow passing prior pipeline object? | Should we enable the below API for combined pipelines?
```python
pipe_prior = DiffusionPipeline.from_pretrained(custom_prior_checkpoint, ...)
pipe = AutoPipelineForText2Image.from_pretrained("kandinsky-community/kandinsky-2-2-decoder", prior_pipe = pipe_prior, ...)
```
currently, we will automatically load the ... | closed | completed | false | 1 | [] | [] | 2023-09-05T23:13:48Z | 2023-09-13T23:29:49Z | 2023-09-13T23:29:49Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 1,882,894,230 | I_kwDOHa8MBc5wOq-W | 4,909 | https://github.com/huggingface/diffusers/issues/4909 | https://api.github.com/repos/huggingface/diffusers/issues/4909 | change train_text_to_image.py to use comet_ml tracker | Hello,
I would like to use train_text_to_image.py script and I see is using wandb or tensordboard. How can this be edited to use cometml? I see accelerate support this tracker but not sure how to check parts of the code and init it with my apy_key and workspace
thank you | closed | completed | false | 1 | [] | [] | 2023-09-06T00:01:32Z | 2023-09-07T05:47:31Z | 2023-09-07T05:47:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | segalinc | 9,353,106 | MDQ6VXNlcjkzNTMxMDY= | User | false |
huggingface/diffusers | 1,370,966,003 | I_kwDOHa8MBc5Rt0fz | 491 | https://github.com/huggingface/diffusers/issues/491 | https://api.github.com/repos/huggingface/diffusers/issues/491 | Script for converting SD weights from compvis to diffusers | It is desirable to have a script for converting SD weights from compvis to diffusers and vice versa. People are already finetuning models, but they do it on different bases. Examples: [for diffusers](https://huggingface.co/hakurei/waifu-diffusion), [for compvis](https://huggingface.co/naclbit/trinart_stable_diffusion_v... | closed | completed | false | 17 | [
"good first issue"
] | [
"patil-suraj"
] | 2022-09-13T07:08:45Z | 2023-04-19T07:59:33Z | 2022-09-27T10:15:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Inkorak | 52,286,717 | MDQ6VXNlcjUyMjg2NzE3 | User | false |
huggingface/diffusers | 1,883,288,462 | I_kwDOHa8MBc5wQLOO | 4,911 | https://github.com/huggingface/diffusers/issues/4911 | https://api.github.com/repos/huggingface/diffusers/issues/4911 | Error in loading custom text_encoder in StableDiffusionPipeline.from_single_file | ### Describe the bug
When i am doing
pipe = diffusers.StableDiffusionPipeline.from_single_file(
model_path,
load_safety_checker=False,
scheduler_type='dpm',
text_encoder=text_encoder
)
I am getting an error , this is weird because when I am specifying a text_encoder in from_pretrained() , it i... | closed | completed | false | 8 | [
"bug"
] | [
"DN6"
] | 2023-09-06T05:55:52Z | 2023-09-26T14:24:39Z | 2023-09-26T14:24:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Garry435 | 143,331,750 | U_kgDOCIsRpg | User | false |
huggingface/diffusers | 1,883,338,127 | I_kwDOHa8MBc5wQXWP | 4,912 | https://github.com/huggingface/diffusers/issues/4912 | https://api.github.com/repos/huggingface/diffusers/issues/4912 | Ability to convert custom CivitAI checkpoints of SDXL to diffusers | **Is your feature request related to a problem? Please describe.**
https://github.com/TheLastBen/fast-stable-diffusion/issues/2528#issuecomment-1705740810
**Describe the solution you'd like**
convert custom civitai checkpoint of sdxl 1.0 to diffusers
**Describe alternatives you've considered**
I tried using T... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-09-06T06:34:41Z | 2023-11-09T15:06:02Z | 2023-11-09T15:06:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MaxTran96 | 15,274,946 | MDQ6VXNlcjE1Mjc0OTQ2 | User | false |
huggingface/diffusers | 1,883,495,417 | I_kwDOHa8MBc5wQ9v5 | 4,913 | https://github.com/huggingface/diffusers/issues/4913 | https://api.github.com/repos/huggingface/diffusers/issues/4913 | controlnet-openpose-sdxl-1.0 | ### Model/Pipeline/Scheduler description
Is this model coming?
https://huggingface.co/thibaud/controlnet-openpose-sdxl-1.0
There is such a model.
### Open source status
- [ ] The model implementation is available
- [ ] The model weights are available (Only relevant if addition is not a scheduler).
### ... | closed | completed | false | 3 | [] | [] | 2023-09-06T08:17:17Z | 2023-09-06T11:36:52Z | 2023-09-06T11:36:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Yakonrus | 4,978,662 | MDQ6VXNlcjQ5Nzg2NjI= | User | false |
huggingface/diffusers | 1,883,500,734 | I_kwDOHa8MBc5wQ_C- | 4,914 | https://github.com/huggingface/diffusers/issues/4914 | https://api.github.com/repos/huggingface/diffusers/issues/4914 | StableDiffusionXLControlNetInpaintPipeline | ### Model/Pipeline/Scheduler description
At the moment it is not possible to use ControlNet with standard models at SD XL inpaint pipeline.
StableDiffusionControlNetInpaintPipeline
Will there be an analogue for SD XL?
If yes, approximately when?
### Open source status
- [ ] The model implementation is ava... | closed | completed | false | 4 | [] | [] | 2023-09-06T08:20:39Z | 2023-09-07T05:44:44Z | 2023-09-07T05:44:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Yakonrus | 4,978,662 | MDQ6VXNlcjQ5Nzg2NjI= | User | false |
huggingface/diffusers | 1,883,657,744 | I_kwDOHa8MBc5wRlYQ | 4,915 | https://github.com/huggingface/diffusers/issues/4915 | https://api.github.com/repos/huggingface/diffusers/issues/4915 | Finetuning image-to-image StableDiffusionXLImg2ImgPipeline on a custom dataset | Hi,
Is there any example/tutorial/script for fine-tuning StableDiffusionXLImg2ImgPipeline?
It would really help to get up to speed with custom datasets.
Is there any current example or script that could be easily adapted for the purpose?
Thanks in advance, | closed | completed | false | 2 | [] | [] | 2023-09-06T09:52:23Z | 2023-09-07T05:39:13Z | 2023-09-07T05:39:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | miquel-espinosa | 34,089,118 | MDQ6VXNlcjM0MDg5MTE4 | User | false |
huggingface/diffusers | 1,883,695,803 | I_kwDOHa8MBc5wRuq7 | 4,916 | https://github.com/huggingface/diffusers/issues/4916 | https://api.github.com/repos/huggingface/diffusers/issues/4916 | Enabling gradient checkpointing and deepspeed ZeRO3 raise train failure | ### Describe the bug
When using DeepSpeed ZeRO3 and enabling unet gradient checkpointing at the same time, training with diffusers <= 0.16.1 works fine but training with diffusers >= 0.17.0 will raise following exception:
(I’ve checked 0.15.1, 0.16.1, 0.17.0, 0.17.1, 0.18.2, 0.19.1, and 0.20.2)
```
Traceback (m... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-09-06T10:15:32Z | 2023-11-09T15:06:00Z | 2023-11-09T15:06:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ryanzhangfan | 16,065,909 | MDQ6VXNlcjE2MDY1OTA5 | User | false |
huggingface/diffusers | 1,883,847,759 | I_kwDOHa8MBc5wSTxP | 4,918 | https://github.com/huggingface/diffusers/issues/4918 | https://api.github.com/repos/huggingface/diffusers/issues/4918 | StableDiffusionXLControlNetInpaintPipeline failed outpainting attempts | Hi,
Using the new pipeline I am trying to take an object with transparent background and outpaint its backgorund.
I copied the `flatten` method from images, apply to the image and create a mask using a custom method that I found in sd-ui code in cases in which a mask is not provided ( that's how it worked in the UI )... | closed | completed | false | 0 | [] | [] | 2023-09-06T11:50:13Z | 2023-09-06T17:47:36Z | 2023-09-06T17:47:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sebi75 | 36,008,268 | MDQ6VXNlcjM2MDA4MjY4 | User | false |
huggingface/diffusers | 1,883,862,456 | I_kwDOHa8MBc5wSXW4 | 4,919 | https://github.com/huggingface/diffusers/issues/4919 | https://api.github.com/repos/huggingface/diffusers/issues/4919 | Multi Lora support using "fuse_lora" | I have experimented with the "fuse_lora" function to check if multi-lora can be supported, and it works.
I can see all 3 Lora influencing results.
```
from diffusers import DiffusionPipeline
import torch
pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.f... | closed | completed | false | 22 | [
"stale"
] | [] | 2023-09-06T11:59:10Z | 2024-08-16T01:59:24Z | 2023-11-26T15:07:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | adhikjoshi | 11,740,719 | MDQ6VXNlcjExNzQwNzE5 | User | false |
huggingface/diffusers | 1,883,898,655 | I_kwDOHa8MBc5wSgMf | 4,920 | https://github.com/huggingface/diffusers/issues/4920 | https://api.github.com/repos/huggingface/diffusers/issues/4920 | multi sd pipeline inference concurrently | Assume there are two different sd pipeline and two gpus. The thing I want is doing inference these two pipelines concurrently, each one on a separate gpu, and then doing something with the outcomes.
