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 | 1,922,693,111 | I_kwDOHa8MBc5ymff3 | 5,270 | https://github.com/huggingface/diffusers/issues/5270 | https://api.github.com/repos/huggingface/diffusers/issues/5270 | Memory leak in notebook environments | ### Describe the bug
When trying to clear memory for a pipeline some is still in use.
### Reproduction
This becomes evident when we want to test and compare different Stable Diffusion checkpoints (have tried with SDXL).
```
import gc
import torch
def clear_memory(pipeline):
pipeline.to('cpu')
gc.... | closed | not_planned | false | 7 | [
"bug"
] | [] | 2023-10-02T20:53:44Z | 2023-10-06T14:14:34Z | 2023-10-04T14:17:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TimothyAlexisVass | 55,708,319 | MDQ6VXNlcjU1NzA4MzE5 | User | false |
huggingface/diffusers | 1,922,989,827 | I_kwDOHa8MBc5ynn8D | 5,271 | https://github.com/huggingface/diffusers/issues/5271 | https://api.github.com/repos/huggingface/diffusers/issues/5271 | Different learning rates for text encoders and unet in dreambooth setting | **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 [...]
> It has been observed/noted in several studies that text encoders necessitate a diminished learning rate compared to UNet. Please refer to https://rentry.o... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-10-02T23:30:09Z | 2023-12-26T15:10:34Z | 2023-12-26T15:10:34Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shyammarjit | 54,628,184 | MDQ6VXNlcjU0NjI4MTg0 | User | false |
huggingface/diffusers | 1,923,261,828 | I_kwDOHa8MBc5yoqWE | 5,274 | https://github.com/huggingface/diffusers/issues/5274 | https://api.github.com/repos/huggingface/diffusers/issues/5274 | "RuntimeError: Found dtype Half but expected Float" when trying to train unconditional_image_generation using `--mixed_precision=fp16' | ### Describe the bug
"RuntimeError: Found dtype Half but expected Float" when trying to train unconditional_image_generation using `--mixed_precision=fp16'
But it works if I set `--mixed_precision=no`
### Reproduction
interactive --gpus=4 -t 2:20:00
accelerate config defaults
accelerate launch --mixed... | closed | completed | false | 3 | [
"bug"
] | [
"sayakpaul"
] | 2023-10-03T04:09:49Z | 2023-10-16T12:01:57Z | 2023-10-16T08:41:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | linhduongtuan | 22,388,092 | MDQ6VXNlcjIyMzg4MDky | User | false |
huggingface/diffusers | 1,923,940,590 | I_kwDOHa8MBc5yrQDu | 5,277 | https://github.com/huggingface/diffusers/issues/5277 | https://api.github.com/repos/huggingface/diffusers/issues/5277 | Jax to numpy process takes too long time on TPU v4-8 for FlaxStableDiffusionXLPipeline | Hello, I'm trying to FlaxStableDiffusionXLPipeline implementation on TPU v4-8. And I measure the time of each step.
When I give the prompts in groups of 8, the 8-image generation time takes 0.02 seconds (after compiling).
Although this time seems very good, converting the `images` variable from jax to numpy takes... | closed | completed | false | 5 | [] | [] | 2023-10-03T11:41:21Z | 2023-10-13T10:39:20Z | 2023-10-13T10:39:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | gokerguner | 20,363,072 | MDQ6VXNlcjIwMzYzMDcy | User | false |
huggingface/diffusers | 1,375,827,423 | I_kwDOHa8MBc5SAXXf | 528 | https://github.com/huggingface/diffusers/issues/528 | https://api.github.com/repos/huggingface/diffusers/issues/528 | Question about flax model output classes | For example here, it is using `from dataclasses import dataclass`
https://github.com/huggingface/diffusers/blob/d8b0e4f433778f9d4555cfab642e43db3aff4833/src/diffusers/models/unet_2d_condition_flax.py#L22-L23
But transformers equivalents use `@flax.struct.dataclass`. For example [here](https://github.com/huggingface... | closed | completed | false | 3 | [] | [
"kashif",
"pcuenca",
"patrickvonplaten",
"patil-suraj"
] | 2022-09-16T11:39:24Z | 2022-09-18T17:35:38Z | 2022-09-18T17:35:38Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mishig25 | 11,827,707 | MDQ6VXNlcjExODI3NzA3 | User | false |
huggingface/diffusers | 1,924,380,642 | I_kwDOHa8MBc5ys7fi | 5,281 | https://github.com/huggingface/diffusers/issues/5281 | https://api.github.com/repos/huggingface/diffusers/issues/5281 | Using `enable_model_cpu_offload` with multiple XL pipelines triggers error Expected all tensors to be on the same device | Hello,
I am having difficulties to use `enable_model_cpu_offload` in the context of my application where I am loading multiple pipelines in memory (`StableDiffusionXLPipeline`, `StableDiffusionXLImg2ImgPipeline` and `StableDiffusionXLControlNetPipeline`). I read through the documentation here: [Model offloading for ... | closed | completed | false | 9 | [
"stale"
] | [] | 2023-10-03T15:19:35Z | 2023-11-20T15:05:46Z | 2023-11-20T15:05:46Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexisrolland | 13,064,696 | MDQ6VXNlcjEzMDY0Njk2 | User | false |
huggingface/diffusers | 1,924,482,384 | I_kwDOHa8MBc5ytUVQ | 5,282 | https://github.com/huggingface/diffusers/issues/5282 | https://api.github.com/repos/huggingface/diffusers/issues/5282 | KeyError: 'time_embed.0.weight' | ### Describe the bug
Trying to create an EndpointHandler on Hugging Face. Running into KeyError: 'time_embed.0.weight'
### Reproduction
```
from typing import Dict, List, Any
from diffusers import StableDiffusionPipeline
class EndpointHandler():
def __init__(self, path=""):
self.pipeline = StableD... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-10-03T16:12:27Z | 2023-11-20T15:05:44Z | 2023-11-20T15:05:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jonathanrstern | 51,684,334 | MDQ6VXNlcjUxNjg0MzM0 | User | false |
huggingface/diffusers | 1,924,598,822 | I_kwDOHa8MBc5ytwwm | 5,283 | https://github.com/huggingface/diffusers/issues/5283 | https://api.github.com/repos/huggingface/diffusers/issues/5283 | Merging Audio and Video Tracks with Captions | ### Describe the bug
I was not able to add transcripts within the video, output mp4 is in audio format
```
Here is the code for reference:
https://colab.research.google.com/drive/1pRFlp1-OVZmRwyGLlMpexe6tSKVAKmwG?usp=sharing
```
### Reproduction
shared the notebook above
### Logs
_No response_
##... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-10-03T17:22:56Z | 2023-10-04T10:36:55Z | 2023-10-04T10:36:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,925,093,265 | I_kwDOHa8MBc5yvpeR | 5,284 | https://github.com/huggingface/diffusers/issues/5284 | https://api.github.com/repos/huggingface/diffusers/issues/5284 | ControlNet-LLLite support | ### Model/Pipeline/Scheduler description
Hi,
Is it possible to add support for ControlNet-LLLite to diffusers? The architecture is slightly different, but the results and performance look good.
### Open source status
- [X] The model implementation is available
- [X] The model weights are available (Only relevant if... | closed | completed | false | 7 | [
"stale"
] | [
"sayakpaul"
] | 2023-10-03T23:16:34Z | 2023-11-21T03:33:54Z | 2023-11-20T15:05:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mycodeiscat | 47,571,787 | MDQ6VXNlcjQ3NTcxNzg3 | User | false |
huggingface/diffusers | 1,925,447,804 | I_kwDOHa8MBc5yxAB8 | 5,285 | https://github.com/huggingface/diffusers/issues/5285 | https://api.github.com/repos/huggingface/diffusers/issues/5285 | add japanese translation for docs translation | **Is your feature request related to a problem? Please describe.**
Add translation of documentation
**Describe the solution you'd like**
Add translation in japanese
| closed | completed | false | 2 | [] | [] | 2023-10-04T05:45:16Z | 2023-10-25T05:55:51Z | 2023-10-24T18:30:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rajveer43 | 64,583,161 | MDQ6VXNlcjY0NTgzMTYx | User | false |
huggingface/diffusers | 1,375,949,028 | I_kwDOHa8MBc5SA1Dk | 529 | https://github.com/huggingface/diffusers/issues/529 | https://api.github.com/repos/huggingface/diffusers/issues/529 | Allow gradient calculation in `DiffusionPipeline.__call__` | Nowadays the gradient calculation is disabled when calling the `DiffusionPipeline` object (example in [here](https://github.com/huggingface/diffusers/blob/c0493723f7851c4a5cc05ef55d3320bd9ab7718d/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py#L93)).
IMO, this makes the framework less flexible to... | closed | completed | false | 3 | [
"stale"
] | [] | 2022-09-16T13:25:35Z | 2022-11-01T15:03:34Z | 2022-11-01T15:03:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JoaoLages | 17,574,157 | MDQ6VXNlcjE3NTc0MTU3 | User | false |
huggingface/diffusers | 1,926,980,306 | I_kwDOHa8MBc5y22LS | 5,291 | https://github.com/huggingface/diffusers/issues/5291 | https://api.github.com/repos/huggingface/diffusers/issues/5291 | custom CFG? | is it possible to Set CFG to zero after 40% of steps?
I see in this article he does it: https://www.baseten.co/blog/sdxl-inference-in-under-2-seconds-the-ultimate-guide-to-stable-diffusion-optimiza/
He has sample code, but I dont know how to incorporate it.
