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 | 1,828,439,989 | I_kwDOHa8MBc5s-8e1 | 4,381 | https://github.com/huggingface/diffusers/issues/4381 | https://api.github.com/repos/huggingface/diffusers/issues/4381 | Optimization in Diffusers in comparison with Comfy-UI and Automatic-1111 | I'm using diffusers from since months , and found it very good in terms of customization for own deployment services. But few days ago when SDXL 1.0 came out tried it on google colab and the size of model is crashing the session on colab after one or two image generation, while in same specs on google colab Comfy-UI ca... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-07-31T05:34:14Z | 2023-10-18T15:14:52Z | 2023-10-18T15:14:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tronizx | 87,223,249 | MDQ6VXNlcjg3MjIzMjQ5 | User | false |
huggingface/diffusers | 1,828,595,915 | I_kwDOHa8MBc5s_ijL | 4,382 | https://github.com/huggingface/diffusers/issues/4382 | https://api.github.com/repos/huggingface/diffusers/issues/4382 | HOW TO Overcoming the Influence of Seed and Enhancing the Role of Text Prompts | I fine-tuned a text2img model using Lora, based on the v1.5 version of stable diffusion. The results generated are very good.
But they can’t be controlled. It seems that the generated results are more based on the seed. Changing the seed changes the image, And if I don’t change the seed and only change the text prompt... | closed | completed | false | 1 | [] | [] | 2023-07-31T07:41:03Z | 2023-08-02T09:23:50Z | 2023-08-02T09:23:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | XiaoyuZhuang | 45,529,460 | MDQ6VXNlcjQ1NTI5NDYw | User | false |
huggingface/diffusers | 1,828,621,813 | I_kwDOHa8MBc5s_o31 | 4,383 | https://github.com/huggingface/diffusers/issues/4383 | https://api.github.com/repos/huggingface/diffusers/issues/4383 | how can i put my controlnet model which is traind by train_controlnet.py to the automatic1111 ? | I have traind a controlnet model with the diffuser/examples/controlnet/train_controlnet.py，which i got a diffusion_pytorch_model.bin and a config.json。 how can i use them in stablediffusion webui?
this is my files:
 is not suitable for sdxl pipeline | closed | completed | false | 8 | [] | [] | 2023-07-31T09:23:04Z | 2023-08-04T04:37:53Z | 2023-08-04T04:37:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | WaltBao | 106,211,276 | U_kgDOBlSnzA | User | false |
huggingface/diffusers | 1,828,868,925 | I_kwDOHa8MBc5tAlM9 | 4,387 | https://github.com/huggingface/diffusers/issues/4387 | https://api.github.com/repos/huggingface/diffusers/issues/4387 | Confusion about the interaction between `gradient_accumulation_steps` and `lr_scheduler` | I'm running code from `diffusers/examples/controlnet/train_controlnet.py`, and I'm confused about some interactions between `gradient_accumulation_steps` and `lr_scheduler`
Here's the code
```
# Scheduler and math around the number of training steps.
overrode_max_train_steps = False
num_update_steps_per_epoc... | closed | completed | false | 3 | [] | [] | 2023-07-31T10:17:41Z | 2023-08-03T03:25:36Z | 2023-08-03T03:25:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | KyleYueye | 34,768,573 | MDQ6VXNlcjM0NzY4NTcz | User | false |
huggingface/diffusers | 1,829,137,934 | I_kwDOHa8MBc5tBm4O | 4,389 | https://github.com/huggingface/diffusers/issues/4389 | https://api.github.com/repos/huggingface/diffusers/issues/4389 | Error when loading ControlNetModel | ### Describe the bug
When creating a `ControlNetModel` object an error is raised. If I run `pip install accelerate` as suggested in the logs it work fine.
My question is: why `accelerate` is not in the requirements of `diffusers` if it is necessary to load some models?
### Reproduction
```python
from diffuse... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-07-31T12:57:41Z | 2023-08-02T09:16:07Z | 2023-08-02T09:16:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | domef | 38,183,728 | MDQ6VXNlcjM4MTgzNzI4 | User | false |
huggingface/diffusers | 1,829,154,459 | I_kwDOHa8MBc5tBq6b | 4,390 | https://github.com/huggingface/diffusers/issues/4390 | https://api.github.com/repos/huggingface/diffusers/issues/4390 | Missing requirements | ### Describe the bug
When I try to use `StableDiffusionControlNetPipeline` or `StableDiffusionControlNetImg2ImgPipeline` I get an error. If I install `transformers` it works correctly.
My question is: why `transformers` is not in the requirements of `diffusers` if it is necessary to load some models?
### Reprodu... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-07-31T13:07:02Z | 2023-08-13T18:46:45Z | 2023-08-02T09:15:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | domef | 38,183,728 | MDQ6VXNlcjM4MTgzNzI4 | User | false |
huggingface/diffusers | 1,829,239,865 | I_kwDOHa8MBc5tB_w5 | 4,391 | https://github.com/huggingface/diffusers/issues/4391 | https://api.github.com/repos/huggingface/diffusers/issues/4391 | Wrong default value (or documentation) in StableDiffusionXLPipeline | ### Describe the bug
According to the documentation(1), the default value for guidance_scale is 7.5. After rendering a couple of thousand images, I set this value explicitly, only to find out that in fact, it defaults to 5.0. This is unlikely to be a documentation error, as 7.5 is the most sensible value, also used in... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-07-31T13:51:06Z | 2023-08-01T23:40:16Z | 2023-08-01T23:40:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rolux | 152,646 | MDQ6VXNlcjE1MjY0Ng== | User | false |
huggingface/diffusers | 1,829,261,118 | I_kwDOHa8MBc5tCE8- | 4,392 | https://github.com/huggingface/diffusers/issues/4392 | https://api.github.com/repos/huggingface/diffusers/issues/4392 | SDXL 1.0 Inpainting - Lower result quality with certain masks | Hi there! I've been conducting tests using the sd-xl-1-0-base model with the pipeline `StableDiffusionXLInpaintPipeline`.
However, I've noticed that the model generates results where the "white zone" doesn't seem to have any relation to the rest of the image. It appears as if another image has been pasted over the ... | closed | completed | false | 18 | [
"stale"
] | [] | 2023-07-31T14:00:56Z | 2023-11-08T15:10:07Z | 2023-11-08T15:10:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Madot10 | 30,395,090 | MDQ6VXNlcjMwMzk1MDkw | User | false |
huggingface/diffusers | 1,829,483,642 | I_kwDOHa8MBc5tC7R6 | 4,393 | https://github.com/huggingface/diffusers/issues/4393 | https://api.github.com/repos/huggingface/diffusers/issues/4393 | AttributeError: module 'diffusers' has no attribute 'StableDiffusionXLPipeline' | ### Describe the bug
Diffusers version 0.15.1
Using SDXL in any capacity is broken
### Reproduction
from diffusers import DiffusionPipeline
import torch
pipe = DiffusionPipeline.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1.0",
torch_dtype=torch.float16,
variant="fp16",
use_saf... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-07-31T15:32:03Z | 2024-02-11T04:44:10Z | 2023-07-31T15:37:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | IMJONEZZ | 46,764,336 | MDQ6VXNlcjQ2NzY0MzM2 | User | false |
huggingface/diffusers | 1,829,983,901 | I_kwDOHa8MBc5tE1ad | 4,398 | https://github.com/huggingface/diffusers/issues/4398 | https://api.github.com/repos/huggingface/diffusers/issues/4398 | All black images generated from new SDXL LORA-dreambooth pipeline | ### Describe the bug
The validation images are all black, and they are not nude just all black images.
Here is my launch script:
accelerate launch --mixed_precision="fp16" train_dreambooth_lora_sdxl.py \
--pretrained_model_name_or_path="stabilityai/stable-diffusion-xl-base-1.0" \
--instance_data_dir=xx \
--... | closed | completed | false | 11 | [
"bug"
] | [] | 2023-07-31T20:31:37Z | 2023-08-01T05:01:54Z | 2023-08-01T05:01:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | linnanwang | 4,675,611 | MDQ6VXNlcjQ2NzU2MTE= | User | false |
huggingface/diffusers | 1,830,002,356 | I_kwDOHa8MBc5tE560 | 4,399 | https://github.com/huggingface/diffusers/issues/4399 | https://api.github.com/repos/huggingface/diffusers/issues/4399 | Pytorch 2 compile and enable cpu model offload do not work together | ### Describe the bug
The system will crash when creating a second image.
### Reproduction
from diffusers import StableDiffusionXLPipeline
import torch
pipe = StableDiffusionXLPipeline.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16, variant="fp16", use_safetensors=True... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-07-31T20:47:00Z | 2023-11-02T08:40:14Z | 2023-10-25T15:10:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | frankjoshua | 230,666 | MDQ6VXNlcjIzMDY2Ng== | User | false |
huggingface/diffusers | 1,830,175,105 | I_kwDOHa8MBc5tFkGB | 4,400 | https://github.com/huggingface/diffusers/issues/4400 | https://api.github.com/repos/huggingface/diffusers/issues/4400 | Question about text_encoder support in train_dreambooth_inpaint_lora.py | Hi, it seems that although `args.train_text_encoder` is included in [train_dreambooth_inpaint_lora.py](https://github.com/huggingface/diffusers/blob/main/examples/research_projects/dreambooth_inpaint/train_dreambooth_inpaint_lora.py), the script does not add LoRA layers to the text encoder part. Is it safe to follow [t... | closed | completed | false | 1 | [] | [] | 2023-07-31T23:04:29Z | 2023-08-02T09:04:43Z | 2023-08-02T09:04:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hlyu09 | 43,687,446 | MDQ6VXNlcjQzNjg3NDQ2 | User | false |
huggingface/diffusers | 1,830,380,595 | I_kwDOHa8MBc5tGWQz | 4,402 | https://github.com/huggingface/diffusers/issues/4402 | https://api.github.com/repos/huggingface/diffusers/issues/4402 | StableDiffusionXLInpaintPipeline.from_single_file() cant correct use | ### Describe the bug
StableDiffusionXLInpaintPipeline.from_single_file(.....)