This is easy for simple nn.Module like torchvision.models:
```py
import torch
import torchvision
model0 = torchvis... | closed | completed | false | 2 | [] | [] | 2023-09-06T12:21:24Z | 2023-09-11T11:41:13Z | 2023-09-11T11:41:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LSC527 | 34,333,110 | MDQ6VXNlcjM0MzMzMTEw | User | false |
huggingface/diffusers | 1,883,980,551 | I_kwDOHa8MBc5wS0MH | 4,921 | https://github.com/huggingface/diffusers/issues/4921 | https://api.github.com/repos/huggingface/diffusers/issues/4921 | Refiner erases details | ### Describe the bug
I'm experimenting with refiner and fine-tuned models, and I'm discovered that refiner erases details.
### Reproduction
1) download https://civitai.com/models/122822/crystal-clear-xl?modelVersionId=133832
2) run in notebook
```
from diffusers import DiffusionPipeline, StableDiffusionXL... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2023-09-06T13:06:40Z | 2023-11-09T15:05:57Z | 2023-11-09T15:05:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Alexadar | 14,125,937 | MDQ6VXNlcjE0MTI1OTM3 | User | false |
huggingface/diffusers | 1,884,132,535 | I_kwDOHa8MBc5wTZS3 | 4,922 | https://github.com/huggingface/diffusers/issues/4922 | https://api.github.com/repos/huggingface/diffusers/issues/4922 | Will there be an android demo？ | Very honored to experience the ios version of https://github.com/huggingface/swift-coreml-diffusers, could you tell me whether there will be an android version?
thanks very much！
**Is your feature request related to a problem? Please describe.**
A clear and concise description of what the problem is. Ex. I'm alway... | closed | completed | false | 1 | [] | [] | 2023-09-06T14:24:56Z | 2023-09-07T05:30:50Z | 2023-09-07T05:30:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AndreaChiChengdu | 31,582,949 | MDQ6VXNlcjMxNTgyOTQ5 | User | false |
huggingface/diffusers | 1,884,139,782 | I_kwDOHa8MBc5wTbEG | 4,923 | https://github.com/huggingface/diffusers/issues/4923 | https://api.github.com/repos/huggingface/diffusers/issues/4923 | Feature request: Update the pipeline for AudioLDM 2 so that 'transcript' can be consumed and text to speech created | **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 [...]
The current pipeline for AudioLDM 2 does not take in "transcript" field.
Hence, it does not create phonemes and hence does not allow for text-to-speech ge... | open | null | false | 5 | [
"enhancement",
"Good second issue"
] | [] | 2023-09-06T14:28:19Z | 2024-01-31T14:54:12Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | filip-michalsky | 31,483,888 | MDQ6VXNlcjMxNDgzODg4 | User | false |
huggingface/diffusers | 1,884,182,604 | I_kwDOHa8MBc5wTlhM | 4,924 | https://github.com/huggingface/diffusers/issues/4924 | https://api.github.com/repos/huggingface/diffusers/issues/4924 | StableDiffusionSAGPipeline doesn't work with any Euler schedulers | so here's the bug:
lib/python3.10/site-packages/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_sag.py", line 732, in pred_x0
alpha_prod_t = self.scheduler.alphas_cumprod[timestep]
RuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cpu)
here's the code us... | closed | completed | false | 11 | [
"stale"
] | [] | 2023-09-06T14:47:55Z | 2023-11-21T15:06:03Z | 2023-11-21T15:06:03Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,884,285,824 | I_kwDOHa8MBc5wT-uA | 4,925 | https://github.com/huggingface/diffusers/issues/4925 | https://api.github.com/repos/huggingface/diffusers/issues/4925 | Training Controlnet SDXL distributed gives out-of-memory errors | ### Describe the bug
Hi.
I am running the Controlnet SDXL example as it is shown in the `examples` section [[example-link](https://github.com/huggingface/diffusers/blob/main/examples/controlnet/README_sdxl.md)].
I am unable to reproduce the results in a SLURM managed environment, where I have 8 A100 40GB GPUs in the... | closed | completed | false | 14 | [
"bug",
"stale"
] | [] | 2023-09-06T15:40:28Z | 2024-01-09T15:08:31Z | 2024-01-09T15:08:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | miquel-espinosa | 34,089,118 | MDQ6VXNlcjM0MDg5MTE4 | User | false |
huggingface/diffusers | 1,884,856,270 | I_kwDOHa8MBc5wWJ_O | 4,928 | https://github.com/huggingface/diffusers/issues/4928 | https://api.github.com/repos/huggingface/diffusers/issues/4928 | undo 4927 once huggingface/accelerate#1880 put into accelerate release | undo https://github.com/huggingface/diffusers/pull/4927 once huggingface/accelerate#1880 put into accelerate release | closed | completed | false | 0 | [] | [
"williamberman",
"sayakpaul"
] | 2023-09-06T22:49:36Z | 2023-09-11T15:53:54Z | 2023-09-11T15:53:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,885,131,112 | I_kwDOHa8MBc5wXNFo | 4,931 | https://github.com/huggingface/diffusers/issues/4931 | https://api.github.com/repos/huggingface/diffusers/issues/4931 | Mysterious weights when training UNET | I was training [sdxl](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0) UNET base model which was going great until around step 210k when the weights suddenly turned back to their original values and stayed that way. I also tried with the ema version, which didn’t change at all. I also looked at the tens... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-09-07T05:16:07Z | 2023-11-09T15:05:45Z | 2023-11-09T15:05:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yeonsikch | 79,968,598 | MDQ6VXNlcjc5OTY4NTk4 | User | false |
huggingface/diffusers | 1,885,185,140 | I_kwDOHa8MBc5wXaR0 | 4,932 | https://github.com/huggingface/diffusers/issues/4932 | https://api.github.com/repos/huggingface/diffusers/issues/4932 | Dreambooth doesn't return pytorch_model.bin | ### Describe the bug
I successfully trained my DreamBooth model following the [tutorial](https://huggingface.co/docs/diffusers/training/dreambooth), and I pushed it to the [hub](https://huggingface.co/ckandemir/dog) with the validation images.
However, when I call my model from the hub using `pipeline = Diffusi... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-09-07T06:15:19Z | 2023-09-07T06:28:10Z | 2023-09-07T06:28:10Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | canberk17 | 33,362,633 | MDQ6VXNlcjMzMzYyNjMz | User | false |
huggingface/diffusers | 1,885,245,214 | I_kwDOHa8MBc5wXo8e | 4,933 | https://github.com/huggingface/diffusers/issues/4933 | https://api.github.com/repos/huggingface/diffusers/issues/4933 | Dreambooth doesn't return pytorch_model.bin after training succesfully | ### Describe the bug
I had to to point back to this issue #4932 😅, I successfully trained my DreamBooth model following the [tutorial](https://huggingface.co/docs/diffusers/training/dreambooth), and I pushed it to the [hub](https://huggingface.co/ckandemir/dog) with the validation images.
However, when I call... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-09-07T07:04:49Z | 2023-09-07T08:16:26Z | 2023-09-07T07:19:45Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | canberk17 | 33,362,633 | MDQ6VXNlcjMzMzYyNjMz | User | false |
huggingface/diffusers | 1,885,898,616 | I_kwDOHa8MBc5waId4 | 4,935 | https://github.com/huggingface/diffusers/issues/4935 | https://api.github.com/repos/huggingface/diffusers/issues/4935 | Can "lora_layers = AttnProcsLayers(unet.attn_processors) " be used in "train_text_to_image_lora_sdxl.py"? | ### Describe the bug
For training the lora of SD-XL in "train_text_to_image_lora_sdxl.py", unet are prepared by "accelerator.prepare":
`unet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(unet, optimizer, train_dataloader, lr_scheduler)`
However, for SD-1.5/2.1 in "train_text_to_image_lora.py", ... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-09-07T13:26:49Z | 2023-11-09T15:05:44Z | 2023-11-09T15:05:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ZihaoW123 | 47,708,655 | MDQ6VXNlcjQ3NzA4NjU1 | User | false |
huggingface/diffusers | 1,886,110,602 | I_kwDOHa8MBc5wa8OK | 4,936 | https://github.com/huggingface/diffusers/issues/4936 | https://api.github.com/repos/huggingface/diffusers/issues/4936 | Resume Function in Dreambooth Lora Training is Broken since 0.19.0 | ### Describe the bug
Resume functionality in train_dreambooth_lora.py example seems to be completely broken since 0.19.0 and narrowed down to https://github.com/huggingface/diffusers/pull/3778
Reverting 0.18.2 style temp_pipeline does seem to fix textencoder resume but not unet.
### Reproduction
The Repro runs tra... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-09-07T15:22:35Z | 2023-11-08T15:08:22Z | 2023-11-08T15:08:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dreamraster | 144,281,363 | U_kgDOCJmPEw | User | false |
huggingface/diffusers | 1,886,321,874 | I_kwDOHa8MBc5wbvzS | 4,940 | https://github.com/huggingface/diffusers/issues/4940 | https://api.github.com/repos/huggingface/diffusers/issues/4940 | lpw_stable_diffusion truncating prompts? | I still get this even with lpw_stable_diffusion. Is there a problem on this?