I couldn't find how to do this in the documentation. he i... | closed | completed | false | 3 | [
"stale"
] | [
"yiyixuxu"
] | 2023-10-04T21:04:31Z | 2023-11-22T15:06:03Z | 2023-11-22T15:06:03Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jtoy | 14,783 | MDQ6VXNlcjE0Nzgz | User | false |
huggingface/diffusers | 1,927,315,901 | I_kwDOHa8MBc5y4IG9 | 5,294 | https://github.com/huggingface/diffusers/issues/5294 | https://api.github.com/repos/huggingface/diffusers/issues/5294 | A small spell issue | ### Describe the bug
Hello! Thank you for spending time to deal with this little issue.
This spell issue comes from ("self.prcoessor.attention_op" in diffusers.models.attention_processor.__file__), and should be replaced to processor. I found it when trying train_text_to_image_lora.py script.
The diffusers.__ver... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-10-05T03:35:51Z | 2023-11-21T15:05:36Z | 2023-11-21T15:05:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | YourLights | 118,877,345 | U_kgDOBxXsoQ | User | false |
huggingface/diffusers | 1,927,619,426 | I_kwDOHa8MBc5y5SNi | 5,295 | https://github.com/huggingface/diffusers/issues/5295 | https://api.github.com/repos/huggingface/diffusers/issues/5295 | LAVIE: High-Quality Video Generation with Cascaded Latent Diffusion Models | ### Model/Pipeline/Scheduler description
This work aims to learn a high-quality text-to-video (T2V) generative model by leveraging a pre-trained text-to-image (T2I) model as a basis. It is a highly desirable yet challenging task to simultaneously a) accomplish the synthesis of visually realistic and temporally coheren... | closed | completed | false | 1 | [
"stale"
] | [] | 2023-10-05T07:48:55Z | 2023-11-21T15:05:34Z | 2023-11-21T15:05:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rajveer43 | 64,583,161 | MDQ6VXNlcjY0NTgzMTYx | User | false |
huggingface/diffusers | 1,928,458,815 | I_kwDOHa8MBc5y8fI_ | 5,300 | https://github.com/huggingface/diffusers/issues/5300 | https://api.github.com/repos/huggingface/diffusers/issues/5300 | Why this package [diffusers] keep connecting the internet! | ### Describe the bug
It takes two minutes to connect to the Internet every time, and I have already downloaded it, so the local files
### Reproduction
It takes two minutes to connect to the Internet every time, and I have already downloaded it, so the local files
### Logs
_No response_
### System Info
Not relate... | closed | completed | false | 26 | [
"bug",
"stale"
] | [] | 2023-10-05T14:35:47Z | 2024-07-30T03:25:27Z | 2023-11-22T15:05:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Louis24 | 18,719,360 | MDQ6VXNlcjE4NzE5MzYw | User | false |
huggingface/diffusers | 1,928,699,810 | I_kwDOHa8MBc5y9Z-i | 5,302 | https://github.com/huggingface/diffusers/issues/5302 | https://api.github.com/repos/huggingface/diffusers/issues/5302 | [Hacktoberfest] DeepFloyd IF docstrings | This feature request is included in this year's Diffusers Hacktoberfest 🎃!
Hacktoberfest's purpose is to give upcoming open-source developers the opportunity to learn about open-source and programming by fixing simple issues (like this one).
How to solve the issue:
- Every issue has a description further belo... | closed | completed | false | 1 | [] | [] | 2023-10-05T16:50:40Z | 2023-10-18T19:40:43Z | 2023-10-18T19:40:43Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | stevhliu | 59,462,357 | MDQ6VXNlcjU5NDYyMzU3 | User | false |
huggingface/diffusers | 1,929,666,410 | I_kwDOHa8MBc5zBF9q | 5,308 | https://github.com/huggingface/diffusers/issues/5308 | https://api.github.com/repos/huggingface/diffusers/issues/5308 | video to gif | ### Describe the bug
```
prior/diffusion_pytorch_model.safetensors not found
---------------------------------------------------------------------------
OSError Traceback (most recent call last)
[<ipython-input-14-505fbd36edae>](https://localhost:8080/#) in <cell line: 7>()
... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-10-06T08:17:57Z | 2023-11-16T15:05:34Z | 2023-11-16T15:05:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andysingal | 20,493,493 | MDQ6VXNlcjIwNDkzNDkz | User | false |
huggingface/diffusers | 1,929,951,939 | I_kwDOHa8MBc5zCLrD | 5,310 | https://github.com/huggingface/diffusers/issues/5310 | https://api.github.com/repos/huggingface/diffusers/issues/5310 | Possibility to set custom timesteps tensor in pipeline | **Is your feature request related to a problem? Please describe.**
I want to be able to manually define timesteps.
**Describe the solution you'd like**
```python
pipe = DiffusionPipeline.from_pretrained(...).to("cuda")
my_timesteps = torch.tensor([901, 801, 701, 601, 501, 401, 301, 201])
pipe.scheduler.timeste... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-10-06T11:17:32Z | 2023-11-16T15:05:32Z | 2023-11-16T15:05:32Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TimothyAlexisVass | 55,708,319 | MDQ6VXNlcjU1NzA4MzE5 | User | false |
huggingface/diffusers | 1,930,013,358 | I_kwDOHa8MBc5zCaqu | 5,311 | https://github.com/huggingface/diffusers/issues/5311 | https://api.github.com/repos/huggingface/diffusers/issues/5311 | AttributeError: 'StableDiffusionXLPipeline' object has no attribute 'load_lora_weights' | from diffusers import DiffusionPipeline
import torch
pipe = DiffusionPipeline.from_pretrained('C:/Users/test/ML/models/majicmix_xl', torch_dtype=torch.bfloat16).to("cuda")
pipe.load_lora_weights("lora.safetensors")
I got AttributeError: 'StableDiffusionXLPipeline' object has no attribute 'load_lora_weights' err... | closed | completed | false | 3 | [] | [
"DN6"
] | 2023-10-06T11:52:55Z | 2023-10-11T17:10:16Z | 2023-10-11T17:09:55Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nullarch | 8,098,724 | MDQ6VXNlcjgwOTg3MjQ= | User | false |
huggingface/diffusers | 1,930,118,235 | I_kwDOHa8MBc5zC0Rb | 5,313 | https://github.com/huggingface/diffusers/issues/5313 | https://api.github.com/repos/huggingface/diffusers/issues/5313 | LoRA adapters: non revertible fuse | ### Describe the bug
When we load certain adapters, a black image is generated.
During inference the following noticeable error occurs:
```
/lib/python3.10/site-packages/diffusers/image_processor.py:88: RuntimeWarning: invalid value encountered in cast
images = (images * 255).round().astype("uint8")
```
... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-10-06T12:50:43Z | 2023-10-09T16:01:57Z | 2023-10-09T16:01:57Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | oOraph | 13,552,058 | MDQ6VXNlcjEzNTUyMDU4 | User | false |
huggingface/diffusers | 1,930,209,940 | I_kwDOHa8MBc5zDKqU | 5,314 | https://github.com/huggingface/diffusers/issues/5314 | https://api.github.com/repos/huggingface/diffusers/issues/5314 | ControlNet Reference Inpainting Pipeline | ### Model/Pipeline/Scheduler description
Can we implement a ControlNet reference Inpainting Pipeline? Currently there is a community pipeline for ControlNet reference but there is no ControlNet reference Inpainting Pipeline.
### Open source status
- [ ] The model implementation is available
- [ ] The model weights a... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-06T13:41:03Z | 2023-11-16T15:05:30Z | 2023-11-16T15:05:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ajkrish95 | 5,734,592 | MDQ6VXNlcjU3MzQ1OTI= | User | false |
huggingface/diffusers | 1,930,291,166 | I_kwDOHa8MBc5zDefe | 5,315 | https://github.com/huggingface/diffusers/issues/5315 | https://api.github.com/repos/huggingface/diffusers/issues/5315 | Bug in DDIMInverseScheduler function step | ### Describe the bug
The `step` function call in `DDIMInverseScheduler` is causing an array out-of-bounds bug. The error is located here.https://github.com/huggingface/diffusers/blob/872ae1dd12b0f7683834fdf9037810f70a661901/src/diffusers/schedulers/scheduling_ddim_inverse.py#L343 The next time step(`prev_timestep`) ... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-10-06T14:26:41Z | 2023-10-11T13:10:04Z | 2023-10-11T13:10:03Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Rashfu | 42,403,175 | MDQ6VXNlcjQyNDAzMTc1 | User | false |
huggingface/diffusers | 1,930,321,870 | I_kwDOHa8MBc5zDl_O | 5,317 | https://github.com/huggingface/diffusers/issues/5317 | https://api.github.com/repos/huggingface/diffusers/issues/5317 | [`PEFT`] Documenting PEFT API | # Feature request
Let's properly document PEFT API !
PEFT integration (LoRA, AdaLora, IA3, ..) is coming soon in diffusers in the next releases, allowing many cool applications - let's properly document them to educate users on what is possible to achieve with PEFT backend
Related: https://github.com/huggingfa... | closed | completed | false | 8 | [
"stale"
] | [
"stevhliu"
] | 2023-10-06T14:41:36Z | 2023-12-04T15:52:38Z | 2023-12-04T15:52:38Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | younesbelkada | 49,240,599 | MDQ6VXNlcjQ5MjQwNTk5 | User | false |
huggingface/diffusers | 1,930,325,331 | I_kwDOHa8MBc5zDm1T | 5,318 | https://github.com/huggingface/diffusers/issues/5318 | https://api.github.com/repos/huggingface/diffusers/issues/5318 | [`PEFT`] Tutorials with PEFT | # Feature request
PEFT is going to be used soon as backend for adapters in diffusers - let's make some nice tutorials that leverage PEFT in diffusers!