The pipe will become of type StableDiffusionXLPipeline and cannot be used
>>>Init_ Image=load_ Image (img_url). convert ("RGB")
>>>Mask_ Image=load_ Image (mask_url). convert ("RGB")
>>>Prompt="A Majestic tiger sitting on... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-08-01T03:09:23Z | 2023-08-09T08:50:31Z | 2023-08-09T08:50:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | skyheros001 | 32,533,832 | MDQ6VXNlcjMyNTMzODMy | User | false |
huggingface/diffusers | 1,830,409,621 | I_kwDOHa8MBc5tGdWV | 4,403 | https://github.com/huggingface/diffusers/issues/4403 | https://api.github.com/repos/huggingface/diffusers/issues/4403 | "LayerNormKernelImpl" not implemented for 'Half' | ### Describe the bug
Kaggle notebook, GPU P100
### Reproduction
```
from diffusers.utils import load_image
image = load_image(
"https://hf.co/datasets/huggingface/documentation-images/resolve/main/diffusers/input_image_vermeer.png"
)
image
import cv2
from PIL import Image
import numpy as np
image ... | closed | completed | false | 9 | [
"bug"
] | [] | 2023-08-01T03:48:20Z | 2023-08-01T06:46:01Z | 2023-08-01T06:16:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,830,435,656 | I_kwDOHa8MBc5tGjtI | 4,405 | https://github.com/huggingface/diffusers/issues/4405 | https://api.github.com/repos/huggingface/diffusers/issues/4405 | [Examples] Create textual inversion training script for SDXL | We have training for textual inversion in place: https://github.com/huggingface/diffusers/tree/main/examples/textual_inversion.
Might be worth adding one compatible with SDXL. | closed | completed | false | 19 | [
"stale",
"contributions-welcome"
] | [] | 2023-08-01T04:22:10Z | 2024-01-19T15:06:39Z | 2024-01-19T15:06:39Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,830,465,745 | I_kwDOHa8MBc5tGrDR | 4,406 | https://github.com/huggingface/diffusers/issues/4406 | https://api.github.com/repos/huggingface/diffusers/issues/4406 | SafetensorError: Error while deserializing header: HeaderTooLarge | ### Describe the bug
running:
```
!wget https://huggingface.co/XpucT/Deliberate/blob/main/Deliberate-inpainting.safetensors
#checkpoit
!python ./diffusers/scripts/convert_original_stable_diffusion_to_diffusers.py --checkpoint_path ./Deliberate-inpainting.safetensors --dump_path ./test_model --from_safetensors
... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-08-01T04:58:19Z | 2023-08-01T05:57:42Z | 2023-08-01T05:57:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,830,486,512 | I_kwDOHa8MBc5tGwHw | 4,407 | https://github.com/huggingface/diffusers/issues/4407 | https://api.github.com/repos/huggingface/diffusers/issues/4407 | how to store hub_download on local directory? | ### Describe the bug
running:
from huggingface_hub import hf_hub_url, hf_hub_download
```
# Generate/show the URL
hf_hub_url(
repo_id="XpucT/Deliberate",
filename="Deliberate-inpainting.safetensors",
)
# Download the file
hf_hub_download(
repo_id="XpucT/Deliberate",
filename="Deliberate-inpain... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-01T05:21:39Z | 2023-08-01T05:55:46Z | 2023-08-01T05:55:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,830,524,408 | I_kwDOHa8MBc5tG5X4 | 4,409 | https://github.com/huggingface/diffusers/issues/4409 | https://api.github.com/repos/huggingface/diffusers/issues/4409 | 'UNet2DConditionModel' object has no attribute 'alphas_cumprod' | ### Describe the bug
Note: https://colab.research.google.com/drive/1LczvyDueMQ4GB82sibqCrLlUnPafIbmp?usp=sharing
while running:
```
!pip install -q langchain datasets ftfy accelerate scipy safetensors
# Installing transformers from source for now since we need the latest version for Depth2Img:
!pip install -q... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-08-01T06:01:55Z | 2023-09-12T15:03:59Z | 2023-09-12T15:03:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,367,536,115 | I_kwDOHa8MBc5RgvHz | 441 | https://github.com/huggingface/diffusers/issues/441 | https://api.github.com/repos/huggingface/diffusers/issues/441 | How is running locally requiring your tokenized access to ckpt file ? Is that really local ? How is that local ? | ### Describe the bug
this crap
https://discuss.huggingface.co/t/how-to-login-to-huggingface-hub-with-access-token/22498/5
So... just let us use our own path to our folders and ckpt file ok ? Like a local thing you know ?
### Reproduction
try it on miniconda, good luck !!!
### Logs
```shell
none
```
### S... | closed | completed | false | 12 | [
"stale"
] | [] | 2022-09-09T09:19:15Z | 2022-10-31T15:03:22Z | 2022-10-31T15:03:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | 1blackbar | 101,844,463 | U_kgDOBhIF7w | User | false |
huggingface/diffusers | 1,830,566,049 | I_kwDOHa8MBc5tHDih | 4,410 | https://github.com/huggingface/diffusers/issues/4410 | https://api.github.com/repos/huggingface/diffusers/issues/4410 | Cannot load ./test_model/unet because conv_in.weight expected shape tensor(..., device='meta', size=(320, 4, 3, 3)) | ### Describe the bug
Notebook: https://colab.research.google.com/drive/1LczvyDueMQ4GB82sibqCrLlUnPafIbmp?usp=sharing
I was working on:
```
#@title Libraries
!pip install -q langchain datasets ftfy accelerate scipy safetensors
# Installing transformers from source for now since we need the latest version fo... | closed | completed | false | 10 | [
"bug"
] | [] | 2023-08-01T06:37:30Z | 2023-08-03T15:11:54Z | 2023-08-03T14:47:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,830,752,715 | I_kwDOHa8MBc5tHxHL | 4,414 | https://github.com/huggingface/diffusers/issues/4414 | https://api.github.com/repos/huggingface/diffusers/issues/4414 | Sample with Consistency Model | ### Describe the bug
When I use diffusers to sample images with offered code by Consistency Model, I can't get an image that corresponds to the label at all:
label 207 (Golden Retriever)
### Reproduction
I just ... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-08-01T08:37:50Z | 2023-09-12T15:03:57Z | 2023-09-12T15:03:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | RossoneriZhao | 102,647,326 | U_kgDOBh5GHg | User | false |
huggingface/diffusers | 1,830,899,636 | I_kwDOHa8MBc5tIU-0 | 4,416 | https://github.com/huggingface/diffusers/issues/4416 | https://api.github.com/repos/huggingface/diffusers/issues/4416 | Improve T2I Adapter Tests | The current set of T2I adapter tests run a single test for multiple adapters.
https://github.com/huggingface/diffusers/blob/6c49d542a352b556e412d8763592050d3a0dec77/tests/pipelines/stable_diffusion/test_stable_diffusion_adapter.py#L183-L293
These should be broken up into individual tests (since precision requireme... | closed | completed | false | 3 | [
"enhancement"
] | [
"DN6"
] | 2023-08-01T09:57:04Z | 2024-11-22T07:52:10Z | 2024-11-22T07:52:10Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DN6 | 7,529,846 | MDQ6VXNlcjc1Mjk4NDY= | User | false |
huggingface/diffusers | 1,830,985,827 | I_kwDOHa8MBc5tIqBj | 4,418 | https://github.com/huggingface/diffusers/issues/4418 | https://api.github.com/repos/huggingface/diffusers/issues/4418 | denoising_start does not work normally in pipeline_stable_diffusion_xl_inpaint.py | ### Describe the bug
I am using StableDiffusionXLInpaintPipeline with stable-diffusion-xl-base-0.9 weight.
And all I use is
```python
prompt = "2 dogs on the beach"
image = base(
prompt=prompt,
num_inference_steps=n_steps,
image = image,
denoising_start = denois... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-01T10:44:11Z | 2023-08-09T08:45:53Z | 2023-08-09T08:45:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CHR-ray | 75,762,846 | MDQ6VXNlcjc1NzYyODQ2 | User | false |
huggingface/diffusers | 1,831,101,707 | I_kwDOHa8MBc5tJGUL | 4,419 | https://github.com/huggingface/diffusers/issues/4419 | https://api.github.com/repos/huggingface/diffusers/issues/4419 | SDXL trained to use offset noise but not used in inference | ### Describe the bug
It sounds like they trained with offset noise but from I can tell the inference implementation does not use offset noise?
https://arxiv.org/pdf/2307.01952.pdf
_I may have missed it of cour... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-08-01T11:53:51Z | 2023-08-03T14:49:26Z | 2023-08-03T05:40:33Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | samedii | 1,872,050 | MDQ6VXNlcjE4NzIwNTA= | User | false |
huggingface/diffusers | 1,831,228,175 | I_kwDOHa8MBc5tJlMP | 4,421 | https://github.com/huggingface/diffusers/issues/4421 | https://api.github.com/repos/huggingface/diffusers/issues/4421 | train controlnet with img2img methods | I want to train the controlnet with specific input of images and contol condition images instead of the prompts, but I found that the script only supports the prompts input to train the controlnet.