```
Token indices sequence length is longer than the specified maximum sequence length for this model (158 > 77). Running this sequence through the model will result in indexing errors
2023-09-08T01:22:15+0800 [WARNING] [runner:stablediff... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-09-07T17:32:46Z | 2024-07-16T12:10:51Z | 2023-11-08T15:08:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sarmientoj24 | 8,830,319 | MDQ6VXNlcjg4MzAzMTk= | User | false |
huggingface/diffusers | 1,886,346,765 | I_kwDOHa8MBc5wb14N | 4,943 | https://github.com/huggingface/diffusers/issues/4943 | https://api.github.com/repos/huggingface/diffusers/issues/4943 | Object Removal in the the new Inpainting SDXL model? | Hi Everyone!
I just started using it and the results are great! However, it has one big issue: It’s hard to remove a big object out of the input. Let’s say I want to empty a room, instead, it will keep throwing random asterisks into the input.
Any ideas how to implement or setting the parameters to achieve this | closed | completed | false | 1 | [] | [] | 2023-09-07T17:53:23Z | 2023-09-08T04:16:54Z | 2023-09-08T04:16:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JohanHuynh0130 | 86,690,620 | MDQ6VXNlcjg2NjkwNjIw | User | false |
huggingface/diffusers | 1,886,904,979 | I_kwDOHa8MBc5wd-KT | 4,945 | https://github.com/huggingface/diffusers/issues/4945 | https://api.github.com/repos/huggingface/diffusers/issues/4945 | [Tests] Add tests for T2I Adapters' training | https://github.com/huggingface/diffusers/pull/4934 is adding a training script for T2I Adapters, thanks to @patil-suraj!
I tried adding a test for the script: https://github.com/huggingface/diffusers/pull/4934/commits/0f646561856f2977e7c11497065ca1149ca360fd. However, since the UNet used for the test is not compati... | closed | completed | false | 0 | [] | [] | 2023-09-08T04:15:51Z | 2023-09-08T14:15:41Z | 2023-09-08T14:15:41Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,887,685,381 | I_kwDOHa8MBc5wg8sF | 4,949 | https://github.com/huggingface/diffusers/issues/4949 | https://api.github.com/repos/huggingface/diffusers/issues/4949 | ProPainter - In and outpainting for video | ### Model/Pipeline/Scheduler description
ProPainter: Improving Propagation and Transformer for Video Inpainting
Object Removal: remove the object(s) from a video.
Watermark Removal: remove the watermark from a video.
Video Completion: complete the masked video.
Video Outpainting: expand the view of a video.
... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-09-08T13:23:16Z | 2023-12-08T15:06:42Z | 2023-12-08T15:06:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,887,789,579 | I_kwDOHa8MBc5whWIL | 4,950 | https://github.com/huggingface/diffusers/issues/4950 | https://api.github.com/repos/huggingface/diffusers/issues/4950 | StableDiffusionXLPipeline to natively support the TextualInversionLoaderMixin. | **Is your feature request related to a problem? Please describe.**
Yes, while working with the StableDiffusionXLPipeline, I found that it didn't natively support the TextualInversionLoaderMixin. As a result, I had to extend the class manually to incorporate this functionality, which isn't ideal for long-term usage and... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-09-08T14:24:31Z | 2023-10-18T19:59:48Z | 2023-10-18T19:59:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alkollo | 722,274 | MDQ6VXNlcjcyMjI3NA== | User | false |
huggingface/diffusers | 1,888,508,951 | I_kwDOHa8MBc5wkFwX | 4,951 | https://github.com/huggingface/diffusers/issues/4951 | https://api.github.com/repos/huggingface/diffusers/issues/4951 | DiffBIR Diffusers Integration [feature request] | The longstanding issue of blurry photos that has been a nightmare for visual designers is starting to be solved using diffusion technology: DiffBir.
https://github.com/XPixelGroup/DiffBIR
| open | null | false | 2 | [
"New pipeline/model",
"community-examples"
] | [] | 2023-09-09T00:30:05Z | 2024-03-15T16:27:04Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | putuoka | 20,880,839 | MDQ6VXNlcjIwODgwODM5 | User | false |
huggingface/diffusers | 1,888,539,471 | I_kwDOHa8MBc5wkNNP | 4,953 | https://github.com/huggingface/diffusers/issues/4953 | https://api.github.com/repos/huggingface/diffusers/issues/4953 | PR Template Typo | In the "who to tag" comment section of the Markdown template that comes up when submitting a PR, the line for documentation-related PRs says
- Docs: @/stevenliu and @/yiyixu
Is this a typo? Neither of these accounts seem to be active in `diffusers`. | closed | completed | false | 1 | [] | [] | 2023-09-09T02:15:25Z | 2023-09-12T17:36:40Z | 2023-09-12T17:36:40Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dg845 | 58,458,699 | MDQ6VXNlcjU4NDU4Njk5 | User | false |
huggingface/diffusers | 1,888,632,188 | I_kwDOHa8MBc5wkj18 | 4,954 | https://github.com/huggingface/diffusers/issues/4954 | https://api.github.com/repos/huggingface/diffusers/issues/4954 | Textual Inversion _pickle.UnpicklingError | ### Describe the bug
After training Textual Inversion, loading the model gives me a ``` _pickle.UnpicklingError: invalid load key, '`' ```
### Reproduction
```
pipe = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16).to("cuda")
text_inv_id = '/home/sam_motamed/di... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-09-09T07:52:24Z | 2024-01-11T11:47:55Z | 2023-09-11T05:39:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sam-motamed | 61,124,645 | MDQ6VXNlcjYxMTI0NjQ1 | User | false |
huggingface/diffusers | 1,888,657,273 | I_kwDOHa8MBc5wkp95 | 4,956 | https://github.com/huggingface/diffusers/issues/4956 | https://api.github.com/repos/huggingface/diffusers/issues/4956 | Upgrade to Diffusers 0.17 and get CUDA OUT OF MEMORY | Without any change when fine-tuning the sd pretrained model, but when I upgrade diffusers from 0.16 to 0.17, I got a CUDA OUT OF MEMORY. My device is a rtx3090. | closed | completed | false | 2 | [] | [] | 2023-09-09T09:13:10Z | 2023-09-15T07:10:12Z | 2023-09-15T07:10:12Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Junsheng121 | 103,119,560 | U_kgDOBiV6yA | User | false |
huggingface/diffusers | 1,889,078,398 | I_kwDOHa8MBc5wmQx- | 4,957 | https://github.com/huggingface/diffusers/issues/4957 | https://api.github.com/repos/huggingface/diffusers/issues/4957 | OSError: [Errno 28] No space left on device - Controlnet + Inpainting-SDXL | ### Describe the bug
- even after using xformers it is not able to run the model
### Reproduction
```
from diffusers import StableDiffusionControlNetInpaintPipeline
controlnet = [
ControlNetModel.from_pretrained(
"diffusers/controlnet-zoe-depth-sdxl-1.0" ,
torch_dtype=torch.float16
... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-09-10T09:49:29Z | 2025-04-28T16:26:22Z | 2023-10-30T15:07:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,889,110,872 | I_kwDOHa8MBc5wmYtY | 4,958 | https://github.com/huggingface/diffusers/issues/4958 | https://api.github.com/repos/huggingface/diffusers/issues/4958 | Run so slow | StableDiffusionPipeline.from_single_file（）
It is too slow to run this code(my model has been downloaded in my own device）.Are there some ways to address it？ | closed | completed | false | 4 | [
"stale"
] | [] | 2023-09-10T11:34:44Z | 2023-10-30T15:07:05Z | 2023-10-30T15:07:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bogao-code | 120,021,873 | U_kgDOBydjcQ | User | false |
huggingface/diffusers | 1,889,118,137 | I_kwDOHa8MBc5wmae5 | 4,960 | https://github.com/huggingface/diffusers/issues/4960 | https://api.github.com/repos/huggingface/diffusers/issues/4960 | train_instruct_pix2pix.py initialized Unet missing bias copy | on train_instruct_pix2pix.py, when using a modified UNet which is initialized from pre-trained checkpoints, on the update of the first convolution layer, the weights are copied but the bias is not. The bias values needs to be copied as well.
need to add the following line to the code below.