Related: https://github.com/huggingface/diffusers/pull/5151
| closed | completed | false | 4 | [
"stale"
] | [
"sayakpaul"
] | 2023-10-06T14:43:26Z | 2023-11-08T15:30:57Z | 2023-11-08T15:30:56Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | younesbelkada | 49,240,599 | MDQ6VXNlcjQ5MjQwNTk5 | User | false |
huggingface/diffusers | 1,930,328,884 | I_kwDOHa8MBc5zDns0 | 5,319 | https://github.com/huggingface/diffusers/issues/5319 | https://api.github.com/repos/huggingface/diffusers/issues/5319 | [`PEFT`] Speeding up lora loading time with PEFT | # Feature request
Currently if one uses PEFT backend the LoRA loading time is slower than the previous backend
```python
from diffusers import DiffusionPipeline
from safetensors.torch import load_file
import torch
import time
from huggingface_hub import hf_hub_download
pipe = DiffusionPipeline.from_pret... | closed | completed | false | 7 | [] | [] | 2023-10-06T14:44:51Z | 2023-11-13T14:34:41Z | 2023-11-13T14:34:40Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | younesbelkada | 49,240,599 | MDQ6VXNlcjQ5MjQwNTk5 | User | false |
huggingface/diffusers | 1,930,333,153 | I_kwDOHa8MBc5zDovh | 5,320 | https://github.com/huggingface/diffusers/issues/5320 | https://api.github.com/repos/huggingface/diffusers/issues/5320 | [`PEFT`] Add `adapter_name` to `fuse_lora` | # Feature request
Diffusers will be soon using PEFT as backend for adapters such as LoRA - we should make sure users can fuse a specific adapter in the base model. Hence, we should add an argument `adapter_name` in `fuse_lora` and make sure to fuse only that adapter in the base model
Related: https://github.com/h... | closed | completed | false | 5 | [] | [
"sayakpaul"
] | 2023-10-06T14:46:26Z | 2023-12-26T15:54:48Z | 2023-12-26T15:54:48Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | younesbelkada | 49,240,599 | MDQ6VXNlcjQ5MjQwNTk5 | User | false |
huggingface/diffusers | 1,930,447,746 | I_kwDOHa8MBc5zEEuC | 5,321 | https://github.com/huggingface/diffusers/issues/5321 | https://api.github.com/repos/huggingface/diffusers/issues/5321 | SemanticMasks to Image - Diffusion Model | Hi Everyone !
**Is your feature request related to a problem? Please describe.**
I've read the documentation and i saw that there is theses task implemented :
Unconditional Image Generation
Text-Guided Image Generation
Text-Guided Image-to-Image Translation
Text-Guided Image-Inpainting
Text-Guided Depth-to-Ima... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-10-06T15:33:18Z | 2023-11-22T15:05:55Z | 2023-11-22T15:05:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FrsECM | 26,071,804 | MDQ6VXNlcjI2MDcxODA0 | User | false |
huggingface/diffusers | 1,931,162,155 | I_kwDOHa8MBc5zGzIr | 5,325 | https://github.com/huggingface/diffusers/issues/5325 | https://api.github.com/repos/huggingface/diffusers/issues/5325 | Packaging Type Information Not Distributed | ### Describe the bug
Diffusers does not appear to adhere to [PEP 561 – Distributing and Packaging Type Information](https://peps.python.org/pep-0561/). As a result, static type checkers, such as mypy or pyright, report incorrect errors.
According to the docs, this can be remedied by
1. Creating a blank `py.typed` ... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-10-07T03:17:28Z | 2023-10-09T15:17:52Z | 2023-10-09T15:17:52Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | byarbrough | 6,315,292 | MDQ6VXNlcjYzMTUyOTI= | User | false |
huggingface/diffusers | 1,931,191,404 | I_kwDOHa8MBc5zG6Rs | 5,327 | https://github.com/huggingface/diffusers/issues/5327 | https://api.github.com/repos/huggingface/diffusers/issues/5327 | ERROR LOADING TEXTUAL INVERSION IN SDXL | ### Describe the bug
I am trying to load an embedding using `load_textual_inversion` in SDXL but for any embedding I try to load , I get an error , this error is only occuring in SDXL and this error is occurring with every embedding, heres the full code :
```
from diffusers import StableDiffusionPipeline
import tor... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-10-07T05:03:18Z | 2023-10-07T09:19:04Z | 2023-10-07T09:19:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Garry435 | 143,331,750 | U_kgDOCIsRpg | User | false |
huggingface/diffusers | 1,931,228,202 | I_kwDOHa8MBc5zHDQq | 5,328 | https://github.com/huggingface/diffusers/issues/5328 | https://api.github.com/repos/huggingface/diffusers/issues/5328 | LoRA Civital all failing but previously okay | ### Describe the bug
Here is the error message:
File "/home/linnan/codes/lora_test.py", line 7, in <module>
pipe.load_lora_weights("/home/linnan/LORAS/dressed-animals/pytorch_lora_weights.safetensors")
File "/home/linnan/envs/sd/lib/python3.10/site-packages/diffusers/loaders.py", line 2787, in load_lora_w... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-10-07T07:04:08Z | 2023-10-09T15:15:25Z | 2023-10-09T15:15:25Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | linnanwang | 4,675,611 | MDQ6VXNlcjQ2NzU2MTE= | User | false |
huggingface/diffusers | 1,931,426,387 | I_kwDOHa8MBc5zHzpT | 5,330 | https://github.com/huggingface/diffusers/issues/5330 | https://api.github.com/repos/huggingface/diffusers/issues/5330 | UNet2DModel ignores class_labels parameter in forward method | ### Describe the bug
Hello I was trying to add condition to a pre-trained UNet2DModel and re-train it with condition. I was thinking I could do that just by passing condition with class_labels parameter. After sometime I realized that if unet model is not initialized with class_embed_type the forward method completely... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-10-07T17:06:47Z | 2023-11-09T15:20:31Z | 2023-11-09T15:20:31Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kesimeg | 48,391,912 | MDQ6VXNlcjQ4MzkxOTEy | User | false |
huggingface/diffusers | 1,931,794,957 | I_kwDOHa8MBc5zJNoN | 5,333 | https://github.com/huggingface/diffusers/issues/5333 | https://api.github.com/repos/huggingface/diffusers/issues/5333 | StableDiffusionControlNetInpaintImg2ImgPipeline and StableDiffusionControlNetInpaintPipeline do not have the attribute -- from_single_file | when I use the community pipeline StableDiffusionControlNetInpaintImg2ImgPipeline and StableDiffusionControlNetInpaintPipeline, they both produce the error do not have the attribute -- from_single_file, I saw the source code, they both inherit from DiffusionPipeline which does not have the FromSingleFileMixin ... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-08T12:53:19Z | 2023-11-16T15:05:26Z | 2023-11-16T15:05:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zcfrank1st | 1,767,976 | MDQ6VXNlcjE3Njc5NzY= | User | false |
huggingface/diffusers | 1,931,866,157 | I_kwDOHa8MBc5zJfAt | 5,334 | https://github.com/huggingface/diffusers/issues/5334 | https://api.github.com/repos/huggingface/diffusers/issues/5334 | Accelerate.prepare() and Accelerate.accumulate() need multiple models passed in during training | ### Describe the bug
As indicated [here](https://github.com/huggingface/accelerate/pull/1708) we need to pass all of the models we want to apply gradient accumulation steps to into the `accumulate()` method
### Reproduction
Run the `train_dreambooth.py` example script with `--gradient_accumulation_steps` > 1 and mul... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-10-08T15:34:10Z | 2023-11-16T15:05:24Z | 2023-11-16T15:05:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 1,932,181,897 | I_kwDOHa8MBc5zKsGJ | 5,335 | https://github.com/huggingface/diffusers/issues/5335 | https://api.github.com/repos/huggingface/diffusers/issues/5335 | how to deploy locally as chinese gov has block huggingface? | ### Describe the bug
got all the models ckpt safetensor, it still try to connect the /CompVis/stable-diffusion/main/configs/stable-diffusion/v1-infer
### Reproduction
pipe = diffusers.StableDiffusionPipeline.from_single_file(base_model,
torch_dtype=torch... | closed | completed | false | 12 | [
"bug",
"stale"
] | [] | 2023-10-09T01:55:44Z | 2024-01-17T10:44:31Z | 2023-11-16T15:05:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Louis24 | 18,719,360 | MDQ6VXNlcjE4NzE5MzYw | User | false |
huggingface/diffusers | 1,932,203,194 | I_kwDOHa8MBc5zKxS6 | 5,336 | https://github.com/huggingface/diffusers/issues/5336 | https://api.github.com/repos/huggingface/diffusers/issues/5336 | Need Parameter `strength` in `StableDiffusionControlNetImg2ImgPipeline` | **Is your feature request related to a problem? Please describe.**
The parameter `strength` is only in `StableDiffusionControlNetInpaintPipeline` now which is also useful in `StableDiffusionControlNetImg2ImgPipeline` .
**Describe the solution you'd like**
I wonder if someone can add this feature which is already ... | closed | completed | false | 1 | [] | [] | 2023-10-09T02:29:36Z | 2023-10-09T02:34:24Z | 2023-10-09T02:34:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | youyuge34 | 15,784,256 | MDQ6VXNlcjE1Nzg0MjU2 | User | false |
huggingface/diffusers | 1,932,358,894 | I_kwDOHa8MBc5zLXTu | 5,337 | https://github.com/huggingface/diffusers/issues/5337 | https://api.github.com/repos/huggingface/diffusers/issues/5337 | What is the function of `callback` in stable diffusion? | I am reading the source code for stable diffusion pipeline. I wonder what is the function of `callback`? How to use it? Is there an example?
https://github.com/huggingface/diffusers/blob/29f15673ed5c14e4843d7c837890910207f72129/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py#L585C13-L585C21 | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-09T06:02:13Z | 2023-11-16T15:05:20Z | 2023-11-16T15:05:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | g-jing | 44,223,191 | MDQ6VXNlcjQ0MjIzMTkx | User | false |
huggingface/diffusers | 1,932,529,836 | I_kwDOHa8MBc5zMBCs | 5,338 | https://github.com/huggingface/diffusers/issues/5338 | https://api.github.com/repos/huggingface/diffusers/issues/5338 | Bug on checkpoints_total_limit when runing mutiple GPUs | Here is some bug issue in examples/unconditional_image_generation/train_unconditional.py. When running this file on mutiple NPUs or GPUs with args checkpoints_total_limit, it will save the wrong number of checkpoints.