Is any script which I didn't find can deal with this problem? | closed | completed | false | 9 | [] | [] | 2023-08-01T12:57:44Z | 2024-08-02T08:50:52Z | 2023-08-02T08:35:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | VV-A-VV | 59,974,890 | MDQ6VXNlcjU5OTc0ODkw | User | false |
huggingface/diffusers | 1,831,609,867 | I_kwDOHa8MBc5tLCYL | 4,424 | https://github.com/huggingface/diffusers/issues/4424 | https://api.github.com/repos/huggingface/diffusers/issues/4424 | [SD-XL] Instruct Pix2Pix fine-tuning script unconditionally upcasts VAE to float32 when built-in VAE is used | ```py
# Move vae, unet and text_encoder to device and cast to weight_dtype
# The VAE is in float32 to avoid NaN losses.
if args.pretrained_vae_model_name_or_path is not None:
vae.to(accelerator.device, dtype=weight_dtype)
else:
vae.to(accelerator.device, dtype=torch.float32)
```
... | closed | completed | false | 9 | [
"stale"
] | [
"bghira"
] | 2023-08-01T16:10:07Z | 2023-09-05T07:04:08Z | 2023-09-05T07:04:08Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 1,831,735,577 | I_kwDOHa8MBc5tLhEZ | 4,425 | https://github.com/huggingface/diffusers/issues/4425 | https://api.github.com/repos/huggingface/diffusers/issues/4425 | 3D diffuser consists of Conv2D layers | **What API design would you like to have changed or added to the library? Why?**
I was looking for unconditional 3D diffusion models and possibly creating it myself. However, the current 3d_conditional model that uses the 3Dblocks I was intending to use seem to not be 3D, but at the base they are Lora compatible conv2... | closed | completed | false | 1 | [] | [] | 2023-08-01T17:31:38Z | 2023-08-02T08:32:41Z | 2023-08-02T08:32:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Meghpal | 40,922,889 | MDQ6VXNlcjQwOTIyODg5 | User | false |
huggingface/diffusers | 1,831,886,513 | I_kwDOHa8MBc5tMF6x | 4,427 | https://github.com/huggingface/diffusers/issues/4427 | https://api.github.com/repos/huggingface/diffusers/issues/4427 | T2I Multiadapter example not working | ### Describe the bug
I tried running the example provided in documentation [here](https://huggingface.co/docs/diffusers/api/pipelines/stable_diffusion/adapter#combining-multiple-adapters) for multi adapters, I'm getting this error:
`AttributeError: 'MultiAdapter' object has no attribute 'total_downscale_factor'`
... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-08-01T19:23:34Z | 2023-09-02T17:28:25Z | 2023-08-24T19:01:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | amrakm | 8,410,055 | MDQ6VXNlcjg0MTAwNTU= | User | false |
huggingface/diffusers | 1,832,241,201 | I_kwDOHa8MBc5tNcgx | 4,429 | https://github.com/huggingface/diffusers/issues/4429 | https://api.github.com/repos/huggingface/diffusers/issues/4429 | Init_image behaviour is inconsistent across pipelines | ### Describe the bug
The expectations of image-to-image pipelines regarding the format of the input image are unclear and inconsistent:
- `StableDiffusionImg2ImgPipeline` can take a PIL Image, or a numpy float array in the range [0, 1], or a torch float tensor in the range [0, 1].
- `StableDiffusionInpaintPipeline... | closed | completed | false | 9 | [
"stale"
] | [
"yiyixuxu"
] | 2023-08-02T01:06:03Z | 2023-12-02T15:06:54Z | 2023-12-02T15:06:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | antoche | 1,627,384 | MDQ6VXNlcjE2MjczODQ= | User | false |
huggingface/diffusers | 1,367,656,208 | I_kwDOHa8MBc5RhMcQ | 443 | https://github.com/huggingface/diffusers/issues/443 | https://api.github.com/repos/huggingface/diffusers/issues/443 | [CoreML] Make diffusers work natively with Mac CoreML | Required changes to make diffusers work out of the book with CoreML: https://gist.github.com/madebyollin/86b9596ffa4ab0fa7674a16ca2aeab3d?permalink_comment_id=4295017 | closed | completed | false | 4 | [
"stale"
] | [] | 2022-09-09T10:54:40Z | 2022-10-31T15:03:21Z | 2022-10-31T15:03:21Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,832,395,479 | I_kwDOHa8MBc5tOCLX | 4,430 | https://github.com/huggingface/diffusers/issues/4430 | https://api.github.com/repos/huggingface/diffusers/issues/4430 | I want to use multi-node training, but the process is stuck | 
The thing is, I use multiple machines, 2x6 A100, to train controlnet, but I don't quite understand why the... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-08-02T04:10:39Z | 2023-09-12T15:03:54Z | 2023-09-12T15:03:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pixeli99 | 46,072,190 | MDQ6VXNlcjQ2MDcyMTkw | User | false |
huggingface/diffusers | 1,832,724,294 | I_kwDOHa8MBc5tPSdG | 4,432 | https://github.com/huggingface/diffusers/issues/4432 | https://api.github.com/repos/huggingface/diffusers/issues/4432 | Question on memory doubling | I am running SD-XL 1.0 inference using the following code:
```python
import torch
import matplotlib.pyplot as plt
from PIL import Image
from diffusers import StableDiffusionXLPipeline
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
pipe = StableDiffusionXLPipeline.from_pretrained("s... | closed | completed | false | 1 | [] | [] | 2023-08-02T08:44:36Z | 2023-08-02T09:01:54Z | 2023-08-02T09:01:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | VigneshBaskar | 11,005,940 | MDQ6VXNlcjExMDA1OTQw | User | false |
huggingface/diffusers | 1,832,774,857 | I_kwDOHa8MBc5tPezJ | 4,433 | https://github.com/huggingface/diffusers/issues/4433 | https://api.github.com/repos/huggingface/diffusers/issues/4433 | segment anything model for text-guided image editing- GroundingDINO with SD | @patrickvonplaten @sayakpaul Any updates on this pipeline. I am trying to use GroundingDINO with Stable diffusion . While trying on a custom dataset i got the following error(Since there is no documentation of HF it would be a nice addition):
Issue: https://github.com/IDEA-Research/GroundingDINO/issues/184
Colab: h... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-08-02T09:13:49Z | 2023-09-12T15:03:52Z | 2023-09-12T15:03:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,833,048,876 | I_kwDOHa8MBc5tQhss | 4,435 | https://github.com/huggingface/diffusers/issues/4435 | https://api.github.com/repos/huggingface/diffusers/issues/4435 | SDXL model offload doesn't unload Unet if latent output is requested | If VEA decode is skipped (through `output_type = "latent"`), the `self.final_offload_hook.offload()` won't be called, and even if it was, it would trigger offload of the VAE, not the Unet. This can be worked around by calling `.to("cpu")` on the pipeline afterwards.
https://github.com/huggingface/diffusers/blob/816c... | closed | completed | false | 5 | [
"wip"
] | [] | 2023-08-02T12:02:54Z | 2025-01-27T01:56:16Z | 2025-01-27T01:56:15Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Kubuxu | 2,259,282 | MDQ6VXNlcjIyNTkyODI= | User | false |
huggingface/diffusers | 1,833,076,326 | I_kwDOHa8MBc5tQoZm | 4,436 | https://github.com/huggingface/diffusers/issues/4436 | https://api.github.com/repos/huggingface/diffusers/issues/4436 | Converting CIVITAI safetensors with VAE to diffusers | ### Describe the bug
I can convert CIVITAI safetensors to diffusers. But most CIVITAI have VAE with them.
How do I convert CIVITAI safetensors w/ VAE to diffusers?
### Reproduction
Just using the script from diffusers
### Logs
_No response_
### System Info
Latest diffusers on Colab
### Who can help?
_No respo... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-08-02T12:19:52Z | 2023-09-12T15:03:50Z | 2023-09-12T15:03:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sarmientoj24 | 8,830,319 | MDQ6VXNlcjg4MzAzMTk= | User | false |
huggingface/diffusers | 1,833,314,815 | I_kwDOHa8MBc5tRin_ | 4,438 | https://github.com/huggingface/diffusers/issues/4438 | https://api.github.com/repos/huggingface/diffusers/issues/4438 | StableDiffusionXLPipeline Error | ### Describe the bug
When I try to set the pipe with `StableDiffusionXLImg2ImgPipeline`, I think it returns the pipe with 'StableDiffusionXLPipeline'
I downloaded the model in civitiai
https://civitai.com/models/112902/dreamshaper-xl10-alpha2
https://civitai.com/models/118223/sdxloil-paintingoil-brush-strok... | closed | completed | false | 8 | [
"bug"
] | [] | 2023-08-02T14:35:09Z | 2023-08-17T11:35:05Z | 2023-08-17T11:35:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Gynjn | 106,458,325 | U_kgDOBlhs1Q | User | false |
huggingface/diffusers | 1,367,662,974 | I_kwDOHa8MBc5RhOF- | 444 | https://github.com/huggingface/diffusers/issues/444 | https://api.github.com/repos/huggingface/diffusers/issues/444 | IndexError on num_inference_steps values: 3, 9, 27, ...? | ### Describe the bug
Some `num_inference_steps` values seem to be "problematic" and throw an IndexError.
Also the iteration status output does not end on that value in any relation. If it's on purpose, it's still confusing:
num_inference_steps: **5**
**6**it [00:49, 8.27s/it]
num_inference_steps: **6**
*... | closed | completed | false | 20 | [
"bug"
] | [] | 2022-09-09T11:00:54Z | 2022-09-17T12:29:35Z | 2022-09-17T12:29:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sausix | 21,084,645 | MDQ6VXNlcjIxMDg0NjQ1 | User | false |
huggingface/diffusers | 1,833,756,590 | I_kwDOHa8MBc5tTOeu | 4,440 | https://github.com/huggingface/diffusers/issues/4440 | https://api.github.com/repos/huggingface/diffusers/issues/4440 | Disentangle ONNX and non-ONNX pipeline | The `OnnxStableDiffusionUpscalePipeline` pipeline should not inherit from `StableDiffusionUpscalePipeline`. This is not in-sync with our design philosophy and has led to multiple errors already as can be seen here: https://github.com/huggingface/diffusers/pull/2158#issuecomment-1650580180 .