` new_conv_in.bias.... | closed | completed | false | 2 | [] | [] | 2023-09-10T11:59:20Z | 2023-09-15T10:03:59Z | 2023-09-15T10:03:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ori-liberman | 89,383,027 | MDQ6VXNlcjg5MzgzMDI3 | User | false |
huggingface/diffusers | 1,889,317,089 | I_kwDOHa8MBc5wnLDh | 4,964 | https://github.com/huggingface/diffusers/issues/4964 | https://api.github.com/repos/huggingface/diffusers/issues/4964 | Converting SDXL models to diffuser | ### Describe the bug
When trying to convert a SDXL to diffuser on Colab and it won't completely finish the process and always ends with ^C as you can see in the image:
I've tried with different SDXL models an... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-09-10T20:42:41Z | 2023-09-14T17:14:07Z | 2023-09-14T17:14:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JoaoMaio | 73,291,213 | MDQ6VXNlcjczMjkxMjEz | User | false |
huggingface/diffusers | 1,889,871,430 | I_kwDOHa8MBc5wpSZG | 4,966 | https://github.com/huggingface/diffusers/issues/4966 | https://api.github.com/repos/huggingface/diffusers/issues/4966 | Allow downloading remote code of subcomponents | It could make a lot of sense to allow people to download remote code along side the model checkpoint, e.g. as proposed here: https://huggingface.co/tomerkeren42/DiffusionExperimental-5/blob/main/unet_model/modeling_unet.py
We could add a new mechanism to the `model_index.json`, possibly coupled with `trust_remote_c... | closed | completed | false | 8 | [
"stale"
] | [] | 2023-09-11T07:40:18Z | 2023-11-13T16:53:48Z | 2023-11-13T16:53:48Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,890,027,233 | I_kwDOHa8MBc5wp4bh | 4,969 | https://github.com/huggingface/diffusers/issues/4969 | https://api.github.com/repos/huggingface/diffusers/issues/4969 | about Import error? diffusers v0.20.1, ImportError: cannot import name 'CrossAttention' | ### Describe the bug
from diffusers.models.attention import CrossAttention, FeedForward, AdaLayerNorm
ImportError: cannot import name 'CrossAttention' from 'diffusers.models.attention'
### Reproduction
from diffusers.models.attention import CrossAttention, FeedForward, AdaLayerNorm
### Logs
_No response_
### Sy... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-09-11T09:04:43Z | 2023-11-24T12:59:32Z | 2023-09-14T07:58:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhanghongyong123456 | 48,466,610 | MDQ6VXNlcjQ4NDY2NjEw | User | false |
huggingface/diffusers | 1,371,222,983 | I_kwDOHa8MBc5RuzPH | 497 | https://github.com/huggingface/diffusers/issues/497 | https://api.github.com/repos/huggingface/diffusers/issues/497 | Typo in StableDiffusionInpaintPipeline.disable_attention_slicing() method | ### Describe the bug
To enable/disable attention slicing, one need to use enable/disable_attention_slicing() method.
but in StableDiffusionInpaintPipeline class, there is misspelling in disable_attention_slicing() method.
`self.enable_attention_slice(None)` should be replaced as `self.enable_attention_slicing(No... | closed | completed | false | 1 | [
"bug"
] | [] | 2022-09-13T10:29:20Z | 2022-09-13T12:25:23Z | 2022-09-13T12:25:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | juno-hwang | 52,326,857 | MDQ6VXNlcjUyMzI2ODU3 | User | false |
huggingface/diffusers | 1,890,028,855 | I_kwDOHa8MBc5wp403 | 4,970 | https://github.com/huggingface/diffusers/issues/4970 | https://api.github.com/repos/huggingface/diffusers/issues/4970 | tile_colorfix in diffusers | for the tile controlnet, in sd-webui-controlnet. It supports [tile-colorfix and sharpness] (https://github.com/Mikubill/sd-webui-controlnet/pull/1434) to fix the color difference problem during tile model inference. How can we implement this in diffusers? | closed | completed | false | 6 | [
"stale"
] | [] | 2023-09-11T09:05:38Z | 2024-05-08T09:09:26Z | 2023-11-22T15:06:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Laidawang | 85,244,566 | MDQ6VXNlcjg1MjQ0NTY2 | User | false |
huggingface/diffusers | 1,890,408,013 | I_kwDOHa8MBc5wrVZN | 4,972 | https://github.com/huggingface/diffusers/issues/4972 | https://api.github.com/repos/huggingface/diffusers/issues/4972 | wuerstchen example code fails to run | ### Describe the bug
While running [wuerstchen pipeline as per official doc](https://huggingface.co/docs/diffusers/main/en/api/pipelines/wuerstchen), the following errors occur:
File /opt/conda/lib/python3.10/site-packages/diffusers/pipelines/wuerstchen/modeling_wuerstchen_diffnext.py:252, in ResBlockStageB.forward... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-09-11T12:32:10Z | 2023-09-14T18:40:18Z | 2023-09-14T18:40:18Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Lime-Cakes | 91,322,985 | MDQ6VXNlcjkxMzIyOTg1 | User | false |
huggingface/diffusers | 1,890,551,415 | I_kwDOHa8MBc5wr4Z3 | 4,974 | https://github.com/huggingface/diffusers/issues/4974 | https://api.github.com/repos/huggingface/diffusers/issues/4974 | converting .safetensor model to diffusers | ### Describe the bug
using the convert_original_stable_diffusion_to_diffusers.py to convert safetensor model to diffusers. but the script always end with this error:
Traceback (most recent call last):
File "convert_original_stable_diffusion_to_diffusers.py", line 154, in <module>
pipe = download_from_origin... | closed | completed | false | 9 | [
"bug"
] | [] | 2023-09-11T13:44:01Z | 2025-06-13T05:46:40Z | 2023-09-12T02:33:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | leebeet | 9,269,981 | MDQ6VXNlcjkyNjk5ODE= | User | false |
huggingface/diffusers | 1,890,664,625 | I_kwDOHa8MBc5wsUCx | 4,975 | https://github.com/huggingface/diffusers/issues/4975 | https://api.github.com/repos/huggingface/diffusers/issues/4975 | LoRa loading is extremely inefficient due to repeated datatype queries | ### Describe the bug
When applying a LoRA state dict that is already loaded into the memory, the `load_lora_weights()` + `unload_lora_weights()` cycle takes ~5.5 seconds which a majority of the time is spent on repeated dtype queries.
### Reproduction
```py
import time
import torch
import safetensors.torch
from... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-09-11T14:37:00Z | 2023-09-12T17:20:25Z | 2023-09-12T17:20:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | isidentical | 47,358,913 | MDQ6VXNlcjQ3MzU4OTEz | User | false |
huggingface/diffusers | 1,890,871,497 | I_kwDOHa8MBc5wtGjJ | 4,977 | https://github.com/huggingface/diffusers/issues/4977 | https://api.github.com/repos/huggingface/diffusers/issues/4977 | Remove checkpoints at all processes in `train_unconditonal.py` example | ### Describe the bug
I encountered the bug on removing the redundant checkpoint.
```bash
09/11/2023 10:33:09 - INFO - accelerate.accelerator - Saving current state to training_runs/diff_0910/checkpoint-12000
{'downsample_type', 'block_out_channels', 'norm_num_groups', 'downsample_padding', 'up_block_types', 'flip... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-09-11T16:32:49Z | 2023-10-30T15:06:59Z | 2023-10-30T15:06:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | santisy | 8,842,080 | MDQ6VXNlcjg4NDIwODA= | User | false |
huggingface/diffusers | 1,890,903,162 | I_kwDOHa8MBc5wtOR6 | 4,978 | https://github.com/huggingface/diffusers/issues/4978 | https://api.github.com/repos/huggingface/diffusers/issues/4978 | Using preprocessor as inpaint_global_harmonious | ### Model/Pipeline/Scheduler description
**Model/Pipeline/Scheduler description**
Automatic1111 supports preprocessor of inpaint_global_harmonious for the Inpaint ControlNet model over web ui. Is this functionality already present in the DIffusers library and if not can it be added?
**Provide useful links fo... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-09-11T16:53:17Z | 2023-10-30T15:06:57Z | 2023-10-30T15:06:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ajkrish95 | 5,734,592 | MDQ6VXNlcjU3MzQ1OTI= | User | false |
huggingface/diffusers | 1,890,969,305 | I_kwDOHa8MBc5wtebZ | 4,981 | https://github.com/huggingface/diffusers/issues/4981 | https://api.github.com/repos/huggingface/diffusers/issues/4981 | `StableDiffusionPipeline`: Failure with OpenvinoEP using Onnxruntime | ### Describe the bug
Facing issues with `OpenvinoEP` with `OnnxStableDiffusionPipeline.from_pretrained` function
### Reproduction
options = SessionOptions()
options.graph_optimization_level = GraphOptimizationLevel.ORT_DISABLE_ALL
onnxPipeline = OnnxStableDiffusionPipeline.from_pretrained(self.model, revision="on... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-09-11T17:39:48Z | 2023-10-30T15:06:55Z | 2023-10-30T15:06:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lalith-mcw | 84,698,958 | MDQ6VXNlcjg0Njk4OTU4 | User | false |
huggingface/diffusers | 1,891,124,787 | I_kwDOHa8MBc5wuEYz | 4,983 | https://github.com/huggingface/diffusers/issues/4983 | https://api.github.com/repos/huggingface/diffusers/issues/4983 | Schedulers to have functions for converting between x0 and epsilon | **Is your feature request related to a problem? Please describe.**
There are some things I have been messing with such as dynamic thresholding and other operations which are typically done to x0/pred_original_sample. There have been a number of different ways I've gone about this from patching scheduler classes to usi... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-09-11T19:23:23Z | 2023-10-30T15:06:54Z | 2023-10-30T15:06:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ethansmith2000 | 98,723,285 | U_kgDOBeJl1Q | User | false |
huggingface/diffusers | 1,891,455,232 | I_kwDOHa8MBc5wvVEA | 4,985 | https://github.com/huggingface/diffusers/issues/4985 | https://api.github.com/repos/huggingface/diffusers/issues/4985 | Training InstructPix2Pix SDXL has OOM issue | ### Describe the bug
Hello,
I used a 40GB A100 machine to train InstructPix2Pix SDXL using the toy data following the instruction.