For example, here is my code
```
examples/unconditional_image_generation/train_unconditional.... | closed | completed | false | 2 | [] | [] | 2023-10-09T08:13:51Z | 2023-10-10T10:09:24Z | 2023-10-10T10:09:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiaqiw09 | 60,021,713 | MDQ6VXNlcjYwMDIxNzEz | User | false |
huggingface/diffusers | 1,932,600,960 | I_kwDOHa8MBc5zMSaA | 5,339 | https://github.com/huggingface/diffusers/issues/5339 | https://api.github.com/repos/huggingface/diffusers/issues/5339 | why can not load local safetensor 🐞🐞🐞??? | ### Describe the bug
Same Code + Same Env,
works fine last week
Now: requests.exceptions.ConnectionError: HTTPSConnectionPool(host='raw.githubusercontent.com', port=443): Max retries exceeded with url: /CompVis/stable-diffusion/main/configs/stable-diffusion/v1-inference.yaml (Caused by NewConnectionError('<urllib3.... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-10-09T08:59:27Z | 2024-01-16T08:52:00Z | 2023-11-16T15:05:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Louis24 | 18,719,360 | MDQ6VXNlcjE4NzE5MzYw | User | false |
huggingface/diffusers | 1,932,815,427 | I_kwDOHa8MBc5zNGxD | 5,341 | https://github.com/huggingface/diffusers/issues/5341 | https://api.github.com/repos/huggingface/diffusers/issues/5341 | [`PEFT`] Update the example scripts to reflect PEFT API | PEFT will be soon used as a backend for LoRA related logic !
Let's make sure to reflect the new API in the example scripts, e.g. change the monkey patch calls to `add_adapter`
Related: https://github.com/huggingface/diffusers/pull/5151
| closed | completed | false | 1 | [
"stale"
] | [
"sayakpaul",
"younesbelkada"
] | 2023-10-09T11:02:40Z | 2023-12-07T04:09:31Z | 2023-12-07T04:09:30Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | younesbelkada | 49,240,599 | MDQ6VXNlcjQ5MjQwNTk5 | User | false |
huggingface/diffusers | 1,932,818,303 | I_kwDOHa8MBc5zNHd_ | 5,342 | https://github.com/huggingface/diffusers/issues/5342 | https://api.github.com/repos/huggingface/diffusers/issues/5342 | Non-consecutive added token '!' found. Should have index 49408 but has index 0 in saved vocabulary. | ### Describe the bug
`Non-consecutive added token '!' found. Should have index 49408 but has index 0 in saved vocabulary.`
### Reproduction
https://huggingface.co/haarchri/dogbooth?text=a+photo+of+%5Bv%5Ddog+on+the+beach
result is:
Non-consecutive added token '!' found. Should have index 49408 but has index 0 in... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-10-09T11:04:25Z | 2023-11-16T15:05:15Z | 2023-11-16T15:05:15Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | haarchri | 21,083,497 | MDQ6VXNlcjIxMDgzNDk3 | User | false |
huggingface/diffusers | 1,933,047,497 | I_kwDOHa8MBc5zN_bJ | 5,344 | https://github.com/huggingface/diffusers/issues/5344 | https://api.github.com/repos/huggingface/diffusers/issues/5344 | Requires the omegaconf library but it was not found in your environment. | ### Describe the bug
i've already installed omegaconf, but this error still exists.
### Reproduction
[Untitled.ipynb.zip](https://github.com/huggingface/diffusers/files/12846537/Untitled.ipynb.zip)
### Logs
```shell
---------------------------------------------------------------------------
ValueError ... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-10-09T13:16:29Z | 2023-10-10T15:31:10Z | 2023-10-10T15:31:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | crapthings | 1,147,704 | MDQ6VXNlcjExNDc3MDQ= | User | false |
huggingface/diffusers | 1,933,405,356 | I_kwDOHa8MBc5zPWys | 5,346 | https://github.com/huggingface/diffusers/issues/5346 | https://api.github.com/repos/huggingface/diffusers/issues/5346 | Unittest Discovery in VScode | ### Describe the bug
In vscode, the test are not correctly discovered by default, there is an error due to relative imports in some tests.
It seems that the issue is linked with this line :
```python
from ..o... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-10-09T16:14:34Z | 2023-11-20T15:05:32Z | 2023-11-20T15:05:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FrsECM | 26,071,804 | MDQ6VXNlcjI2MDcxODA0 | User | false |
huggingface/diffusers | 1,933,463,936 | I_kwDOHa8MBc5zPlGA | 5,348 | https://github.com/huggingface/diffusers/issues/5348 | https://api.github.com/repos/huggingface/diffusers/issues/5348 | Implementing EDICT for memory efficient backpropagation through chain/time/steps | Currently backpropagation through chain is impossible in diffusers, due to obvious memory limitations.
However recently through [invertible networks](https://openreview.net/pdf?id=HJsjkMb0Z) and [MemCNN](https://memcnn.readthedocs.io/en/latest/index.html) a recent technique known as EDICT [[paper](https://arxiv.org/ab... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-09T16:56:35Z | 2023-11-20T15:05:30Z | 2023-11-20T15:05:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ysig | 28,439,529 | MDQ6VXNlcjI4NDM5NTI5 | User | false |
huggingface/diffusers | 1,934,069,983 | I_kwDOHa8MBc5zR5Df | 5,350 | https://github.com/huggingface/diffusers/issues/5350 | https://api.github.com/repos/huggingface/diffusers/issues/5350 | Is xformer necessary for SDXL? | I wonder if we have to use xformer so to use SDXL? I am trying to get the attention map of the cross attention and self attention inside of the Unet, but it is hard to get that with xformer since it directly get the result from q, k, and v instead of getting attention map from q and k first. | closed | completed | false | 3 | [] | [] | 2023-10-10T00:50:09Z | 2023-10-13T12:09:10Z | 2023-10-13T12:09:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | g-jing | 44,223,191 | MDQ6VXNlcjQ0MjIzMTkx | User | false |
huggingface/diffusers | 1,934,254,531 | I_kwDOHa8MBc5zSmHD | 5,351 | https://github.com/huggingface/diffusers/issues/5351 | https://api.github.com/repos/huggingface/diffusers/issues/5351 | ZoeDetector Error(s) in loading state_dict for ZoeDepthNK |
zoe_depth = ZoeDetector.from_pretrained(
"valhalla/t2iadapter-aux-models", filename="zoed_nk.pth", model_type="zoedepth_nk"
).to("cuda")
RuntimeError: Error(s) in loading state_dict for ZoeDepthNK:
Unexpected key(s) in state_dict: "core.core.pretrained.model.blocks.0.attn.relative_position_index",... | closed | completed | false | 2 | [] | [] | 2023-10-10T03:26:24Z | 2023-11-01T18:17:01Z | 2023-11-01T18:17:01Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wangyong860401 | 756,537 | MDQ6VXNlcjc1NjUzNw== | User | false |
huggingface/diffusers | 1,934,279,342 | I_kwDOHa8MBc5zSsKu | 5,352 | https://github.com/huggingface/diffusers/issues/5352 | https://api.github.com/repos/huggingface/diffusers/issues/5352 | Use of LoRa sliders | **Is your feature request related to a problem? Please describe.**
In civitai, a lot of LoRa "sliders" are popping up like this one: [hair-length-slider-lora](https://civitai.com/models/114215?modelVersionId=123434). It would allow one to play around with different weights values to apply the LoRa effect. In A1111, in... | closed | completed | false | 12 | [] | [] | 2023-10-10T03:49:25Z | 2023-10-13T13:46:56Z | 2023-10-13T13:39:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jonarod | 6,122,703 | MDQ6VXNlcjYxMjI3MDM= | User | false |
huggingface/diffusers | 1,934,855,654 | I_kwDOHa8MBc5zU43m | 5,353 | https://github.com/huggingface/diffusers/issues/5353 | https://api.github.com/repos/huggingface/diffusers/issues/5353 | How to use FreeU in SimpleCrossAttnUpBlock2D? | I've tried to change your code in order to maintain SimpleCrossAttnUpBlock2D however it seems that shapes doesn't fit up. How can I do it? Thanks!