Anybody keen on removing ... | closed | completed | false | 5 | [
"stale"
] | [
"DN6"
] | 2023-08-02T19:09:03Z | 2023-09-04T06:41:36Z | 2023-09-04T06:41:36Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,834,094,828 | I_kwDOHa8MBc5tUhDs | 4,443 | https://github.com/huggingface/diffusers/issues/4443 | https://api.github.com/repos/huggingface/diffusers/issues/4443 | loading multi Lora models with different weights | Thank you to the developers for their great work. Currently, Diffusers has very complete functionality, especially in supporting single SD models.
However, in the AI painting industry, loading multi Lora models with different weights has been proven effective by many users. So, when can diffusers add multi Lora model... | closed | completed | false | 2 | [] | [] | 2023-08-03T00:44:57Z | 2023-08-04T02:59:26Z | 2023-08-04T02:59:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhaoyun0071 | 35,762,050 | MDQ6VXNlcjM1NzYyMDUw | User | false |
huggingface/diffusers | 1,834,145,277 | I_kwDOHa8MBc5tUtX9 | 4,445 | https://github.com/huggingface/diffusers/issues/4445 | https://api.github.com/repos/huggingface/diffusers/issues/4445 | How to finetune lora model ? | **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 [...]
If I have a model from civitai , how to finetune it in sd1.5 and sdxl?
**Describe the solution you'd like**
A clear and concise description of what you w... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-08-03T01:55:15Z | 2023-09-12T15:03:49Z | 2023-09-12T15:03:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kelisiya | 23,169,766 | MDQ6VXNlcjIzMTY5NzY2 | User | false |
huggingface/diffusers | 1,834,211,507 | I_kwDOHa8MBc5tU9iz | 4,446 | https://github.com/huggingface/diffusers/issues/4446 | https://api.github.com/repos/huggingface/diffusers/issues/4446 | Unable to use custom text_encoder in `StableDiffusionPipeline.from_single_file` | ### Describe the bug
Hi, I was trying to use load a custom `text_encoder` with a CIVITAI downloaded safetensors in `StableDiffusionPipeline`, but the following errors occurd.
Is this a bug or am I missing something?
### Reproduction
```python
import torch
from diffusers import EulerDiscreteScheduler
fr... | closed | completed | false | 4 | [
"bug"
] | [
"DN6"
] | 2023-08-03T03:33:53Z | 2023-12-14T07:15:55Z | 2023-08-17T03:15:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lss233 | 8,984,680 | MDQ6VXNlcjg5ODQ2ODA= | User | false |
huggingface/diffusers | 1,834,294,997 | I_kwDOHa8MBc5tVR7V | 4,447 | https://github.com/huggingface/diffusers/issues/4447 | https://api.github.com/repos/huggingface/diffusers/issues/4447 | DeepFloyd/IF-I-XL-v1.0 is restricted and you are not in the authorized list | ### Describe the bug
While running:
```
# stage 1
stage_1 = DiffusionPipeline.from_pretrained("DeepFloyd/IF-I-XL-v1.0", variant="fp16", torch_dtype=torch.float16)
stage_1.enable_xformers_memory_efficient_attention() # remove line if torch.__version__ >= 2.0.0
stage_1.enable_model_cpu_offload()
# stage 2
stag... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-03T05:21:44Z | 2023-08-03T05:34:35Z | 2023-08-03T05:30:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,834,443,445 | I_kwDOHa8MBc5tV2K1 | 4,448 | https://github.com/huggingface/diffusers/issues/4448 | https://api.github.com/repos/huggingface/diffusers/issues/4448 | Outpainting results from diffusers' StableDiffusionControlNetPipeline is much worse than those from A1111 webui. How to improve? | I am trying to outpaint some human images (mainly the lower-body part) with SD 1.5 conditioned on ControlNet's inpainting and openpose. I have been using A1111 webui with ControlNet extension and it has been working quite well:
Here are my settings in the webui:
<img width="774" alt="Screenshot 2023-08-03 at 15 08 30... | closed | completed | false | 4 | [] | [] | 2023-08-03T07:19:12Z | 2023-08-30T05:35:03Z | 2023-08-03T11:39:51Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xiyichen | 50,854,238 | MDQ6VXNlcjUwODU0MjM4 | User | false |
huggingface/diffusers | 1,834,444,613 | I_kwDOHa8MBc5tV2dF | 4,449 | https://github.com/huggingface/diffusers/issues/4449 | https://api.github.com/repos/huggingface/diffusers/issues/4449 | Backpropagation in zeroscope / TextToVideoSDPipeline | Hi,
What are the GPU memory requirements for backpropagation through zeroscope (TextToVideoSDPipeline)?
Thanks! | closed | completed | false | 2 | [
"stale"
] | [] | 2023-08-03T07:20:04Z | 2023-09-12T15:03:48Z | 2023-09-12T15:03:48Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | omerbt | 52,277,000 | MDQ6VXNlcjUyMjc3MDAw | User | false |
huggingface/diffusers | 1,834,508,483 | I_kwDOHa8MBc5tWGDD | 4,450 | https://github.com/huggingface/diffusers/issues/4450 | https://api.github.com/repos/huggingface/diffusers/issues/4450 | example/texual_inversion/ multi text inversion is actually not working. | ### Describe the bug
https://github.com/huggingface/diffusers/blob/4c4fe042a7f718f30f6bd897b967ff800e035347/examples/textual_inversion/textual_inversion.py#L629C1-L648
https://github.com/huggingface/diffusers/blob/4c4fe042a7f718f30f6bd897b967ff800e035347/examples/textual_inversion/textual_inversion.py#L706-L716https:... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-08-03T08:03:37Z | 2023-08-04T10:21:33Z | 2023-08-04T10:21:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Suho0922 | 47,239,412 | MDQ6VXNlcjQ3MjM5NDEy | User | false |
huggingface/diffusers | 1,834,866,507 | I_kwDOHa8MBc5tXddL | 4,453 | https://github.com/huggingface/diffusers/issues/4453 | https://api.github.com/repos/huggingface/diffusers/issues/4453 | How to convert diffusers SDXL lora into safetensors that works with AUTO1111 webui | ### Describe the bug
I trained a lora on SDXL with this diffusers script: https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth_lora_sdxl.py
I get great results when using the output .bin with the diffusers inference code.
How can I convert the .bin to .safetensors that can be loa... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-08-03T11:23:25Z | 2023-09-12T15:03:46Z | 2023-09-12T15:03:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wangqyqq | 11,223,661 | MDQ6VXNlcjExMjIzNjYx | User | false |
huggingface/diffusers | 1,835,449,039 | I_kwDOHa8MBc5tZrrP | 4,456 | https://github.com/huggingface/diffusers/issues/4456 | https://api.github.com/repos/huggingface/diffusers/issues/4456 | The loaders load_lora_into_unet() function throw out "AttributeError: 'ModuleList' object has no attribute '4'" | ### Describe the bug
In Diffusers Version: 0.20.0.dev0
When loading SDXL LoRA to SDXL pipeline.
Works by
```python
lora_path = r"path/to/SDXLLoRA.safetensors"
base_pipe.load_lora_weights(lora_path)
```
However, when extracting state_dict, and network_alphas manually feed into unet like this:
```... | closed | completed | false | 16 | [
"bug"
] | [] | 2023-08-03T17:21:24Z | 2024-02-23T12:57:21Z | 2023-08-06T05:45:40Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xhinker | 7,898,116 | MDQ6VXNlcjc4OTgxMTY= | User | false |
huggingface/diffusers | 1,835,941,247 | I_kwDOHa8MBc5tbj1_ | 4,459 | https://github.com/huggingface/diffusers/issues/4459 | https://api.github.com/repos/huggingface/diffusers/issues/4459 | how to convert a picture to text embedding, without training these image model like Textual Inversion | clip text: tokens -> text_embedding -> text_features
clip img: img -> img_embedding -> img_features
how inversion without training every time: img -> text_embedding | closed | completed | false | 2 | [
"stale"
] | [] | 2023-08-04T01:46:25Z | 2023-09-12T15:03:45Z | 2023-09-12T15:03:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yanchaoguo | 11,827,802 | MDQ6VXNlcjExODI3ODAy | User | false |
huggingface/diffusers | 1,835,965,936 | I_kwDOHa8MBc5tbp3w | 4,460 | https://github.com/huggingface/diffusers/issues/4460 | https://api.github.com/repos/huggingface/diffusers/issues/4460 | Different results in stable diffusion and diffusers | I can get the same result in stable diffusion and diffusers using txt2img.
diffusers
stable diffusion webui
... | closed | completed | false | 2 | [] | [] | 2023-08-04T02:25:25Z | 2023-08-14T02:01:53Z | 2023-08-04T02:34:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Laidawang | 85,244,566 | MDQ6VXNlcjg1MjQ0NTY2 | User | false |
huggingface/diffusers | 1,836,208,431 | I_kwDOHa8MBc5tclEv | 4,463 | https://github.com/huggingface/diffusers/issues/4463 | https://api.github.com/repos/huggingface/diffusers/issues/4463 | When will guide diffusion and unet_1d_condition be supported ？ | Just like titles,
I need guide diffusion’s feature and unet_1d_condition, but I don't find them in this repo,
please support them,
thx !