However, even though I reduce the batch size to 1, I still have the OOM issue. Wondering if anyone has observed the same issue? Thanks.
### Reproduction
```
accelerate launch tra... | closed | completed | false | 16 | [
"bug",
"stale"
] | [] | 2023-09-12T00:05:09Z | 2024-07-02T09:53:58Z | 2023-10-30T15:06:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shugerdou | 10,605,184 | MDQ6VXNlcjEwNjA1MTg0 | User | false |
huggingface/diffusers | 1,891,809,973 | I_kwDOHa8MBc5wwrq1 | 4,988 | https://github.com/huggingface/diffusers/issues/4988 | https://api.github.com/repos/huggingface/diffusers/issues/4988 | StableDiffusionSAGPipeline doesn't work with any Euler schedulers, but I have the fix | Hi, the current StableDiffusionSAGPipeline will not work with any Euler schedulers. I have a detailed solution, so please implement it.
Replace this function:
```
def sag_masking(self, original_latents, attn_map, map_size, t, eps):
# Same masking process as in SAG paper: https://arxiv.org/pdf/2210.00939.p... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-09-12T06:57:14Z | 2023-10-30T15:06:50Z | 2023-10-30T15:06:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,371,536,484 | I_kwDOHa8MBc5Rv_xk | 499 | https://github.com/huggingface/diffusers/issues/499 | https://api.github.com/repos/huggingface/diffusers/issues/499 | Example unconditional training script does not work | ### Describe the bug
When running the [example unconditional training script](https://github.com/huggingface/diffusers/blob/main/examples/unconditional_image_generation/train_unconditional.py)
on GPU using `--mixed_precision=fp16`, the autocasting of accelerate will break:
```
Traceback (most recent call last):
... | closed | completed | false | 6 | [
"bug"
] | [
"anton-l"
] | 2022-09-13T14:19:09Z | 2022-10-04T17:04:07Z | 2022-10-04T17:04:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | c-goldschmidt | 16,539,922 | MDQ6VXNlcjE2NTM5OTIy | User | false |
huggingface/diffusers | 1,892,107,066 | I_kwDOHa8MBc5wx0M6 | 4,991 | https://github.com/huggingface/diffusers/issues/4991 | https://api.github.com/repos/huggingface/diffusers/issues/4991 | The `train_dataloader` was not passed into `acceler.prepare()`. | ### Describe the bug
Hi @sayakpaul,
I encountered a bug in the file located in the folder examples/t2i_adapter/train_t2i_adapter_sdxl.py. At line 1067 of the file, it seems to be a bug if you leave it as follows:
t2iadapter, optimizer, lr_scheduler = accelerator.prepare(t2iadapter, optimizer, lr_scheduler)
It s... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-09-12T09:43:29Z | 2023-09-14T09:23:13Z | 2023-09-14T09:23:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | linhduongtuan | 22,388,092 | MDQ6VXNlcjIyMzg4MDky | User | false |
huggingface/diffusers | 1,892,236,502 | I_kwDOHa8MBc5wyTzW | 4,993 | https://github.com/huggingface/diffusers/issues/4993 | https://api.github.com/repos/huggingface/diffusers/issues/4993 | CLIPImageProjection not in _import_structure on '0.21.0.dev0' | ### Describe the bug
I'm having an error running the StableDiffusionGLIGENTextImagePipeline example. The error says that "module diffusers has no attribute CLIPImageProjection". That's probably because CLIPImageProjection is not added to the _import_structure dictionary on the __init__.py file.
### Reproduction
Try ... | closed | completed | false | 4 | [
"bug"
] | [
"DN6"
] | 2023-09-12T10:41:57Z | 2023-09-30T13:52:19Z | 2023-09-13T13:39:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | MrDasix | 22,715,371 | MDQ6VXNlcjIyNzE1Mzcx | User | false |
huggingface/diffusers | 1,289,759,818 | I_kwDOHa8MBc5M4CxK | 50 | https://github.com/huggingface/diffusers/issues/50 | https://api.github.com/repos/huggingface/diffusers/issues/50 | No attribute 'LatentDiffusionPipeline' | **Error**: ` AttributeError: module 'diffusers' has no attribute 'LatentDiffusionPipeline'`
**On:**
```python
pipeline = DiffusionPipeline.from_pretrained("fusing/latent-diffusion-text2im-large")
```
This happened after the model files got successfully downloaded.
**Environment**:
- `python`: `3.10.5`
- ... | closed | completed | false | 2 | [] | [] | 2022-06-30T08:09:15Z | 2022-07-01T04:27:42Z | 2022-07-01T04:27:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | creatorrr | 931,887 | MDQ6VXNlcjkzMTg4Nw== | User | false |
huggingface/diffusers | 1,892,908,763 | I_kwDOHa8MBc5w037b | 5,000 | https://github.com/huggingface/diffusers/issues/5000 | https://api.github.com/repos/huggingface/diffusers/issues/5000 | Regression 🚨 Unfusing LoRAs is broken after the first fuse/unfuse operation | ### Describe the bug
For some combination of LoRAs, if I try to `pipe.unfuse_lora()`, after a LoRA has already been fused/unfused to the pipe, I get an error:
```py
import torch
from diffusers import StableDiffusionXLPipeline
pipe = StableDiffusionXLPipeline.from_pretrained(
"stabilityai/stable-diffu... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-09-12T16:53:15Z | 2023-09-29T16:00:00Z | 2023-09-29T16:00:00Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | apolinario | 788,417 | MDQ6VXNlcjc4ODQxNw== | User | false |
huggingface/diffusers | 1,893,014,455 | I_kwDOHa8MBc5w1Ru3 | 5,002 | https://github.com/huggingface/diffusers/issues/5002 | https://api.github.com/repos/huggingface/diffusers/issues/5002 | `KDPM2DiscreteScheduler` and `KDPM2AncestralDiscreteScheduler` Missing `use_karras_sigmas` argument | ### Describe the bug
Hey HF team, I'm trying to mimic the A1111 samplers using the table listed [in the docs](https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/schedulers/overview.md) but it seems the [`KDPM2DiscreteScheduler`](https://github.com/huggingface/diffusers/blob/0e0db625d0b7da2e6c2733... | closed | completed | false | 1 | [
"bug"
] | [
"yiyixuxu"
] | 2023-09-12T18:03:16Z | 2023-09-21T23:50:42Z | 2023-09-21T23:50:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CodeCorrupt | 8,364,947 | MDQ6VXNlcjgzNjQ5NDc= | User | false |
huggingface/diffusers | 1,893,080,495 | I_kwDOHa8MBc5w1h2v | 5,004 | https://github.com/huggingface/diffusers/issues/5004 | https://api.github.com/repos/huggingface/diffusers/issues/5004 | SDXL dreambooth can't be resumed from a checkpoint at fp16 training | ### Describe the bug
`train_dreambooth_lora_sdxl.py` can't be resumed from a checkpoint using fp16. The log error is `Attempting to unscale FP16 gradients.`
This is a big blocker from being able to train on the free colab tier since you need fp16 to fit in vram, but also need to resume from checkpoints since it c... | closed | completed | false | 39 | [
"bug",
"stale"
] | [] | 2023-09-12T18:53:54Z | 2024-01-17T05:19:31Z | 2024-01-17T05:19:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | epi-morphism | 74,825,640 | MDQ6VXNlcjc0ODI1NjQw | User | false |
huggingface/diffusers | 1,893,655,086 | I_kwDOHa8MBc5w3uIu | 5,005 | https://github.com/huggingface/diffusers/issues/5005 | https://api.github.com/repos/huggingface/diffusers/issues/5005 | example training scripts end up truncating progress bar on resume | ### Describe the bug
I noticed some discrepancies in the progress bar tracking, as it seemed like the numbers weren't adding up.
### Reproduction
The progress bar begins on a certain step already, with a limited range.
Later, we skip through steps, hoping to meet the resume step. We update the progress bar ... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-09-13T03:29:32Z | 2023-10-30T15:06:45Z | 2023-10-30T15:06:45Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 1,893,767,431 | I_kwDOHa8MBc5w4JkH | 5,006 | https://github.com/huggingface/diffusers/issues/5006 | https://api.github.com/repos/huggingface/diffusers/issues/5006 | SDXL CUDA OOM at end of training with BF16 but not with FP16 | ### Describe the bug
Hi, I'm attempting to run `train_dreambooth_lora_sdxl.py` on SageMaker with an A10 GPU (ml.g5.2xlarge).
When I set `--mixed_precision` to `fp16` the training completes without error, however, when I set it to `bf16`, I get a CUDA out of memory error at the very end of training.