```Traceback (most recent call last):
File "/usr/local/lib/python3.9/dist-packages/gradio/routes.py", line 523, in run_predict
output = await app.get_blocks().p... | closed | completed | false | 0 | [] | [] | 2023-10-10T09:13:22Z | 2023-10-11T05:11:38Z | 2023-10-11T05:11:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | americanexplorer13 | 57,260,643 | MDQ6VXNlcjU3MjYwNjQz | User | false |
huggingface/diffusers | 1,934,885,864 | I_kwDOHa8MBc5zVAPo | 5,354 | https://github.com/huggingface/diffusers/issues/5354 | https://api.github.com/repos/huggingface/diffusers/issues/5354 | Report error about "find_unused_parameters=True" running in mutiple GPUs | the error will be reported running below UT cases in mutiple NPUs or GPUs:
- ExamplesTestsAccelerate.test_dreambooth_lora_sdxl_with_text_encoder
- ExamplesTestsAccelerate.test_dreambooth_lora_sdxl_text_encoder_checkpointing_checkpoints_total_limit
- ExamplesTestsAccelerate.test_text_to_image_lora_sdxl_with_text_enco... | closed | completed | false | 1 | [] | [] | 2023-10-10T09:29:27Z | 2023-10-27T17:19:15Z | 2023-10-27T17:19:15Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiaqiw09 | 60,021,713 | MDQ6VXNlcjYwMDIxNzEz | User | false |
huggingface/diffusers | 1,935,319,998 | I_kwDOHa8MBc5zWqO- | 5,356 | https://github.com/huggingface/diffusers/issues/5356 | https://api.github.com/repos/huggingface/diffusers/issues/5356 | [`PEFT`] Document properly advanced features for PEFT integration | Let's properly document advanced features users can play with in a relevant section of the documentation. E.g. `set_lora_device` to set the LoRA weights in a specific device (for example to offload lora weights on CPU)
Related: #5151
cc @sayakpaul | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-10T12:59:36Z | 2023-11-20T15:05:28Z | 2023-11-20T15:05:28Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | younesbelkada | 49,240,599 | MDQ6VXNlcjQ5MjQwNTk5 | User | false |
huggingface/diffusers | 1,935,372,380 | I_kwDOHa8MBc5zW3Bc | 5,357 | https://github.com/huggingface/diffusers/issues/5357 | https://api.github.com/repos/huggingface/diffusers/issues/5357 | question about sdxl+img2img+controlnet pipeline | it seems sdxlimg2imgcontrolnet pipeline does not work very well in some promt style
for example
anime style
prompt：listen, anime style artwork, cartoon, makoto shinkai and lois van baarle, unreal engine, loish, rhads, beeple, ilya kuvshinov, rossdraws, tom bagshaw, alphonse mucha, global illumination, soft colors, ... | closed | completed | false | 7 | [] | [] | 2023-10-10T13:26:07Z | 2023-10-16T09:22:31Z | 2023-10-16T09:21:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | syyxsxx | 32,666,364 | MDQ6VXNlcjMyNjY2MzY0 | User | false |
huggingface/diffusers | 1,935,410,649 | I_kwDOHa8MBc5zXAXZ | 5,358 | https://github.com/huggingface/diffusers/issues/5358 | https://api.github.com/repos/huggingface/diffusers/issues/5358 | Issue with Custom Attention Processors Compatibility | Version of diffusers: **0.21.4**
**Description**
I've encountered difficulties with the compatibility of custom attention processors in the diffusers library, particularly when attempting to use the [IP-Adapter](https://github.com/tencent-ailab/IP-Adapter) with custom attention processors.
**Details of the Issue... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-10T13:45:18Z | 2023-11-20T15:05:26Z | 2023-11-20T15:05:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | denemmy | 5,609,333 | MDQ6VXNlcjU2MDkzMzM= | User | false |
huggingface/diffusers | 1,376,228,928 | I_kwDOHa8MBc5SB5ZA | 536 | https://github.com/huggingface/diffusers/issues/536 | https://api.github.com/repos/huggingface/diffusers/issues/536 | Add `output_attentions` option to `DiffusionPipeline.__call__` | In the same way that `transformers` has this option for generate/forward methods (example in [here](https://github.com/huggingface/transformers/blob/5e636eee4af48ccd03b4d9c1a1e6f7a1b92a643f/src/transformers/generation_utils.py#L927)), we could have the same option in `DiffusionPipeline.__call__`.
I started the imple... | closed | completed | false | 3 | [
"stale"
] | [] | 2022-09-16T17:38:06Z | 2022-10-31T15:03:09Z | 2022-10-31T15:03:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JoaoLages | 17,574,157 | MDQ6VXNlcjE3NTc0MTU3 | User | false |
huggingface/diffusers | 1,935,720,302 | I_kwDOHa8MBc5zYL9u | 5,360 | https://github.com/huggingface/diffusers/issues/5360 | https://api.github.com/repos/huggingface/diffusers/issues/5360 | Provide single file safetensors of hf-internal-testing/tiny-stable-diffusion-torch and SDXL | Such files would be useful for tests.
Is there an option to convert hf-internal-testing/tiny-stable-diffusion-torch to a single safetensors file?
I tried convert_diffusers_to_original_stable_diffusion.py but when loading it produces "NotImplementedError: Cannot copy out of meta tensor; no data!". I think it's due ... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-10-10T16:08:39Z | 2024-03-15T10:22:43Z | 2023-11-21T15:05:22Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | noskill | 733,626 | MDQ6VXNlcjczMzYyNg== | User | false |
huggingface/diffusers | 1,935,913,017 | I_kwDOHa8MBc5zY7A5 | 5,363 | https://github.com/huggingface/diffusers/issues/5363 | https://api.github.com/repos/huggingface/diffusers/issues/5363 | [`PEFT` / `Tests` ] Re-design testing suite for `Unet2DConditionModel` & `Unet3DConditionModel` | In the current PEFT integration the `Unet2DConditionModel` and `Unet3DConditionModel` testers are being silently tested in `PeftLoraLoaderMixinTests`. Since the API changed a bit - i.e. no need to call `set_attn_procs` but call `add_adapter` instead, I propose to refactor the testing suite of the unets and make sure t... | closed | completed | false | 11 | [
"stale"
] | [] | 2023-10-10T17:56:23Z | 2024-02-03T04:18:27Z | 2024-02-03T04:18:26Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | younesbelkada | 49,240,599 | MDQ6VXNlcjQ5MjQwNTk5 | User | false |
huggingface/diffusers | 1,936,288,597 | I_kwDOHa8MBc5zaWtV | 5,364 | https://github.com/huggingface/diffusers/issues/5364 | https://api.github.com/repos/huggingface/diffusers/issues/5364 | Early stopping and validation metrics logging during training T2I | Any plan to add a early stopping approach and loggin validation metrics like FID and IS when training text to image model ? | closed | completed | false | 3 | [
"stale"
] | [] | 2023-10-10T21:25:48Z | 2023-11-21T15:05:19Z | 2023-11-21T15:05:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | segalinc | 9,353,106 | MDQ6VXNlcjkzNTMxMDY= | User | false |
huggingface/diffusers | 1,936,422,317 | I_kwDOHa8MBc5za3Wt | 5,365 | https://github.com/huggingface/diffusers/issues/5365 | https://api.github.com/repos/huggingface/diffusers/issues/5365 | Error when setting `num_inference_steps` in txt2video | ### Describe the bug
If I set `num_inference_steps` to 30 in txt2video, an error occurs, but if I set it to 60, no error occurs.
### Reproduction
import torch
import imageio
from diffusers import TextToVideoZeroPipeline
import numpy as np
model_id = "runwayml/stable-diffusion-v1-5"
pipe = TextToVideoZeroP... | closed | completed | false | 2 | [
"bug"
] | [
"DN6"
] | 2023-10-10T23:01:26Z | 2023-10-12T22:50:47Z | 2023-10-12T22:50:47Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | suzukimain | 131,413,573 | U_kgDOB9U2RQ | User | false |
huggingface/diffusers | 1,938,046,702 | I_kwDOHa8MBc5zhD7u | 5,368 | https://github.com/huggingface/diffusers/issues/5368 | https://api.github.com/repos/huggingface/diffusers/issues/5368 | Using --mixed_precision="fp16" brings ValueError: Query/Key/Value should all have the same dtype | ### Describe the bug
ValueError: Query/Key/Value should all have the same dtype
query.dtype: torch.float32
key.dtype : torch.float16
value.dtype: torch.float16
### Reproduction
Use --mixed_precision="fp16" in Dreambooth script
Without --mixed_precision="fp16" it runs fine.