| closed | completed | false | 3 | [
"stale"
] | [] | 2023-08-04T07:20:28Z | 2023-09-12T15:03:44Z | 2023-09-12T15:03:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | YChienHung | 41,673,864 | MDQ6VXNlcjQxNjczODY0 | User | false |
huggingface/diffusers | 1,836,308,865 | I_kwDOHa8MBc5tc9mB | 4,464 | https://github.com/huggingface/diffusers/issues/4464 | https://api.github.com/repos/huggingface/diffusers/issues/4464 | CLIPSeg + StableDiffusionInpainting | ### Describe the bug
```
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
[<ipython-input-10-1ff2265a51a8>](https://localhost:8080/#) in <cell line: 7>()
5 # model = CLIPSegForImageSegmentation.from_pretr... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-08-04T08:29:43Z | 2023-08-09T10:15:17Z | 2023-08-09T09:58:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,836,431,466 | I_kwDOHa8MBc5tdbhq | 4,466 | https://github.com/huggingface/diffusers/issues/4466 | https://api.github.com/repos/huggingface/diffusers/issues/4466 | Dynamic threasholding example and mimic scale please | Hi. I know that dynamic threasholding is implemented, however there are no examples on how to use them. Also, in this repo: https://github.com/mcmonkeyprojects/sd-dynamic-thresholding/tree/master there's a variable called mimic scale which is nowhere to be found in diffusers | closed | completed | false | 8 | [] | [] | 2023-08-04T09:51:53Z | 2023-09-14T14:11:18Z | 2023-09-14T14:11:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,836,436,890 | I_kwDOHa8MBc5tdc2a | 4,467 | https://github.com/huggingface/diffusers/issues/4467 | https://api.github.com/repos/huggingface/diffusers/issues/4467 | DWPose controlnet | here's the most precise openpose controlnet derivative I've ever seen: https://github.com/IDEA-Research/DWPose/tree/main
Absolutely worth implementing! This is beyond amazing and all who tried using the regular controlnet openpose can tell you how much better it is and the paper shows direct examples! Insane precisi... | closed | completed | false | 6 | [
"community-examples",
"stale",
"contributions-welcome"
] | [] | 2023-08-04T09:55:39Z | 2023-12-20T10:34:05Z | 2023-11-26T15:07:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,836,569,964 | I_kwDOHa8MBc5td9Vs | 4,468 | https://github.com/huggingface/diffusers/issues/4468 | https://api.github.com/repos/huggingface/diffusers/issues/4468 | The performance of the sample code is different when running on multiple GPUs versus a single GPU. | ### Describe the bug
I tried to train unconditional diffusion with accelerate on a dataset generated from .ttf files. And I find that the performance of the sample code is different when running on multiple GPUs versus a single GPU.
I trained on two A100 GPUs with a learning rate of 1e-4 and batch size of 32, and ... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-08-04T11:24:02Z | 2023-08-04T19:40:48Z | 2023-08-04T19:40:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pengyrDL | 138,228,198 | U_kgDOCD0x5g | User | false |
huggingface/diffusers | 1,367,748,440 | I_kwDOHa8MBc5Rhi9Y | 447 | https://github.com/huggingface/diffusers/issues/447 | https://api.github.com/repos/huggingface/diffusers/issues/447 | Prompt truncation detection and return of tokenized prompts | **Is your feature request related to a problem? Please describe.**
There are examples of long prompts on the web, but many people are unaware that long strings are truncated.
However, there is no way to know wheter `StableDifusionPipelineOutputprompt` truncated long `prompt` or not.
Long prompts are truncated [h... | closed | completed | false | 1 | [] | [] | 2022-09-09T12:18:09Z | 2022-09-27T09:14:17Z | 2022-09-27T09:14:17Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shirayu | 963,961 | MDQ6VXNlcjk2Mzk2MQ== | User | false |
huggingface/diffusers | 1,836,667,488 | I_kwDOHa8MBc5teVJg | 4,471 | https://github.com/huggingface/diffusers/issues/4471 | https://api.github.com/repos/huggingface/diffusers/issues/4471 | StableDiffusionXLImg2ImgPipeline can't load lora_weight | This script returns error
~~~python
from diffusers import StableDiffusionXLImg2ImgPipeline
import torch
pipe = StableDiffusionXLImg2ImgPipeline.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1.0",
torch_dtype=torch.float16)
pipe.load_lora_weights(
"stabilityai/stable-diffusion-xl-base-1.... | closed | completed | false | 1 | [] | [] | 2023-08-04T12:33:23Z | 2023-08-08T23:08:07Z | 2023-08-08T23:08:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dai-ichiro | 53,703,225 | MDQ6VXNlcjUzNzAzMjI1 | User | false |
huggingface/diffusers | 1,837,477,680 | I_kwDOHa8MBc5tha8w | 4,476 | https://github.com/huggingface/diffusers/issues/4476 | https://api.github.com/repos/huggingface/diffusers/issues/4476 | fail to get expected results on sdxl+controlnet on fill50k | ### Describe the bug
The generated images seem not to correspond to the input condition images after around 10000 iterations.
### Reproduction
```
CUDA_VISIBLE_DEVICES='0' accelerate launch train_controlnet_sdxl.py \
--pretrained_model_name_or_path=$MODEL_DIR \
--output_dir=$OUTPUT_DIR \
--pretrained_va... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-08-05T00:41:06Z | 2023-11-08T15:09:53Z | 2023-11-08T15:09:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | YongtaoGE | 22,744,013 | MDQ6VXNlcjIyNzQ0MDEz | User | false |
huggingface/diffusers | 1,837,534,201 | I_kwDOHa8MBc5thov5 | 4,478 | https://github.com/huggingface/diffusers/issues/4478 | https://api.github.com/repos/huggingface/diffusers/issues/4478 | DiffusionPipeline.from_pretrained SDXL model: Input type and bias type should be the same | ### Describe the bug
I am running into an error when generating images. I tried the below fix and it only let me generate the first image, subsequent images would fail to the error below.
for module in pipe.vae.modules():
if isinstance(module, torch.nn.Conv2d):
module.half()
### Reproduction
... | closed | not_planned | false | 8 | [
"bug"
] | [] | 2023-08-05T02:43:50Z | 2023-08-08T14:02:59Z | 2023-08-05T15:26:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | caldwecg | 69,778,732 | MDQ6VXNlcjY5Nzc4NzMy | User | false |
huggingface/diffusers | 1,367,767,900 | I_kwDOHa8MBc5Rhntc | 448 | https://github.com/huggingface/diffusers/issues/448 | https://api.github.com/repos/huggingface/diffusers/issues/448 | I made a web GUI using diffusers in the backend for Stable Diffusion | It's using a unified pipeline which allows to do either text-to-image, image-to-image or inpainting using the same backend.
Including doing image-to-image inside an inpainting region!
I made a short video to demo the proces... | closed | completed | false | 0 | [] | [] | 2022-09-09T12:35:16Z | 2022-09-22T17:08:36Z | 2022-09-22T17:08:36Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | leszekhanusz | 4,641,499 | MDQ6VXNlcjQ2NDE0OTk= | User | false |
huggingface/diffusers | 1,837,588,768 | I_kwDOHa8MBc5th2Eg | 4,480 | https://github.com/huggingface/diffusers/issues/4480 | https://api.github.com/repos/huggingface/diffusers/issues/4480 | Generation using StableDiffusionPipeline with torch_dtype=torch.float16 and mps crashes the kernel on Mac M1 | ### Describe the bug
If I add torch_dtype=torch.float16 to any model the Python kernel stops/crashes when trying to generate images, works well when I don't add that setting. I have a Mac M1 so I run with the .to("mps").
### Reproduction
# Recreate bug code
from diffusers import StableDiffusionPipeline
import torc... | closed | completed | false | 19 | [] | [] | 2023-08-05T05:39:57Z | 2024-11-25T13:08:29Z | 2024-11-25T13:08:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jorgen-h | 36,112,509 | MDQ6VXNlcjM2MTEyNTA5 | User | false |
huggingface/diffusers | 1,837,761,435 | I_kwDOHa8MBc5tigOb | 4,483 | https://github.com/huggingface/diffusers/issues/4483 | https://api.github.com/repos/huggingface/diffusers/issues/4483 | Stable Diffusion XL Pipeline generator argument is not deterministic | ### Describe the bug
Passing a torch seeded generator
```
generator = torch.Generator("cuda")
generator.manual_seed(1337)
```
Does not seem to matter when passed to the SDXL pipe, the generator seems to be completely random. I did some digging but could not find the root cause of the issue.
### Reproductio... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-05T13:26:08Z | 2023-08-09T02:36:46Z | 2023-08-09T02:36:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CoffeeVampir3 | 48,565,901 | MDQ6VXNlcjQ4NTY1OTAx | User | false |
huggingface/diffusers | 1,837,788,101 | I_kwDOHa8MBc5timvF | 4,484 | https://github.com/huggingface/diffusers/issues/4484 | https://api.github.com/repos/huggingface/diffusers/issues/4484 | SDXL Kohya LoRA sometimes produces weird results | ### Describe the bug
I am using both the SDXL 1.0 Base and the refiner model to generate images. I have two different LoRAs being loaded after each other into the base model (see reproduction)
Both LoRAs are trained using the Kohya GUI.