Should I be ... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-09-13T05:27:13Z | 2023-10-30T15:06:43Z | 2023-10-30T15:06:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | austinmw | 12,224,358 | MDQ6VXNlcjEyMjI0MzU4 | User | false |
huggingface/diffusers | 1,893,838,447 | I_kwDOHa8MBc5w4a5v | 5,008 | https://github.com/huggingface/diffusers/issues/5008 | https://api.github.com/repos/huggingface/diffusers/issues/5008 | TryOnDiffusion: A Tale of Two UNets Implementation | ### Model/Pipeline/Scheduler description
Is there anyone working on the "TryOnDiffision: A Tale of Two UNets" implementation in diffusers? It looks like a promising approach for virtual try-on with groundbreaking results. We have been trying to implement it and we welcome contributions from the Hugging Face and diffus... | closed | completed | false | 22 | [
"community-examples",
"stale",
"contributions-welcome"
] | [] | 2023-09-13T06:31:51Z | 2024-12-18T05:48:13Z | 2023-12-26T15:10:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kailashahirwar | 3,731,558 | MDQ6VXNlcjM3MzE1NTg= | User | false |
huggingface/diffusers | 1,893,921,925 | I_kwDOHa8MBc5w4vSF | 5,009 | https://github.com/huggingface/diffusers/issues/5009 | https://api.github.com/repos/huggingface/diffusers/issues/5009 | wuerstchen error | ### Describe the bug
```
Traceback (most recent call last):
File "/home/primeline-ai/atra/w.py", line 28, in <module>
height=1024,
File "/home/primeline-ai/.local/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/home/primeli... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-09-13T07:30:33Z | 2023-09-13T07:36:04Z | 2023-09-13T07:36:03Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | flozi00 | 47,894,090 | MDQ6VXNlcjQ3ODk0MDkw | User | false |
huggingface/diffusers | 1,371,698,965 | I_kwDOHa8MBc5RwncV | 501 | https://github.com/huggingface/diffusers/issues/501 | https://api.github.com/repos/huggingface/diffusers/issues/501 | MPS: scheduler indices must be in the same device as the indexed tensor | Continuation of #239, but specific for `mps`. /cc @FahimF. Being fixed in #450. | closed | completed | false | 1 | [] | [
"pcuenca"
] | 2022-09-13T16:14:59Z | 2022-10-11T12:23:20Z | 2022-10-11T12:23:20Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pcuenca | 1,177,582 | MDQ6VXNlcjExNzc1ODI= | User | false |
huggingface/diffusers | 1,894,099,245 | I_kwDOHa8MBc5w5akt | 5,012 | https://github.com/huggingface/diffusers/issues/5012 | https://api.github.com/repos/huggingface/diffusers/issues/5012 | LoRA can not train text encoder in "train_text_to_image_lora_sdxl.py"? | ### Describe the bug
`train_text_to_image_lora.py` "`--train_text_encoder`" is work, but `train_text_to_image_lora_sdxl.py` "`--train_text_encoder`" is not work.
### Reproduction
accelerate launch train_dreambooth_lora_sdxl.py --pretrained_model_name_or_path="stabilityai/stable-diffusion-xl-base-1.0" --pretrained_... | closed | completed | false | 14 | [
"bug"
] | [] | 2023-09-13T09:17:19Z | 2023-11-07T08:44:33Z | 2023-09-18T05:32:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Kevin7720 | 50,312,749 | MDQ6VXNlcjUwMzEyNzQ5 | User | false |
huggingface/diffusers | 1,894,225,302 | I_kwDOHa8MBc5w55WW | 5,014 | https://github.com/huggingface/diffusers/issues/5014 | https://api.github.com/repos/huggingface/diffusers/issues/5014 | Not getting the results from dreambooth lora training | accelerate launch train_dreambooth_lora_sdxl.py --pretrained_model_name_or_path="stabilityai/stable-diffusion-xl-base-1.0" --instance_data_dir="/home/test/testt" --pretrained_vae_model_name_or_path="madebyollin/sdxl-vae-fp16-fix" --output_dir="/home/test/sandeep" --mixed_precision="fp16" --instance_prompt="a photo of s... | closed | completed | false | 2 | [] | [] | 2023-09-13T10:23:49Z | 2023-10-09T07:37:52Z | 2023-09-13T10:48:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Sandeep-Narahari | 46,187,761 | MDQ6VXNlcjQ2MTg3NzYx | User | false |
huggingface/diffusers | 1,894,596,555 | I_kwDOHa8MBc5w7T_L | 5,016 | https://github.com/huggingface/diffusers/issues/5016 | https://api.github.com/repos/huggingface/diffusers/issues/5016 | SDXL dreambooth LoRA can not train text_encoders in "train_dreambooth_lora_sdxl.py" | ### Describe the bug
While enabling `--train_text_encoder` in the `train_dreambooth_lora_sdxl.py` script, it initializes two text encoder parameters but its require_grad is False. Due to this, the parameters are not being backpropagated and updated.
![image](https://github.com/huggingface/diffusers/assets/54628184... | closed | completed | false | 11 | [
"bug",
"stale"
] | [
"sayakpaul"
] | 2023-09-13T13:57:43Z | 2023-11-29T15:06:13Z | 2023-11-29T15:06:13Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shyammarjit | 54,628,184 | MDQ6VXNlcjU0NjI4MTg0 | User | false |
huggingface/diffusers | 1,894,726,352 | I_kwDOHa8MBc5w7zrQ | 5,017 | https://github.com/huggingface/diffusers/issues/5017 | https://api.github.com/repos/huggingface/diffusers/issues/5017 | Can we have control mode for controlnet? | **Is your feature request related to a problem? Please describe.**
I'd like more options for controlnet usage
**Describe the solution you'd like**
implement this: https://github.com/Mikubill/sd-webui-controlnet/issues/1011
**Describe alternatives you've considered**
Implementing it myself
**Additional conte... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-09-13T15:01:02Z | 2023-12-12T07:47:11Z | 2023-10-30T15:06:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,895,486,214 | I_kwDOHa8MBc5w-tMG | 5,022 | https://github.com/huggingface/diffusers/issues/5022 | https://api.github.com/repos/huggingface/diffusers/issues/5022 | Importing UNet2DModel from diffusers in version 0.21.0 causes error name 'PreTrainedTokenizer' is not defined | ### Describe the bug
When running the line:
```py
from diffusers import UNet2DModel
```
the error
`name 'PreTrainedTokenizer' is not defined`
is thrown, but only in the current latest version v0.21.0 and not version v0.20.2 before that
I thought that it could've been an API change, but it doesn't look that ... | closed | completed | false | 3 | [
"bug"
] | [
"DN6"
] | 2023-09-14T01:18:17Z | 2024-07-25T02:06:24Z | 2023-09-14T09:42:11Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | RandomGamingDev | 83,996,185 | MDQ6VXNlcjgzOTk2MTg1 | User | false |
huggingface/diffusers | 1,895,498,634 | I_kwDOHa8MBc5w-wOK | 5,023 | https://github.com/huggingface/diffusers/issues/5023 | https://api.github.com/repos/huggingface/diffusers/issues/5023 | WuerstchenCombinedPipeline Import error in version 0.21.0 | ### Describe the bug
Failed to import import `WuerstchenCombinedPipeline` from `diffusers.pipelines.wuerstchen`,
Tested on Google Colab
### Reproduction
import torch
from diffusers import AutoPipelineForText2Image
### Logs
```shell
--------------------------------------------------------------------... | closed | completed | false | 13 | [
"bug",
"stale"
] | [] | 2023-09-14T01:35:16Z | 2023-11-08T15:08:06Z | 2023-11-08T15:08:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rupeshs | 3,255,994 | MDQ6VXNlcjMyNTU5OTQ= | User | false |
huggingface/diffusers | 1,895,610,176 | I_kwDOHa8MBc5w_LdA | 5,025 | https://github.com/huggingface/diffusers/issues/5025 | https://api.github.com/repos/huggingface/diffusers/issues/5025 | ImportError: cannot import name 'randn_tensor' from 'diffusers.utils' | ### Describe the bug
In the newest version, whenever you run: from diffusers.utils import randn_tensor
you will get the error: importError: cannot import name 'randn_tensor' from 'diffusers.utils'
But you can fix this by adding "from .torch_utils import randn_tensor" to your pythonxx/site-packages/diffusers/utils/_... | closed | completed | false | 5 | [
"bug"
] | [
"DN6"
] | 2023-09-14T03:55:01Z | 2023-11-06T12:16:21Z | 2023-09-14T07:44:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Zerg-Overmind | 35,461,133 | MDQ6VXNlcjM1NDYxMTMz | User | false |
huggingface/diffusers | 1,895,662,747 | I_kwDOHa8MBc5w_YSb | 5,026 | https://github.com/huggingface/diffusers/issues/5026 | https://api.github.com/repos/huggingface/diffusers/issues/5026 | KeyError: 'text_encoder' in 0.21.0 | ### Describe the bug
The release of 0.21.0 broke my previously working refiner code. I'm getting this error:
```
AppData\Roaming\Python\Python310\site-packages\diffusers\pipelines\pipeline_utils.py", line 1255, in enable_model_cpu_offload
model = all_model_components.pop(model_str)
KeyError: 'text_encoder'... | closed | completed | false | 1 | [
"bug"
] | [
"DN6"
] | 2023-09-14T04:55:13Z | 2023-09-14T14:41:27Z | 2023-09-14T14:41:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,895,813,225 | I_kwDOHa8MBc5w_9Bp | 5,028 | https://github.com/huggingface/diffusers/issues/5028 | https://api.github.com/repos/huggingface/diffusers/issues/5028 | TypeError: StableDiffusionXLReferencePipeline.__call__.<locals>.hacked_DownBlock2D_forward() got an unexpected keyword argument 'scale' | ### Describe the bug
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
Cell In[1], line 18
16 pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config)
17 seed = torch.manual_seed(10240... | closed | completed | false | 15 | [
"bug",
"stale"
] | [] | 2023-09-14T06:29:53Z | 2024-02-04T21:16:46Z | 2024-02-04T21:16:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yanchaoguo | 11,827,802 | MDQ6VXNlcjExODI3ODAy | User | false |
huggingface/diffusers | 1,895,851,210 | I_kwDOHa8MBc5xAGTK | 5,029 | https://github.com/huggingface/diffusers/issues/5029 | https://api.github.com/repos/huggingface/diffusers/issues/5029 | AttributeError: 'AutoencoderTinyOutput' object has no attribute 'latent_dist' | ### Describe the bug
AttributeError Traceback (most recent call last)
Cell In[2], line 3
1 seed = torch.manual_seed(10240)
2 pipe.vae = AutoencoderTiny.from_pretrained("madebyollin/taesdxl", torch_dtype=torch.float16, local_files_only=True).to("cuda")
----> 3 result_img = pip... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-09-14T06:53:39Z | 2023-09-14T08:53:26Z | 2023-09-14T08:53:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yanchaoguo | 11,827,802 | MDQ6VXNlcjExODI3ODAy | User | false |
huggingface/diffusers | 1,895,989,275 | I_kwDOHa8MBc5xAoAb | 5,031 | https://github.com/huggingface/diffusers/issues/5031 | https://api.github.com/repos/huggingface/diffusers/issues/5031 | [Tests] Add tests for refiner architecture of SDXL | Currently we don't really test the refiner architecture in SDXL for StableDiffusionXLImg2Img, but only the case when xl-base is used for img2img.