### Logs
```shell
10/11/202... | closed | completed | false | 16 | [
"bug",
"stale"
] | [] | 2023-10-11T15:08:50Z | 2024-01-09T15:08:17Z | 2024-01-09T15:08:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bluusun | 65,051,271 | MDQ6VXNlcjY1MDUxMjcx | User | false |
huggingface/diffusers | 1,939,102,453 | I_kwDOHa8MBc5zlFr1 | 5,371 | https://github.com/huggingface/diffusers/issues/5371 | https://api.github.com/repos/huggingface/diffusers/issues/5371 | add blip inpainting pipe support | how to use mask inpainting with blip diffusion [blip_diffusion](https://github.com/huggingface/diffusers/tree/main/src/diffusers/pipelines/blip_diffusion) | closed | completed | false | 3 | [
"stale"
] | [] | 2023-10-12T03:19:24Z | 2023-11-21T15:05:16Z | 2023-11-21T15:05:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yanchaoguo | 11,827,802 | MDQ6VXNlcjExODI3ODAy | User | false |
huggingface/diffusers | 1,939,116,832 | I_kwDOHa8MBc5zlJMg | 5,372 | https://github.com/huggingface/diffusers/issues/5372 | https://api.github.com/repos/huggingface/diffusers/issues/5372 | How to use safety_checker in StableDiffusionXLPipeline? | ### Describe the bug
I want to use safety_checker in StableDiffusionXLPipeline, but it seems that `safety_checker` keyword does not take effect
### Reproduction
```python
pipe = StableDiffusionXLPipeline.from_pretrained(
"nyxia/mysterious-xl",
torch_dtype=torch.float16,
safety_checker = StableD... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-10-12T03:39:23Z | 2023-10-12T08:13:28Z | 2023-10-12T08:13:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hundredwz | 10,173,356 | MDQ6VXNlcjEwMTczMzU2 | User | false |
huggingface/diffusers | 1,939,175,758 | I_kwDOHa8MBc5zlXlO | 5,373 | https://github.com/huggingface/diffusers/issues/5373 | https://api.github.com/repos/huggingface/diffusers/issues/5373 | Callback for `WuerstchenCombinedPipeline` | A callback similar to existing pipelines such as [`diffusers.StableDiffusionPipeline`](https://huggingface.co/docs/diffusers/v0.21.0/en/api/pipelines/stable_diffusion/text2img#diffusers.StableDiffusionPipeline) and [`diffusers.KandinskyCombinedPipeline`](https://huggingface.co/docs/diffusers/v0.21.0/en/api/pipelines/ka... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-10-12T04:56:42Z | 2023-11-21T15:05:14Z | 2023-11-21T15:05:14Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | soumik12345 | 19,887,676 | MDQ6VXNlcjE5ODg3Njc2 | User | false |
huggingface/diffusers | 1,939,523,298 | I_kwDOHa8MBc5zmsbi | 5,374 | https://github.com/huggingface/diffusers/issues/5374 | https://api.github.com/repos/huggingface/diffusers/issues/5374 | Getting error got an unexpected keyword argument 'checkpoint_path' | ### Describe the bug
I been trying to convert .safetenor to .diffusers in Colab. After running .py file getting error as
_got an unexpected keyword argument 'checkpoint_path'_
### Reproduction
Traceback (most recent call last):
File "/content/diffusers/scripts/convert_original_stable_diffusion_to_diffuser... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-10-12T08:53:06Z | 2023-11-21T15:05:12Z | 2023-11-21T15:05:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LemonsRobert | 141,733,591 | U_kgDOCHKu1w | User | false |
huggingface/diffusers | 1,939,630,160 | I_kwDOHa8MBc5znGhQ | 5,375 | https://github.com/huggingface/diffusers/issues/5375 | https://api.github.com/repos/huggingface/diffusers/issues/5375 | StableDiffusionXLControlNetPipeline does not have the from_single_file attribute | When i use the StableDiffusionXLControlNetPipeline and call the "from_single_file" method it returns an error that the attribute does not exist, but in the docs it states that the pipeline inherits the method from the DiffusionPipeline. Would like to use this attribute to deploy models with controlnet using only the sa... | closed | completed | false | 5 | [] | [] | 2023-10-12T09:52:29Z | 2023-10-19T04:49:08Z | 2023-10-19T04:49:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rafationgson | 70,418,702 | MDQ6VXNlcjcwNDE4NzAy | User | false |
huggingface/diffusers | 1,940,389,684 | I_kwDOHa8MBc5zp_80 | 5,377 | https://github.com/huggingface/diffusers/issues/5377 | https://api.github.com/repos/huggingface/diffusers/issues/5377 | Make T2I-Adapter downscale padding match the UNet | # Background
T2I-Adapter models produce residual maps that must match the corresponding UNet hidden state dimensions at each downscaling level. This works as expected when the T2I-Adapter's input image is evenly divisible by the T2I-Adapter's `total_downscale_factor`.
For example, a `FullAdapter` used with an SD1... | closed | completed | false | 7 | [] | [] | 2023-10-12T16:45:44Z | 2023-10-23T16:52:33Z | 2023-10-23T16:52:33Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | RyanJDick | 14,897,797 | MDQ6VXNlcjE0ODk3Nzk3 | User | false |
huggingface/diffusers | 1,940,597,927 | I_kwDOHa8MBc5zqyyn | 5,379 | https://github.com/huggingface/diffusers/issues/5379 | https://api.github.com/repos/huggingface/diffusers/issues/5379 | [community pipeline] SpeechToImage | Hi @MikailINTech, the copy-and-paste example from the [Speech to image](https://github.com/huggingface/diffusers/tree/main/examples/community#speech-to-image) pipeline is failing to initialize correctly with the following error:
```py
---------------------------------------------------------------------------
Valu... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-10-12T18:53:38Z | 2024-08-27T15:52:00Z | 2023-11-21T15:05:10Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | stevhliu | 59,462,357 | MDQ6VXNlcjU5NDYyMzU3 | User | false |
huggingface/diffusers | 1,940,610,216 | I_kwDOHa8MBc5zq1yo | 5,380 | https://github.com/huggingface/diffusers/issues/5380 | https://api.github.com/repos/huggingface/diffusers/issues/5380 | [community pipeline] Mixture tiling | Hey @kadirnar, the [Mixture tiling](https://github.com/huggingface/diffusers/tree/main/examples/community#stable-diffusion-mixture-tiling) pipeline is failing because it can't find the module `diffusers.pipeline_utils`.
```py
---------------------------------------------------------------------------
ModuleNotFoun... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-12T19:01:27Z | 2023-11-16T19:46:40Z | 2023-11-16T19:46:39Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | stevhliu | 59,462,357 | MDQ6VXNlcjU5NDYyMzU3 | User | false |
huggingface/diffusers | 1,941,491,771 | I_kwDOHa8MBc5zuNA7 | 5,383 | https://github.com/huggingface/diffusers/issues/5383 | https://api.github.com/repos/huggingface/diffusers/issues/5383 | Use `HfApi` instead of `Repository` in training scripts | `Repository` [is being deprecated](https://github.com/huggingface/huggingface_hub/pull/1724), but we still use it in a couple of training scripts:
- https://github.com/huggingface/diffusers/blob/4d2c981d551566961bbc7254ae9556d76dd95764/examples/dreambooth/train_dreambooth_flax.py#L19
- https://github.com/huggingfac... | closed | completed | false | 2 | [] | [] | 2023-10-13T08:26:51Z | 2023-10-16T14:04:48Z | 2023-10-16T14:04:48Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pcuenca | 1,177,582 | MDQ6VXNlcjExNzc1ODI= | User | false |
huggingface/diffusers | 1,942,150,700 | I_kwDOHa8MBc5zwt4s | 5,387 | https://github.com/huggingface/diffusers/issues/5387 | https://api.github.com/repos/huggingface/diffusers/issues/5387 | Broken Link in Textual Inversion Documentation | Issue Description:
The link to the [Textual Inversion Training Guide](https://[huggingface.co/docs/diffusers/using-diffusers/training/text_inversion](https://huggingface.co/docs/diffusers/using-diffusers/training/text_inversion)) is currently broken on [this](https://huggingface.co/docs/diffusers/using-diffusers/textu... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-10-13T15:16:06Z | 2023-11-21T15:05:06Z | 2023-11-21T15:05:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mhetrerajat | 11,217,092 | MDQ6VXNlcjExMjE3MDky | User | false |
huggingface/diffusers | 1,376,287,615 | I_kwDOHa8MBc5SCHt_ | 539 | https://github.com/huggingface/diffusers/issues/539 | https://api.github.com/repos/huggingface/diffusers/issues/539 | TypeError: snapshot_download() got an unexpected keyword argument 'allow_patterns' | ### Describe the bug
TypeError: snapshot_download() got an unexpected keyword argument 'allow_patterns'
### Reproduction
_No response_
### Logs
_No response_
### System Info
- `diffusers` version: 0.4.0.dev0
- Platform: Linux-5.3.0-64-generic-x86_64-with-glibc2.17
- Python version: 3.8.12
- PyTorch version (G... | closed | completed | false | 1 | [
"bug"
] | [] | 2022-09-16T18:11:25Z | 2022-09-16T18:12:32Z | 2022-09-16T18:12:32Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,942,262,781 | I_kwDOHa8MBc5zxJP9 | 5,390 | https://github.com/huggingface/diffusers/issues/5390 | https://api.github.com/repos/huggingface/diffusers/issues/5390 | SDXL pipeline downloads text encoder even when local_files_only=True | ### Describe the bug
Using diffusers in a docker container, I notice when loading a safetensors SDXL model via `StableDiffusionXLPipeline.from_single_file` triggers the downloading of a text encoder, even when local_files_only=True.
How do I identify what is being downloaded and what is causing a cache miss?
#... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-10-13T16:35:16Z | 2023-10-18T12:03:14Z | 2023-10-18T12:03:14Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | levi | 8,493 | MDQ6VXNlcjg0OTM= | User | false |
huggingface/diffusers | 1,942,926,372 | I_kwDOHa8MBc5zzrQk | 5,392 | https://github.com/huggingface/diffusers/issues/5392 | https://api.github.com/repos/huggingface/diffusers/issues/5392 | How to train an unconditional latent diffusion model ? | It seems that there is only one available unconditional LDM model (CompVis/ldm-celebahq-256).
```python
pipeline = LDMPipeline.from_pretrained("CompVis/ldm-celebahq-256")
```
How can I train this unconditional model on my own dataset? The LDM model includes the training of both `VQModel` and `UNet2DModel`, but the... | closed | completed | false | 2 | [] | [] | 2023-10-14T03:32:34Z | 2024-02-16T08:59:49Z | 2023-10-17T01:54:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Rashfu | 42,403,175 | MDQ6VXNlcjQyNDAzMTc1 | User | false |
huggingface/diffusers | 1,943,073,361 | I_kwDOHa8MBc5z0PJR | 5,394 | https://github.com/huggingface/diffusers/issues/5394 | https://api.github.com/repos/huggingface/diffusers/issues/5394 | "`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly | Getting this error when using the following prompt and negative prompt combination:
RAW photo, a photo of nayansks, (wearing beige blazer and shirt:1.2), 8k uhd, high quality, Fujifilm XT3, medium close up shot, (Soft lighting, Studio backdrop, Headshot photography, Even lighting, Portrait lighting, Professional heads... | closed | completed | false | 2 | [] | [] | 2023-10-14T08:13:32Z | 2023-10-17T07:46:22Z | 2023-10-17T07:46:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nayan-dhabarde | 24,761,808 | MDQ6VXNlcjI0NzYxODA4 | User | false |
huggingface/diffusers | 1,943,217,383 | I_kwDOHa8MBc5z0yTn | 5,395 | https://github.com/huggingface/diffusers/issues/5395 | https://api.github.com/repos/huggingface/diffusers/issues/5395 | [PEFT] Bug combining some loras from civitai using peft | ### Describe the bug
Inference bug when combining some LORAs from Civitai using PEFT.