I uploaded both LoRAs here:
1. https://drive.google.com/file/d/1thZQt5-1uiJ... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-05T14:30:47Z | 2023-08-06T14:56:48Z | 2023-08-06T14:56:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LrsChrSch | 86,013,790 | MDQ6VXNlcjg2MDEzNzkw | User | false |
huggingface/diffusers | 1,837,841,846 | I_kwDOHa8MBc5tiz22 | 4,487 | https://github.com/huggingface/diffusers/issues/4487 | https://api.github.com/repos/huggingface/diffusers/issues/4487 | Remove attention slicing from docs | ### Describe the bug
The attention slicing function as described in:
**Sliced attention for additional memory savings**
under https://huggingface.co/docs/diffusers/optimization/fp16
Is very outdated and we should probably remove the section from the optimizations docs.
90% of users use either xformers or Tor... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-08-05T16:47:13Z | 2023-08-10T18:00:04Z | 2023-08-10T18:00:04Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,837,935,949 | I_kwDOHa8MBc5tjK1N | 4,488 | https://github.com/huggingface/diffusers/issues/4488 | https://api.github.com/repos/huggingface/diffusers/issues/4488 | download_from_original_stable_diffusion_ckpt does not load SDXL checkpoints without specifying pipeline_class | ### Describe the bug
https://github.com/huggingface/diffusers/commit/94b332c47665cb85b22a54545d7fb7f6478ea72d changed a hardcoded class reference of `StableDiffusionXLPipeline` to `pipeline_class`, `pipeline_class` defaults to `StableDiffusionPipeline` so it tries to load the SDXL checkpoint as a SD checkpoint.
P... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-08-05T20:56:18Z | 2023-08-09T08:32:36Z | 2023-08-09T08:32:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | allenbenz | 11,370,198 | MDQ6VXNlcjExMzcwMTk4 | User | false |
huggingface/diffusers | 1,838,016,347 | I_kwDOHa8MBc5tjedb | 4,490 | https://github.com/huggingface/diffusers/issues/4490 | https://api.github.com/repos/huggingface/diffusers/issues/4490 | convert model ckpt to diffusers "text_config_dict...." | Hi, I am training with dreambooth. When I want to convert my model to a diffuser I receive the following message and I don't know if it's a problem with my model or something I'm missing.
Note: the conversion process ends and I can train, but I want to know what these lines mean:
!python /content/diffusers/script... | closed | completed | false | 1 | [] | [] | 2023-08-06T02:36:39Z | 2023-08-09T08:32:05Z | 2023-08-09T08:32:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | josemerinom | 20,940,429 | MDQ6VXNlcjIwOTQwNDI5 | User | false |
huggingface/diffusers | 1,838,023,889 | I_kwDOHa8MBc5tjgTR | 4,491 | https://github.com/huggingface/diffusers/issues/4491 | https://api.github.com/repos/huggingface/diffusers/issues/4491 | SDXL Refiner broken in latest diffusers | ### Describe the bug
Using the example "ensemble of experts" code produces this error: TypeError: StableDiffusionXLPipeline.call() got an unexpected keyword argument 'denoising_start'
### Reproduction
Use example code from e.g. https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0 for ensemble of experts w... | closed | completed | false | 8 | [
"bug"
] | [] | 2023-08-06T03:10:34Z | 2023-09-27T19:39:19Z | 2023-08-07T11:50:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | catid | 252,193 | MDQ6VXNlcjI1MjE5Mw== | User | false |
huggingface/diffusers | 1,838,098,602 | I_kwDOHa8MBc5tjyiq | 4,492 | https://github.com/huggingface/diffusers/issues/4492 | https://api.github.com/repos/huggingface/diffusers/issues/4492 | NameError: name 'false' is not defined. Did you mean: 'False'? | ### Describe the bug
I ran the demo directly
https://huggingface.co/docs/diffusers/training/lora#dreambooth
### Reproduction
https://huggingface.co/docs/diffusers/training/lora#dreambooth
The code is used from here with almost no changes
### Logs
```shell
NameError: name 'false' is not defined. Did you mean: 'Fa... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2023-08-06T07:55:29Z | 2023-11-20T15:06:53Z | 2023-11-20T15:06:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jqsl2012 | 6,802,595 | MDQ6VXNlcjY4MDI1OTU= | User | false |
huggingface/diffusers | 1,838,202,996 | I_kwDOHa8MBc5tkMB0 | 4,494 | https://github.com/huggingface/diffusers/issues/4494 | https://api.github.com/repos/huggingface/diffusers/issues/4494 | How to convert a diffuser pipeline of XL to checkpoint or safetensors | I need to fine-tune stable diffusion unet or something like that. Then I have to convert the pipeline into ckpt for webui usage.
Before I use the `scripts/convert_diffusers_to_original_stable_diffusion.py` for transforming.
But currently it cannot convert correctly for XL pipeline and webui may raise bugs.
Thanks i... | closed | completed | false | 4 | [
"stale",
"contributions-welcome"
] | [] | 2023-08-06T13:06:54Z | 2023-11-06T04:42:19Z | 2023-10-18T15:14:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FeiiYin | 28,101,823 | MDQ6VXNlcjI4MTAxODIz | User | false |
huggingface/diffusers | 1,838,222,162 | I_kwDOHa8MBc5tkQtS | 4,496 | https://github.com/huggingface/diffusers/issues/4496 | https://api.github.com/repos/huggingface/diffusers/issues/4496 | Prompt fusion | **Is your feature request related to a problem? Please describe.**
I would like to get a wider range of results
**Describe the solution you'd like**
Prompt fusion
**Describe alternatives you've considered**
Doing it myself, but I'd have to redo it for every new pipeline
**Additional context**
https://githu... | closed | completed | false | 15 | [] | [] | 2023-08-06T14:05:12Z | 2023-11-08T15:09:43Z | 2023-11-08T15:09:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,838,233,830 | I_kwDOHa8MBc5tkTjm | 4,497 | https://github.com/huggingface/diffusers/issues/4497 | https://api.github.com/repos/huggingface/diffusers/issues/4497 | Dynamic thresholding (full) | **Is your feature request related to a problem? Please describe.**
I am trying to increase the variety and quality of my results
**Describe the solution you'd like**
I would like dynamic thresholding implemented
**Describe alternatives you've considered**
Doing it myself, paying someone better than to do it or... | open | reopened | false | 10 | [
"stale",
"contributions-welcome"
] | [] | 2023-08-06T14:40:42Z | 2024-04-07T15:04:59Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,838,668,105 | I_kwDOHa8MBc5tl9lJ | 4,504 | https://github.com/huggingface/diffusers/issues/4504 | https://api.github.com/repos/huggingface/diffusers/issues/4504 | Loading Lora safetensor: local variable 'state_dict' referenced before assignment | ### Describe the bug
There are 2 errors:
1.
```
!python ./diffusers/scripts/convert_original_stable_diffusion_to_diffusers.py --checkpoint_path ./loras/test_model.safetensors --dump_path ./test_model --from_safetensors
2023-08-07 04:40:53.152295: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-08-07T04:47:30Z | 2023-08-17T04:04:05Z | 2023-08-17T04:04:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,839,161,965 | I_kwDOHa8MBc5tn2Jt | 4,506 | https://github.com/huggingface/diffusers/issues/4506 | https://api.github.com/repos/huggingface/diffusers/issues/4506 | prompt_embeds variable is overwritten | ### Describe the bug
In example dreambooth/train_dreambooth_lora_sdxl.py, variable prompt_embeds is overwritten, making its shape increases in each step.
`
elems_to_repeat = bsz // 2 if args.with_prior_preservation else bsz
if not args.train_text_encoder:
... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-08-07T10:29:35Z | 2023-08-09T08:23:08Z | 2023-08-09T08:23:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghost | 10,137 | MDQ6VXNlcjEwMTM3 | User | false |
huggingface/diffusers | 1,839,175,635 | I_kwDOHa8MBc5tn5fT | 4,507 | https://github.com/huggingface/diffusers/issues/4507 | https://api.github.com/repos/huggingface/diffusers/issues/4507 | How to train stable-diffusion-xl-base-1.0 without lora? | Hi, I want to train `stable-diffusion-xl-base-1.0` without lora, how to do this?
I can run `train_text_to_image_lora_sdxl.py` .
But `train_text_to_image.py` with `MODEL_NAME="stabilityai/stable-diffusion-xl-base-1.0"` with raise an error:
```
diffusers/models/unet_2d_condition.py:836 in forward ... | closed | completed | false | 4 | [] | [] | 2023-08-07T10:38:24Z | 2023-08-14T07:25:49Z | 2023-08-09T08:23:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | KimmiShi | 13,433,751 | MDQ6VXNlcjEzNDMzNzUx | User | false |
huggingface/diffusers | 1,839,434,780 | I_kwDOHa8MBc5to4wc | 4,511 | https://github.com/huggingface/diffusers/issues/4511 | https://api.github.com/repos/huggingface/diffusers/issues/4511 | StableDiffusionXLInpaintPipeline and StableDiffusionXLImg2ImgPipeline | I have tested using the `load_lora_weights()` and `save_lora_weights()` functions from `StableDiffusionXLPipeline` in `StableDiffusionXLInpaintPipeline` and `StableDiffusionXLImg2ImgPipeline`, and they work well.
Can I open a PR?
---
This is likely happening because the definitions of `load_lora... | closed | completed | false | 1 | [] | [] | 2023-08-07T13:13:45Z | 2023-08-08T13:23:24Z | 2023-08-08T13:23:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wooyeolbaek | 75,835,117 | MDQ6VXNlcjc1ODM1MTE3 | User | false |
huggingface/diffusers | 1,368,228,573 | I_kwDOHa8MBc5RjYLd | 452 | https://github.com/huggingface/diffusers/issues/452 | https://api.github.com/repos/huggingface/diffusers/issues/452 | Performance degradation in `mps` after `einsum` replacement | Before #445 was merged I was getting ~31s inference time in `mps`. After the change, time goes up to 42s. I verified again in `main` @ `b2b3b1a`, and time is again 31s.