We should add a new fast testing class below here: https://github.com/huggingface/diffusers/blob/19edca82f1ff194c07317369a92b470dbae97f34/tests/pipelines/stable_diffusion... | closed | completed | false | 6 | [] | [
"DN6"
] | 2023-09-14T08:08:27Z | 2023-09-15T10:36:09Z | 2023-09-15T10:36:09Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,895,993,470 | I_kwDOHa8MBc5xApB- | 5,032 | https://github.com/huggingface/diffusers/issues/5032 | https://api.github.com/repos/huggingface/diffusers/issues/5032 | How to unfuse_lora only the first one after I have added multiple lora? | base.load_lora_weights("models/safetensors/SDXL/国风插画SDXL.safetensors")
base.fuse_lora(lora_scale=.7)
base.load_lora_weights("models/safetensors/SDXL/sd_xl_offset_example-lora_1.0.safetensors")
base.fuse_lora(lora_scale=.8)
Now, When I execute unfuse_lora() only the most recent one has been unfuse .
so,how to un... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-09-14T08:10:46Z | 2023-10-30T15:06:34Z | 2023-10-30T15:06:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yanchaoguo | 11,827,802 | MDQ6VXNlcjExODI3ODAy | User | false |
huggingface/diffusers | 1,896,059,652 | I_kwDOHa8MBc5xA5ME | 5,034 | https://github.com/huggingface/diffusers/issues/5034 | https://api.github.com/repos/huggingface/diffusers/issues/5034 | Loading checkpoints may cause the progress bar to display incorrectly | ### Describe the bug
The part of the training script about loading checkpoints. I noticed that when ```step<resume_step```, ```progress_bar.update(1) ```happened. Is this correct? In my understanding this causes the progress bar to display incorrectly. The progress bar marks the number of gradient descents: ```(global... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-09-14T08:42:46Z | 2023-10-30T15:06:32Z | 2023-10-30T15:06:32Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aihao2000 | 51,043,929 | MDQ6VXNlcjUxMDQzOTI5 | User | false |
huggingface/diffusers | 1,896,155,695 | I_kwDOHa8MBc5xBQov | 5,037 | https://github.com/huggingface/diffusers/issues/5037 | https://api.github.com/repos/huggingface/diffusers/issues/5037 | An error is raised when calling the second validation image with the argument --validation_image | ### Describe the bug
An error is still raised when calling the second validation image with the argument --validation_image like this: "train_t2i_adapter_sdxl.py: error: unrecognized arguments: ./conditioning_image_2.png."
I encounter the error shown in the attachment.
<img width="873" alt="Untitled" src="https... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-09-14T09:31:56Z | 2023-09-15T11:20:52Z | 2023-09-15T11:20:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | linhduongtuan | 22,388,092 | MDQ6VXNlcjIyMzg4MDky | User | false |
huggingface/diffusers | 1,896,808,197 | I_kwDOHa8MBc5xDv8F | 5,042 | https://github.com/huggingface/diffusers/issues/5042 | https://api.github.com/repos/huggingface/diffusers/issues/5042 | How to give number of inference steps to Wuerstchen prior pipeline | **this below working with default DEFAULT_STAGE_C_TIMESTEPS but it always generates with exactly 29 number of prior inference steps**
```
prior_output = prior_pipeline(
prompt=prompt,
height=height,
width=width,
num_inference_steps=prior_num_inference_steps,
timesteps=DEF... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-09-14T15:21:31Z | 2023-09-20T07:41:19Z | 2023-09-20T07:41:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FurkanGozukara | 19,240,467 | MDQ6VXNlcjE5MjQwNDY3 | User | false |
huggingface/diffusers | 1,896,823,225 | I_kwDOHa8MBc5xDzm5 | 5,043 | https://github.com/huggingface/diffusers/issues/5043 | https://api.github.com/repos/huggingface/diffusers/issues/5043 | Add LoRA Finetuning for Würstchen | **Is your feature request related to a problem? Please describe.**
It would be nice to easily finetune [Würstchen](https://huggingface.co/docs/diffusers/api/pipelines/wuerstchen) using Diffusers which we released yesterday with this [blog post](https://huggingface.co/blog/wuerstchen)
**Describe the solution you'd l... | closed | completed | false | 1 | [
"stale"
] | [] | 2023-09-14T15:29:36Z | 2023-10-30T15:06:28Z | 2023-10-30T15:06:28Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dome272 | 61,938,694 | MDQ6VXNlcjYxOTM4Njk0 | User | false |
huggingface/diffusers | 1,896,843,145 | I_kwDOHa8MBc5xD4eJ | 5,044 | https://github.com/huggingface/diffusers/issues/5044 | https://api.github.com/repos/huggingface/diffusers/issues/5044 | Kandinsky 2.2 fails to load | ### Describe the bug
i've tried both `DiffusionPipeline` and `AutoPipelineForText2Image` and Kandinsky 2.2 fails to load:
```python
sd_model = diffusers.DiffusionPipeline.from_pretrained('models/Diffusers/models--kandinsky-community--kandinsky-2-2-decoder')
sd_model = diffusers.AutoPipelineForText2Image.from_pret... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-09-14T15:40:28Z | 2023-09-14T20:06:57Z | 2023-09-14T20:06:56Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 1,897,511,268 | I_kwDOHa8MBc5xGblk | 5,046 | https://github.com/huggingface/diffusers/issues/5046 | https://api.github.com/repos/huggingface/diffusers/issues/5046 | DragonDiffusion: Enabling Drag-style Manipulation on Diffusion Models [No code yet] | ### Model/Pipeline/Scheduler description
The authors have developed this technique as a fine-grained image editing scheme based on the strong correspondence of intermediate features in diffusion models. To this end, they designed a classifier-guidance-based method to transform the editing signals into gradients via fe... | closed | completed | false | 1 | [
"community-examples",
"stale"
] | [] | 2023-09-15T00:16:30Z | 2023-10-30T15:06:26Z | 2023-10-30T15:06:26Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | clarencechen | 8,482,341 | MDQ6VXNlcjg0ODIzNDE= | User | false |
huggingface/diffusers | 1,897,885,576 | I_kwDOHa8MBc5xH2-I | 5,049 | https://github.com/huggingface/diffusers/issues/5049 | https://api.github.com/repos/huggingface/diffusers/issues/5049 | In Pipelines can't load LoRA, previously it could | `pipe.load_lora_weights(".", weight_name="/content/model/lora/mjty-000016.safetensors")`
```
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-i... | closed | completed | false | 2 | [] | [] | 2023-09-15T07:40:08Z | 2023-11-06T13:28:22Z | 2023-09-15T08:45:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Damarcreative | 106,896,172 | U_kgDOBl8bLA | User | false |
huggingface/diffusers | 1,897,925,526 | I_kwDOHa8MBc5xIAuW | 5,050 | https://github.com/huggingface/diffusers/issues/5050 | https://api.github.com/repos/huggingface/diffusers/issues/5050 | change config in LoRA changing | **Is your feature request related to a problem? Please describe.**
When training a LoRA, sometimes I get the impression that the LoRA model didn't fully capture the characteristics of the training set. I suspect this is related to underfitting in DL. I wonder whether changing the model configuration, e.g. increasing t... | closed | completed | false | 2 | [] | [] | 2023-09-15T08:06:59Z | 2023-09-16T08:26:12Z | 2023-09-16T08:26:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sunhaozhepy | 73,462,159 | MDQ6VXNlcjczNDYyMTU5 | User | false |
huggingface/diffusers | 1,898,189,433 | I_kwDOHa8MBc5xJBJ5 | 5,054 | https://github.com/huggingface/diffusers/issues/5054 | https://api.github.com/repos/huggingface/diffusers/issues/5054 | AttributeError: 'StableDiffusionXLPipeline' object has no attribute 'final_offload_hook' | ### Describe the bug
After upgrading my code to `0.21.1` it triggers the error:
> AttributeError: 'StableDiffusionXLPipeline' object has no attribute 'final_offload_hook'
### Reproduction
```py
import os
import torch
from diffusers import StableDiffusionXLPipeline
MODEL_PATH = os.getenv('MODEL_PAT... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-09-15T10:58:27Z | 2023-10-05T05:58:39Z | 2023-09-17T09:37:07Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexisrolland | 13,064,696 | MDQ6VXNlcjEzMDY0Njk2 | User | false |
huggingface/diffusers | 1,898,248,523 | I_kwDOHa8MBc5xJPlL | 5,055 | https://github.com/huggingface/diffusers/issues/5055 | https://api.github.com/repos/huggingface/diffusers/issues/5055 | How to download config.json if it is not in the root directory. | Is there any way to download vae for a model where config.json is not in the root directory?