Google colab with reproduction: https://colab.research.google.com/drive/1bpVehy9NQDEIRt5JeYfKGFlw7KZdlDrm?usp=sharing
in peft/tuners/lora/layer.py
(self.active_adapters contain adapters not present in self.lora_alpha)
##... | closed | completed | false | 5 | [
"bug",
"stale"
] | [
"sayakpaul"
] | 2023-10-14T11:48:25Z | 2023-11-13T15:13:48Z | 2023-11-13T15:13:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jaimereyley | 27,781,671 | MDQ6VXNlcjI3NzgxNjcx | User | false |
huggingface/diffusers | 1,943,695,257 | I_kwDOHa8MBc5z2m-Z | 5,398 | https://github.com/huggingface/diffusers/issues/5398 | https://api.github.com/repos/huggingface/diffusers/issues/5398 | [Hacktoberfest] Deepspeed for dreambooth | **Is your feature request related to a problem? Please describe.**
I plan to add deepspeed to dreambooth following the same design as train_text_to_image.py. Also might copy open-muse for the config.
**Describe the solution you'd like**
I'll make a pr and test if it saves vram.
| closed | completed | false | 2 | [] | [] | 2023-10-15T04:05:47Z | 2023-10-17T20:06:52Z | 2023-10-17T20:06:52Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | isamu-isozaki | 23,430,101 | MDQ6VXNlcjIzNDMwMTAx | User | false |
huggingface/diffusers | 1,943,885,185 | I_kwDOHa8MBc5z3VWB | 5,399 | https://github.com/huggingface/diffusers/issues/5399 | https://api.github.com/repos/huggingface/diffusers/issues/5399 | Inference with custom inpainting model triggers: RuntimeError: mat1 and mat2 must have the same dtype | ### Describe the bug
I have converted [DreamShaper v8 Inpainting](https://civitai.com/models/4384?modelVersionId=131004) to diffusers format using [OneTrainer](https://github.com/Nerogar/OneTrainer) and everything went smoothly, but when I call `StableDiffusionInpaintPipeline` I get the error message: **RuntimeError... | closed | completed | false | 13 | [
"bug",
"stale"
] | [] | 2023-10-15T13:15:40Z | 2024-01-18T15:06:19Z | 2024-01-18T15:06:19Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexisrolland | 13,064,696 | MDQ6VXNlcjEzMDY0Njk2 | User | false |
huggingface/diffusers | 1,376,290,373 | I_kwDOHa8MBc5SCIZF | 540 | https://github.com/huggingface/diffusers/issues/540 | https://api.github.com/repos/huggingface/diffusers/issues/540 | Reduce Stable Diffusion memory usage by keeping unet only on GPU. | **Is your feature request related to a problem? Please describe.**
Stable Diffusion is not compute heavy on all its steps. If we keep the diffusion unet on fp16 on GPU and everything else on CPU, we could reduce the GPU usage to 2.2GB while having a non-so-big impact on performance. It should democratize Stable Diffus... | closed | completed | false | 7 | [] | [] | 2022-09-16T18:12:59Z | 2022-10-27T18:25:11Z | 2022-10-27T18:25:11Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | piEsposito | 47,679,710 | MDQ6VXNlcjQ3Njc5NzEw | User | false |
huggingface/diffusers | 1,944,265,334 | I_kwDOHa8MBc5z4yJ2 | 5,402 | https://github.com/huggingface/diffusers/issues/5402 | https://api.github.com/repos/huggingface/diffusers/issues/5402 | cannot import name 'compute_snr' from 'diffusers.training_utils' | I'm trying to train using the script train_text_to_image_lora.py and I am getting the following error:
```
Traceback (most recent call last):
File "G:\Project Files\Kohya_SS - Data\Text to Image LoRA\train_text_to_image_lora.py", line 46, in <module>
from diffusers.training_utils import compute_snr
ImportEr... | closed | completed | false | 6 | [] | [
"sayakpaul"
] | 2023-10-16T03:10:40Z | 2023-10-18T03:29:59Z | 2023-10-18T03:29:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | RhodianSK | 99,561,171 | U_kgDOBe8u0w | User | false |
huggingface/diffusers | 1,944,279,360 | I_kwDOHa8MBc5z41lA | 5,403 | https://github.com/huggingface/diffusers/issues/5403 | https://api.github.com/repos/huggingface/diffusers/issues/5403 | Training script for T2I adapter bug? | Hi,
The training script for t2i adapter also sets the unet for train:
https://github.com/huggingface/diffusers/blob/07b297e7dec3f475a880c15addcb6c02c0690992/examples/t2i_adapter/train_t2i_adapter_sdxl.py#L900C5-L900C9
Is this by design or a bug? My understanding is that only adapter is trained and unet does not ... | closed | completed | false | 4 | [
"stale"
] | [
"sayakpaul"
] | 2023-10-16T03:26:49Z | 2023-11-27T02:47:51Z | 2023-11-27T02:47:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | darshats | 56,063,876 | MDQ6VXNlcjU2MDYzODc2 | User | false |
huggingface/diffusers | 1,944,330,647 | I_kwDOHa8MBc5z5CGX | 5,404 | https://github.com/huggingface/diffusers/issues/5404 | https://api.github.com/repos/huggingface/diffusers/issues/5404 | [Feature Request]: Pipeline for Expressive Text-to-Image Generation with Rich Text | ### Is your feature request related to a problem? Please describe.
Event though plain text is currently the most dominant interface for text-to-image synthesis, its limited customization options
hinder users from accurately describing desired outputs. {lain text makes it hard to specify continuous quantities, such as... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-16T04:28:58Z | 2023-11-24T15:05:22Z | 2023-11-24T15:05:22Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | soumik12345 | 19,887,676 | MDQ6VXNlcjE5ODg3Njc2 | User | false |
huggingface/diffusers | 1,944,397,555 | I_kwDOHa8MBc5z5Sbz | 5,405 | https://github.com/huggingface/diffusers/issues/5405 | https://api.github.com/repos/huggingface/diffusers/issues/5405 | sdxl.from_single_file not overwriting the weights | ### Describe the bug
ckpt1
ckpt2
SDXL.from_single_file(ckpt1)
img1 = SDXL.generate_image()
SDXL.from_single_file(ckpt2)
img2 = SDXL.generate_image()
It seems img1 and img2 are all generate from ckpt1. And the weights from ckpt2 is not reloaded.
Is it designed like this? Thanks.
### Reproduction
s... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-10-16T05:30:36Z | 2023-11-24T15:05:20Z | 2023-11-24T15:05:20Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | linnanwang | 4,675,611 | MDQ6VXNlcjQ2NzU2MTE= | User | false |
huggingface/diffusers | 1,944,516,580 | I_kwDOHa8MBc5z5vfk | 5,406 | https://github.com/huggingface/diffusers/issues/5406 | https://api.github.com/repos/huggingface/diffusers/issues/5406 | Seeking Advice on ControlNet training | Greetings,
I tried to train my own inpaint version of controlnet on COCO datasets several times, but found it was hard to train it well. Basically, I have 330k amplified samples of COCO dataset, each sample has image, mask and caption. I have 8 A100 GPUs and I train controlnets from 30 to 90 epochs (some are way mor... | closed | completed | false | 19 | [
"community-examples",
"stale"
] | [] | 2023-10-16T07:02:39Z | 2024-01-17T02:59:55Z | 2023-11-24T15:05:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lucasgblu | 129,648,867 | U_kgDOB7pI4w | User | false |
huggingface/diffusers | 1,944,564,861 | I_kwDOHa8MBc5z57R9 | 5,408 | https://github.com/huggingface/diffusers/issues/5408 | https://api.github.com/repos/huggingface/diffusers/issues/5408 | bad inference result on diffusers (kohya style lora) | I use v0.21.4 version of diffusers
and I m using this code for inference image with kohya style lora
pipe = StableDiffusionXLPipeline.from_single_file("model.safetensors", torch_dtype=torch.bfloat16)
pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)
pipe.load_lora_weights('woma... | closed | completed | false | 1 | [
"stale"
] | [] | 2023-10-16T07:32:50Z | 2023-11-24T15:05:16Z | 2023-11-24T15:05:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nullarch | 8,098,724 | MDQ6VXNlcjgwOTg3MjQ= | User | false |
huggingface/diffusers | 1,944,744,351 | I_kwDOHa8MBc5z6nGf | 5,409 | https://github.com/huggingface/diffusers/issues/5409 | https://api.github.com/repos/huggingface/diffusers/issues/5409 | Using the SDXL refiner with SD 1.x models | I've been attempting to use the SDXL refiner together with SD 1.x models but so far haven't been able to produce good results. I think this is an idea many people will explore at some point, so I figured it justifies a ticket here. Sorry if this is too off topic.
I have looked into the Diffusers source code for SDXL... | closed | completed | false | 4 | [
"community-examples"
] | [] | 2023-10-16T09:11:38Z | 2023-10-27T14:22:52Z | 2023-10-27T14:22:51Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | holwech | 7,474,295 | MDQ6VXNlcjc0NzQyOTU= | User | false |
huggingface/diffusers | 1,376,352,376 | I_kwDOHa8MBc5SCXh4 | 541 | https://github.com/huggingface/diffusers/issues/541 | https://api.github.com/repos/huggingface/diffusers/issues/541 | Diffusers on SageMaker | **Is your feature request related to a problem? Please describe.**
I'd like to run inference with stable diffusion through an API request!
**Describe the solution you'd like**
Either an Inference API or Sagemaker integration!