I haven't checked other platforms yet.
Any ideas, @patil-suraj? | closed | completed | false | 15 | [
"stale"
] | [] | 2022-09-09T19:15:43Z | 2022-10-29T18:39:45Z | 2022-10-29T18:39:45Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pcuenca | 1,177,582 | MDQ6VXNlcjExNzc1ODI= | User | false |
huggingface/diffusers | 1,840,231,464 | I_kwDOHa8MBc5tr7Qo | 4,520 | https://github.com/huggingface/diffusers/issues/4520 | https://api.github.com/repos/huggingface/diffusers/issues/4520 | LoRA scale kwarg doesn't work with Attention Slicing | ### Describe the bug
When I try to use LoRA scale with Attention Slicing, it results in this error:
```shell
TypeError: SlicedAttnProcessor.__call__() got an unexpected keyword argument 'scale'
```
### Reproduction
Here's a simplified reproduction:
I load my pipeline and then LoRA weights.
After my pipe... | closed | completed | false | 10 | [
"bug"
] | [] | 2023-08-07T21:28:23Z | 2023-08-10T19:32:25Z | 2023-08-10T19:32:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | soten355 | 8,527,497 | MDQ6VXNlcjg1Mjc0OTc= | User | false |
huggingface/diffusers | 1,840,477,819 | I_kwDOHa8MBc5ts3Z7 | 4,522 | https://github.com/huggingface/diffusers/issues/4522 | https://api.github.com/repos/huggingface/diffusers/issues/4522 | Custom Inpainting models on StableDiffusionControlNetInpaintPipeline | Is there a way to use custom Inpainting models like SG161222/Realistic_Vision_V5.1_noVAE on StableDiffusionControlNetInpaintPipeline?
Can somebody help?
| closed | completed | false | 1 | [] | [] | 2023-08-08T02:57:08Z | 2023-08-09T08:17:02Z | 2023-08-09T08:17:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vijishmadhavan | 53,169,213 | MDQ6VXNlcjUzMTY5MjEz | User | false |
huggingface/diffusers | 1,840,585,694 | I_kwDOHa8MBc5ttRve | 4,524 | https://github.com/huggingface/diffusers/issues/4524 | https://api.github.com/repos/huggingface/diffusers/issues/4524 | AnimateDiff in Diffusers Library | ### Model/Pipeline/Scheduler description
I think it would be cool to work on turning AnimateDiff into a diffusers pipeline that can be used via the library.
### Open source status
- [X] The model implementation is available
- [X] The model weights are available (Only relevant if addition is not a scheduler).
### Pr... | closed | completed | false | 20 | [
"stale"
] | [
"DN6",
"aandyw"
] | 2023-08-08T05:22:57Z | 2023-11-13T17:26:58Z | 2023-11-08T15:09:39Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aandyw | 37,781,802 | MDQ6VXNlcjM3NzgxODAy | User | false |
huggingface/diffusers | 1,840,596,504 | I_kwDOHa8MBc5ttUYY | 4,525 | https://github.com/huggingface/diffusers/issues/4525 | https://api.github.com/repos/huggingface/diffusers/issues/4525 | TensorRT inpainting bug . | When I load Engine model , there is a error when I inference single image ok .
`cost time , 1.0900559425354004
inference ok .
Exception ignored in: <function Engine.__del__ at 0x7f90f32c44c0>
Traceback (most recent call last):
File "/root/.cache/huggingface/modules/diffusers_modules/git/stable_diffusion_tensor... | closed | completed | false | 1 | [] | [] | 2023-08-08T05:36:03Z | 2023-08-09T08:12:20Z | 2023-08-09T08:12:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kelisiya | 23,169,766 | MDQ6VXNlcjIzMTY5NzY2 | User | false |
huggingface/diffusers | 1,840,665,874 | I_kwDOHa8MBc5ttlUS | 4,527 | https://github.com/huggingface/diffusers/issues/4527 | https://api.github.com/repos/huggingface/diffusers/issues/4527 | Prioritise Pipelines for CI/CD testing | BTW here a list of the importance on pipelines. This list comes from looking at:
- a) downloads of these pipeline classes (a bit inaccurate since we don't always know when a certain pipeline class is used)
- b) downloads of specific checkpoint that we know are related to certain pipelines
- c) pageview... | closed | completed | false | 1 | [
"stale"
] | [] | 2023-08-08T06:43:27Z | 2023-10-18T15:14:11Z | 2023-10-18T15:14:11Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DN6 | 7,529,846 | MDQ6VXNlcjc1Mjk4NDY= | User | false |
huggingface/diffusers | 1,840,694,316 | I_kwDOHa8MBc5ttsQs | 4,528 | https://github.com/huggingface/diffusers/issues/4528 | https://api.github.com/repos/huggingface/diffusers/issues/4528 | why some lora not effective, get snowy/blurry images | Hi, I got a problem when using lora, I downloaded a lora with safetensors from civitai, this is the link: https://civitai.com/models/121706, I used the script[ convert_lora_safetensor_to_diffusers.py ](https://github.com/huggingface/diffusers/blob/v0.17.0/scripts/convert_lora_safetensor_to_diffusers.py)to load lora, th... | closed | completed | false | 18 | [
"stale"
] | [] | 2023-08-08T07:05:34Z | 2023-11-08T15:09:37Z | 2023-11-08T15:09:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zxf2012 | 2,860,824 | MDQ6VXNlcjI4NjA4MjQ= | User | false |
huggingface/diffusers | 1,840,986,655 | I_kwDOHa8MBc5tuzof | 4,529 | https://github.com/huggingface/diffusers/issues/4529 | https://api.github.com/repos/huggingface/diffusers/issues/4529 | TypeError: LoraLoaderMixin.load_lora_into_unet() got an unexpected keyword argument 'network_alpha' | ### Describe the bug
TypeError: LoraLoaderMixin.load_lora_into_unet() got an unexpected keyword argument 'network_alpha'
应该是network_alphas
### Reproduction
accelerate launch train_text_to_image_lora_sdxl.py using 【--resume_from_checkpoint="latest" 】
### Logs
_No response_
### System Info
Copy-and-paste the te... | closed | completed | false | 13 | [
"bug"
] | [] | 2023-08-08T09:58:57Z | 2023-08-12T06:16:03Z | 2023-08-11T08:48:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yanchaoguo | 11,827,802 | MDQ6VXNlcjExODI3ODAy | User | false |
huggingface/diffusers | 1,841,032,540 | I_kwDOHa8MBc5tu-1c | 4,530 | https://github.com/huggingface/diffusers/issues/4530 | https://api.github.com/repos/huggingface/diffusers/issues/4530 | Little artifacts on the img generated by sdxl refiner 1.0, mostly red green and blue | ### Describe the bug
I don't know if it is the problem of the model itself
### Reproduction
pipe = StableDiffusionXLImg2ImgPipeline.from_pretrained(
sd_model_path, torch_dtype=torch.float16, requ... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-08T10:28:17Z | 2023-08-09T08:08:35Z | 2023-08-09T08:08:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | WaltBao | 106,211,276 | U_kgDOBlSnzA | User | false |
huggingface/diffusers | 1,841,549,484 | I_kwDOHa8MBc5tw9Cs | 4,533 | https://github.com/huggingface/diffusers/issues/4533 | https://api.github.com/repos/huggingface/diffusers/issues/4533 | How to debug custom pipeline locally ? | Hi,
I build diffusers from source, and I am using ControlNet. However, diffusers seems not to load the custom pipeline from ```diffusers/examples/community/stable_diffusion_controlnet_img2img.py``` as I expected. Instead, it seems to download from the hub and cache a new ```stable_diffusion_controlnet_img2img.py`... | closed | completed | false | 4 | [] | [] | 2023-08-08T15:34:40Z | 2023-08-09T12:17:42Z | 2023-08-09T08:06:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pansanity666 | 47,111,102 | MDQ6VXNlcjQ3MTExMTAy | User | false |
huggingface/diffusers | 1,841,731,959 | I_kwDOHa8MBc5txpl3 | 4,534 | https://github.com/huggingface/diffusers/issues/4534 | https://api.github.com/repos/huggingface/diffusers/issues/4534 | SDXL dreambooth crashes during validation on mps: Expected dst.dim() >= src.dim() to be true, but got false | ### Describe the bug
When running the dreambooth SDXL training, I get a crash during validation
Expected dst.dim() >= src.dim() to be true, but got false (see below)
### Reproduction
Run the tutorial at examples/dreambooth/README_sdxl.md
my training settings
```
export MODEL_NAME="stabilityai/stable-... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-08-08T17:36:52Z | 2023-08-09T08:05:26Z | 2023-08-09T08:05:25Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rrva | 887,132 | MDQ6VXNlcjg4NzEzMg== | User | false |
huggingface/diffusers | 1,842,383,721 | I_kwDOHa8MBc5t0Itp | 4,537 | https://github.com/huggingface/diffusers/issues/4537 | https://api.github.com/repos/huggingface/diffusers/issues/4537 | Question regarding SDXL micro-conditioning and diffusers SDXL training | In the past with SD 1.x and SD 2.x, we just need to set the resolution with the SD training scripts.