```python
vae = AutoencoderKL.from_pretrained("redstonehero/kl-f8-anime2")
```
For example, as shown above, there is no problem if config.json exists in the root directory, but if it does not exist, an error will occur... | closed | completed | false | 2 | [] | [] | 2023-09-15T11:37:47Z | 2023-09-16T00:15:58Z | 2023-09-16T00:15:57Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | suzukimain | 131,413,573 | U_kgDOB9U2RQ | User | false |
huggingface/diffusers | 1,898,582,890 | I_kwDOHa8MBc5xKhNq | 5,059 | https://github.com/huggingface/diffusers/issues/5059 | https://api.github.com/repos/huggingface/diffusers/issues/5059 | Finetune Stable Diffusion XL on multiple concepts | Hey,
I was looking into the LoRA training script for Stable diffusion XL. Would it be possible in the future to add support of being able to provide prompts for the training images from the name of the image or separate text file (similar to training in stable diffusion like here:
parser.add_argument(
... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-09-15T14:51:13Z | 2023-10-30T15:06:23Z | 2023-10-30T15:06:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Maximilianwte | 22,194,558 | MDQ6VXNlcjIyMTk0NTU4 | User | false |
huggingface/diffusers | 1,898,852,807 | I_kwDOHa8MBc5xLjHH | 5,060 | https://github.com/huggingface/diffusers/issues/5060 | https://api.github.com/repos/huggingface/diffusers/issues/5060 | Diffusers 0.21 breaks model offloading | ### Describe the bug
upgrading to `diffusers==0.21.1` results in unusable `StableDiffusionPipeline` if either offloading options are used: `enable_model_cpu_offload` or `enable_sequential_cpu_offload`
digging deeper, reason seems to be that `pipeline.text_encoder` component does not get pulled back to execution dev... | closed | completed | false | 12 | [
"bug"
] | [] | 2023-09-15T17:51:30Z | 2023-09-27T03:19:47Z | 2023-09-27T03:19:47Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 1,899,005,671 | I_kwDOHa8MBc5xMIbn | 5,061 | https://github.com/huggingface/diffusers/issues/5061 | https://api.github.com/repos/huggingface/diffusers/issues/5061 | convert_original_stable_diffusion_to_diffusers produces argument error on config_files | ### Describe the bug
Trying to use the latest version of the conversion script does not work since the call to download_from_original_stable_diffusion_ckpt tries to use args.config_files which cannot be passed as an argument.
### Reproduction
Use convert_original_stable_diffusion_to_diffusers.py with any checkpoint_... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-09-15T19:56:12Z | 2023-10-30T15:06:21Z | 2023-10-30T15:06:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | SirRichie | 12,241,553 | MDQ6VXNlcjEyMjQxNTUz | User | false |
huggingface/diffusers | 1,899,122,814 | I_kwDOHa8MBc5xMlB- | 5,062 | https://github.com/huggingface/diffusers/issues/5062 | https://api.github.com/repos/huggingface/diffusers/issues/5062 | `prcoessor` typo in attention | Hi. There is a typo here:
https://github.com/huggingface/diffusers/blob/v0.21.1/src/diffusers/models/attention_processor.py#L385
Instead of `prcoessor` it should be `processor`.
Because of this typo my LoRA training doesn't work. | closed | completed | false | 0 | [] | [] | 2023-09-15T21:56:46Z | 2023-09-16T12:43:20Z | 2023-09-16T12:43:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TropComplique | 19,353,433 | MDQ6VXNlcjE5MzUzNDMz | User | false |
huggingface/diffusers | 1,899,352,244 | I_kwDOHa8MBc5xNdC0 | 5,064 | https://github.com/huggingface/diffusers/issues/5064 | https://api.github.com/repos/huggingface/diffusers/issues/5064 | Out of space while using refiner for inference. | ### Describe the bug
After training, dreambooth_lora_sdxl, when I try to infer using refiner, the device runs out of space.
Initially I was using 100GB in my cloud provider, so I though maybe I need to increase the sapce.
However, after increasing the space to 256 GB and running the inference using refiner, I get t... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-09-16T08:23:01Z | 2023-09-16T15:51:17Z | 2023-09-16T15:51:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | PrakharSaxena24 | 50,725,987 | MDQ6VXNlcjUwNzI1OTg3 | User | false |
huggingface/diffusers | 1,899,373,007 | I_kwDOHa8MBc5xNiHP | 5,067 | https://github.com/huggingface/diffusers/issues/5067 | https://api.github.com/repos/huggingface/diffusers/issues/5067 | Missing requirements in examples\controlnet | ### Describe the bug
Using `examples/controlnet` only installs the requirements needed for CPU training which is terribly slow. It should also install requirements for CUDA training whether or not this will be used by default.
Also some requirements for CPU training are missing on Windows at all.
**Missing requi... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-09-16T09:35:09Z | 2023-10-30T15:06:19Z | 2023-10-30T15:06:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | geroldmeisinger | 112,266,044 | U_kgDOBrELPA | User | false |
huggingface/diffusers | 1,899,377,858 | I_kwDOHa8MBc5xNjTC | 5,069 | https://github.com/huggingface/diffusers/issues/5069 | https://api.github.com/repos/huggingface/diffusers/issues/5069 | examples\controlnet option to save inference validation images to disk | **Is your feature request related to a problem? Please describe.**
I recently tried to train a control net with the `train_controlnet.py` example and documented the whole process [in this article](https://civitai.com/articles/2078#heading-3270). Following the [controlnet tutorial ](https://huggingface.co/docs/diffuser... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-09-16T09:53:52Z | 2023-11-08T15:07:54Z | 2023-11-08T15:07:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | geroldmeisinger | 112,266,044 | U_kgDOBrELPA | User | false |
huggingface/diffusers | 1,372,453,764 | I_kwDOHa8MBc5RzfuE | 507 | https://github.com/huggingface/diffusers/issues/507 | https://api.github.com/repos/huggingface/diffusers/issues/507 | Textual Inversion Broken: it updates entire `embeddings` weights, loss diverges | ### Describe the bug
The weights of the entire `text_encoder` evolve over the course of training, thus breaking the `text_encoder`. I'm not sure why yet, but this in turn breaks Inversion.
To demonstrate it,
1.) save a random token id and it's embedding, outside the main loop:
```python
token_embed_w_... | closed | completed | false | 12 | [] | [
"patil-suraj"
] | 2022-09-14T06:53:46Z | 2022-12-12T16:44:40Z | 2022-12-12T16:44:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | freckletonj | 8,399,149 | MDQ6VXNlcjgzOTkxNDk= | User | false |
huggingface/diffusers | 1,899,379,100 | I_kwDOHa8MBc5xNjmc | 5,070 | https://github.com/huggingface/diffusers/issues/5070 | https://api.github.com/repos/huggingface/diffusers/issues/5070 | examples\controlnet add only_mid_control=true and sd_locked=false | **Is your feature request related to a problem? Please describe.**
The [original controlnet training tutorial](https://github.com/lllyasviel/ControlNet/blob/main/docs/train.md#other-options) mentions the options
* `only_mid_control=True`: _This can be helpful when your computation power is limited and want to speed u... | open | null | false | 9 | [
"Good second issue",
"contributions-welcome"
] | [] | 2023-09-16T09:58:50Z | 2024-05-08T00:57:08Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | geroldmeisinger | 112,266,044 | U_kgDOBrELPA | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.