However, with SDXL 0.9/1.0 with the micro-conditioning with original image size, crop coordinates, target size, does this mean if we finetune the model on 1024x1024, the model COULD be largely unchanged in other reso... | closed | completed | false | 10 | [] | [] | 2023-08-09T03:13:30Z | 2023-08-14T05:04:02Z | 2023-08-14T05:04:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xiankgx | 4,113,258 | MDQ6VXNlcjQxMTMyNTg= | User | false |
huggingface/diffusers | 1,842,642,568 | I_kwDOHa8MBc5t1H6I | 4,538 | https://github.com/huggingface/diffusers/issues/4538 | https://api.github.com/repos/huggingface/diffusers/issues/4538 | convert inpainting tensorRT OutOfMemory in 3090 | Why do we need so much cuda memory? this is unreasonable | closed | completed | false | 2 | [] | [] | 2023-08-09T07:19:58Z | 2024-03-04T10:10:47Z | 2023-08-09T08:04:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kelisiya | 23,169,766 | MDQ6VXNlcjIzMTY5NzY2 | User | false |
huggingface/diffusers | 1,368,312,157 | I_kwDOHa8MBc5Rjsld | 454 | https://github.com/huggingface/diffusers/issues/454 | https://api.github.com/repos/huggingface/diffusers/issues/454 | The first image generated with LMSScheduler is completely green | ### Describe the bug
The first image generated with LMSScheduler is just a green image.
### Reproduction
Run with:
lms = LMSDiscreteScheduler(beta_start=0.00085, beta_end=0.012, beta_schedule="scaled_linear"
pipe = StableDiffusionPipeline.from_pretrained(..., scheduler=lms)
instead of using the ... | closed | completed | false | 6 | [
"stale"
] | [] | 2022-09-09T20:48:54Z | 2022-10-12T08:35:44Z | 2022-10-11T17:57:52Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | allo- | 1,482,341 | MDQ6VXNlcjE0ODIzNDE= | User | false |
huggingface/diffusers | 1,842,868,174 | I_kwDOHa8MBc5t1-_O | 4,541 | https://github.com/huggingface/diffusers/issues/4541 | https://api.github.com/repos/huggingface/diffusers/issues/4541 | Feature request for example training code: add safe_serialization=True to save_lora_weights | I think adding 'safe_serialization=True' to save_lora_weights (or pass a boolean from a new argument) could be nice, given the overall shift towards defaulting to safetensors
https://github.com/huggingface/diffusers/blob/e731ae0ec817649bf2c15f9f719269d57062696c/examples/dreambooth/train_dreambooth_lora_sdxl.py#L824... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-08-09T09:44:21Z | 2023-10-18T15:14:04Z | 2023-10-18T15:14:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Norod | 3,617,152 | MDQ6VXNlcjM2MTcxNTI= | User | false |
huggingface/diffusers | 1,842,927,706 | I_kwDOHa8MBc5t2Nha | 4,542 | https://github.com/huggingface/diffusers/issues/4542 | https://api.github.com/repos/huggingface/diffusers/issues/4542 | StableDiffusionXLImg2ImgPipeline cannot produce more than 1 images | ### Describe the bug
If the `num_images_per_prompt=1`, it works fine but any value bigger than 1 produces the following error.
### Reproduction
```python
import torch
from diffusers import AutoencoderKL, DiffusionPipeline, StableDiffusionXLImg2ImgPipeline
vae = AutoencoderKL.from_pretrained("madebyollin... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-08-09T10:20:36Z | 2023-08-10T09:04:19Z | 2023-08-10T09:04:18Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eminn | 378,108 | MDQ6VXNlcjM3ODEwOA== | User | false |
huggingface/diffusers | 1,843,148,005 | I_kwDOHa8MBc5t3DTl | 4,545 | https://github.com/huggingface/diffusers/issues/4545 | https://api.github.com/repos/huggingface/diffusers/issues/4545 | Support controlnet param in AutoPipeline from_pipe() method | Currently from_pipe() method preserves 'controlnet' property of the pipeline.
So the transitions like:
t2i/i2i/inpaint <-> (t2i/i2i/inpaint)_controlnet
are only possible by using from_pretrained method which will take additional memory and compute time. | closed | completed | false | 7 | [] | [
"yiyixuxu"
] | 2023-08-09T12:37:29Z | 2023-09-01T16:54:37Z | 2023-09-01T16:54:37Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | slep0v | 37,597,789 | MDQ6VXNlcjM3NTk3Nzg5 | User | false |
huggingface/diffusers | 1,368,332,386 | I_kwDOHa8MBc5Rjxhi | 455 | https://github.com/huggingface/diffusers/issues/455 | https://api.github.com/repos/huggingface/diffusers/issues/455 | `NotImplementedError` for `aten::index.Tensor` when using `mps` | ### Describe the bug
I tried to use `StableDiffusionPipeline` with the lately added support for M1 (`mps`) as described in this [tweet](https://twitter.com/pcuenq/status/1567927480253808647).
Despite having installed the latest (preview) nighly PyTorch Build via
`pip3 install --pre torch torchvision torchaudio --... | closed | completed | false | 4 | [
"bug"
] | [] | 2022-09-09T21:08:35Z | 2022-09-10T08:17:37Z | 2022-09-10T08:17:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fabianmax | 17,836,552 | MDQ6VXNlcjE3ODM2NTUy | User | false |
huggingface/diffusers | 1,844,041,655 | I_kwDOHa8MBc5t6de3 | 4,551 | https://github.com/huggingface/diffusers/issues/4551 | https://api.github.com/repos/huggingface/diffusers/issues/4551 | dreambooth crash with stabilityai/stable-diffusion-xl-base-1.0 | ### Describe the bug
Crash during training of SDXL model (stabilityai/stable-diffusion-xl-base-1.0) with dreambooth
>
### Reproduction
Following tutorial here:
https://github.com/huggingface/diffusers/tree/main/examples/dreambooth
I'm trying to train a stabilityai/stable-diffusion-xl-base-1.0 model.
When I ... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-08-09T21:27:55Z | 2023-08-10T05:43:15Z | 2023-08-10T03:40:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | EnricoBeltramo | 22,332,651 | MDQ6VXNlcjIyMzMyNjUx | User | false |
huggingface/diffusers | 1,844,074,670 | I_kwDOHa8MBc5t6liu | 4,552 | https://github.com/huggingface/diffusers/issues/4552 | https://api.github.com/repos/huggingface/diffusers/issues/4552 | [Need Help] A training script for train a t2i model to inpainting model | Hi, recently there is small stable diffusion model (SD-Small) provided by the segmind(https://github.com/segmind/distill-sd)
But it is for text-2-img task.
When I use the SD-Small Unet in the StableDiffusionInpaintPipeline, the generated result looks not right(masked area content is totally different from the non-... | closed | completed | false | 1 | [] | [] | 2023-08-09T21:53:17Z | 2023-08-10T03:52:38Z | 2023-08-10T03:52:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Bikesuffer | 46,605,321 | MDQ6VXNlcjQ2NjA1MzIx | User | false |
huggingface/diffusers | 1,844,134,199 | I_kwDOHa8MBc5t60E3 | 4,553 | https://github.com/huggingface/diffusers/issues/4553 | https://api.github.com/repos/huggingface/diffusers/issues/4553 | Incorrect initialization value of guidance_rescale in different pipelines | Hi! I've been trying to understand how SD inference works, and diffusers has been really helpful with that.
As I was going through the code, I found this line of code, which might be a minor bug. The documentation says that it is defaulted to `0.7` but the code has it defaulted to `0.0`. The [arxiv](https://arxiv.or... | closed | completed | false | 2 | [] | [] | 2023-08-09T22:36:44Z | 2023-08-16T21:39:57Z | 2023-08-16T21:39:57Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | a-r-r-o-w | 72,266,394 | MDQ6VXNlcjcyMjY2Mzk0 | User | false |
huggingface/diffusers | 1,844,190,060 | I_kwDOHa8MBc5t7Bts | 4,554 | https://github.com/huggingface/diffusers/issues/4554 | https://api.github.com/repos/huggingface/diffusers/issues/4554 | Kandinsky 2.2 example fails with OOM | ### Describe the bug
I am attempting to execute the basic example given for the Kandinsky 2.2 pipeline in Colab, but I'm getting OOM errors due to what appears to be a memory leak in the prior pipeline.
### Reproduction
[Gist of minimal notebook](https://gist.github.com/bgjackma/8524a16a87bb4661c99d7422c0ec5974)
##... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-08-09T23:47:01Z | 2023-10-18T15:14:00Z | 2023-10-18T15:14:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bgjackma | 6,486,273 | MDQ6VXNlcjY0ODYyNzM= | User | false |
huggingface/diffusers | 1,844,456,636 | I_kwDOHa8MBc5t8Cy8 | 4,557 | https://github.com/huggingface/diffusers/issues/4557 | https://api.github.com/repos/huggingface/diffusers/issues/4557 | AttributeError: 'super' object has no attribute '__getattr__' | ### Describe the bug
When building the DiffusionPipeline from_pretrained, I get the bug:
"AttributeError: 'super' object has no attribute '__getattr__'"
### Reproduction
from diffusers import DiffusionPipeline
import torch
pipe = DiffusionPipeline.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-08-10T05:46:14Z | 2023-08-24T08:44:33Z | 2023-08-24T08:44:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | RossoneriZhao | 102,647,326 | U_kgDOBh5GHg | User | false |
huggingface/diffusers | 1,844,865,725 | I_kwDOHa8MBc5t9mq9 | 4,559 | https://github.com/huggingface/diffusers/issues/4559 | https://api.github.com/repos/huggingface/diffusers/issues/4559 | long weighted prompt support for SDXL? | The commmunity pipeline https://github.com/huggingface/diffusers/blob/main/examples/community/lpw_stable_diffusion.py supports long prompts that exceed the 77 tokens limit and weighted prompts such as "(((masterpiece)))", but it only works with sd 1/2, not SDXL, since SDXL has two sets of text encoders and tokenizers. ... | closed | completed | false | 8 | [] | [] | 2023-08-10T10:07:28Z | 2023-08-18T06:00:12Z | 2023-08-18T06:00:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | evanferguson28 | 137,776,371 | U_kgDOCDZM8w | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.