**Additional context**
Similar to https://huggingface.co/inference-endpoints
or
ht... | closed | completed | false | 15 | [
"stale"
] | [] | 2022-09-16T18:53:32Z | 2022-12-05T15:03:51Z | 2022-12-05T15:03:51Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AmanKishore | 10,477,380 | MDQ6VXNlcjEwNDc3Mzgw | User | false |
huggingface/diffusers | 1,945,374,165 | I_kwDOHa8MBc5z9A3V | 5,410 | https://github.com/huggingface/diffusers/issues/5410 | https://api.github.com/repos/huggingface/diffusers/issues/5410 | Latent upscale with SDXL gives blurry results | How can one use a latent upscaler with SDXL? I've been playing around with e.g stabilityai/sd-x2-latent-upscaler but nothing I do results in a usable image. Is latent upscaling incompatible with SDXL or is it just a matter of getting the parameters right?
Here's a piece of code I tried to run:
```python
from d... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-10-16T14:40:15Z | 2023-11-24T15:05:14Z | 2023-11-24T15:05:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | agcty | 12,101,091 | MDQ6VXNlcjEyMTAxMDkx | User | false |
huggingface/diffusers | 1,945,603,796 | I_kwDOHa8MBc5z947U | 5,412 | https://github.com/huggingface/diffusers/issues/5412 | https://api.github.com/repos/huggingface/diffusers/issues/5412 | resume_from_checkpoint training fail in train_dreambooth_lora_sdxl.py | ### Describe the bug
I trained dreambooth with lora and sd-xl for 1000 steps, then I try to continue traning resume from the 500th step, however, it seems like the training starts without the 1000's checkpoint, i.e. it starts from the beginning. Training scripts are as below.
btw. I fix another resume bug as advised ... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-10-16T16:07:09Z | 2024-01-15T15:07:59Z | 2024-01-15T15:07:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yuxu915 | 94,051,337 | U_kgDOBZscCQ | User | false |
huggingface/diffusers | 1,945,899,431 | I_kwDOHa8MBc5z_BGn | 5,414 | https://github.com/huggingface/diffusers/issues/5414 | https://api.github.com/repos/huggingface/diffusers/issues/5414 | [`PEFT`] `set_adapters` with `weights` argument currently broken | ### Describe the bug
Currently `set_adapters` with `weights` argument is broken in diffusers. In fact `set_adapters` will correctly call `module.scale_layer` that currently overrides the scale value with the scaled value. But during inference, that information is being lost because we re-compute the original scale her... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-10-16T19:00:40Z | 2023-10-21T16:47:20Z | 2023-10-21T16:47:20Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | younesbelkada | 49,240,599 | MDQ6VXNlcjQ5MjQwNTk5 | User | false |
huggingface/diffusers | 1,946,057,128 | I_kwDOHa8MBc5z_nmo | 5,415 | https://github.com/huggingface/diffusers/issues/5415 | https://api.github.com/repos/huggingface/diffusers/issues/5415 | Allow setting GroupNorm groups in UNet1D classes | **Is your feature request related to a problem? Please describe.**
`ResidualTemporalBlock1D` [currently](https://github.com/huggingface/diffusers/blob/29f15673ed5c14e4843d7c837890910207f72129/src/diffusers/models/resnet.py#L703-L714) does allow setting the number of normalization groups.
This behavior could then be p... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-10-16T20:41:52Z | 2023-11-16T15:18:52Z | 2023-11-16T15:18:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | petrzjunior | 7,000,918 | MDQ6VXNlcjcwMDA5MTg= | User | false |
huggingface/diffusers | 1,946,076,101 | I_kwDOHa8MBc5z_sPF | 5,416 | https://github.com/huggingface/diffusers/issues/5416 | https://api.github.com/repos/huggingface/diffusers/issues/5416 | How to correctly implement a class-conditional model | Hi, I'd like to implement a DDPM that is class-conditioned, but not conditioned on anything else (no text), using `UNet2DConditionModel`. I'm training from scratch.
I'm calling the model with `noise_pred = model(noisy_images, timesteps, class_labels=class_labels, return_dict=False)[0]`, but I get the error `UNet2D... | closed | completed | false | 1 | [] | [] | 2023-10-16T20:53:41Z | 2023-10-16T21:02:39Z | 2023-10-16T21:02:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nickk124 | 37,184,532 | MDQ6VXNlcjM3MTg0NTMy | User | false |
huggingface/diffusers | 1,376,652,561 | I_kwDOHa8MBc5SDg0R | 542 | https://github.com/huggingface/diffusers/issues/542 | https://api.github.com/repos/huggingface/diffusers/issues/542 | unet export to onnx issue | ### Describe the bug
Trying to export the unet model to onnx. Used [this script](https://github.com/huggingface/diffusers/blob/main/scripts/convert_stable_diffusion_checkpoint_to_onnx.py)
`.onnx` file created successfully but export wasn't successful:
```
import onnx
onnx_model = onnx.load("model.onnx")
onnx.... | closed | completed | false | 5 | [
"stale"
] | [] | 2022-09-17T03:38:13Z | 2022-11-02T16:09:36Z | 2022-11-02T16:09:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tanayvarshney | 11,531,975 | MDQ6VXNlcjExNTMxOTc1 | User | false |
huggingface/diffusers | 1,946,852,317 | I_kwDOHa8MBc50Cpvd | 5,423 | https://github.com/huggingface/diffusers/issues/5423 | https://api.github.com/repos/huggingface/diffusers/issues/5423 | Non-consecutive added token '<|startoftext|>' found. Should have index 49408 but has index 49406 in saved vocabulary. | How to solve this??
text to img
Non-consecutive added token '<|startoftext|>' found. Should have index 49408 but has index 49406 in saved vocabulary.
| closed | completed | false | 4 | [
"stale"
] | [
"DN6"
] | 2023-10-17T08:26:28Z | 2023-11-21T03:54:41Z | 2023-11-21T03:54:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Damarcreative | 106,896,172 | U_kgDOBl8bLA | User | false |
huggingface/diffusers | 1,947,095,979 | I_kwDOHa8MBc50DlOr | 5,425 | https://github.com/huggingface/diffusers/issues/5425 | https://api.github.com/repos/huggingface/diffusers/issues/5425 | An Issue may relate to be compatible among accelerate, PT 2.1 and train_unconditional.py while trying to save checkpoints. | ### Describe the bug
An Issue may relate to be compatible among accelerate, PT 2.1 and train_unconditional.py while trying to save checkpoints.
### Reproduction
I am encountering an issue while trying to reimplement the unconditional_image_generation
accelerate launch --mixed_precision="fp16" --multi_gpu train_u... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-10-17T10:32:00Z | 2023-10-18T13:00:58Z | 2023-10-18T13:00:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | linhduongtuan | 22,388,092 | MDQ6VXNlcjIyMzg4MDky | User | false |
huggingface/diffusers | 1,948,246,304 | I_kwDOHa8MBc50H-Eg | 5,429 | https://github.com/huggingface/diffusers/issues/5429 | https://api.github.com/repos/huggingface/diffusers/issues/5429 | [Hacktoberfest] Support models without vae in other examples than dreambooth | **Is your feature request related to a problem? Please describe.**
I noticed that some features in dreambooth seem to not be in other examples. The main one I found was supporting models without vae not being there in textual_inversion and text_to_image example.
**Describe the solution you'd like**
Make a pr for i... | closed | completed | false | 0 | [] | [
"DN6"
] | 2023-10-17T20:55:07Z | 2023-11-06T00:14:56Z | 2023-11-06T00:14:55Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | isamu-isozaki | 23,430,101 | MDQ6VXNlcjIzNDMwMTAx | User | false |
huggingface/diffusers | 1,376,658,673 | I_kwDOHa8MBc5SDiTx | 543 | https://github.com/huggingface/diffusers/issues/543 | https://api.github.com/repos/huggingface/diffusers/issues/543 | DDP Warning: Grad strides do not match bucket view strides warning | ### Describe the bug
During fine tuning the diffusion model with accelerate on multiple GPUs I got the following warning (it does not happen on a single gpu):
```
python3.8/site-packages/torch/autograd/__init__.py:173: UserWarning: Grad strides do not match bucket view strides. This may indicate grad was not created... | closed | completed | false | 7 | [
"stale"
] | [
"patrickvonplaten"
] | 2022-09-17T04:11:21Z | 2023-12-21T19:57:57Z | 2022-10-18T17:18:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mrsalehi | 38,360,019 | MDQ6VXNlcjM4MzYwMDE5 | User | false |
huggingface/diffusers | 1,948,341,090 | I_kwDOHa8MBc50IVNi | 5,432 | https://github.com/huggingface/diffusers/issues/5432 | https://api.github.com/repos/huggingface/diffusers/issues/5432 | train_text_to_image_lora_sdxl.py validation images are completely blank | ### Describe the bug
I am trying to finetune SDXL with train_text_to_image_lora_sdxl.py. Unfortunately, the created validation images are all completely blank.
The test images created at the end of the process, contain the expected output.
### Reproduction
!accelerate launch ./diffusers/examples/text_to_image/tr... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-10-17T21:48:57Z | 2023-10-18T15:50:11Z | 2023-10-18T15:50:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nikkolodeon | 64,075,461 | MDQ6VXNlcjY0MDc1NDYx | User | false |
huggingface/diffusers | 1,948,392,720 | I_kwDOHa8MBc50Ih0Q | 5,433 | https://github.com/huggingface/diffusers/issues/5433 | https://api.github.com/repos/huggingface/diffusers/issues/5433 | Visual artifacts when using `DPM++` schedulers and SDXL without the refiner model | ### Describe the bug
All DPM++ schedulers are showing visual artifacts out of the base model when `denoising_end=1` (skipping the refiner). This effect is most notable with `DPM++ 2M SDE` configured using the flag [from the docs](https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/schedulers/overview... | closed | completed | false | 19 | [
"bug"
] | [
"yiyixuxu"
] | 2023-10-17T22:27:27Z | 2023-10-30T16:36:54Z | 2023-10-30T16:36:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CodeCorrupt | 8,364,947 | MDQ6VXNlcjgzNjQ5NDc= | User | false |
huggingface/diffusers | 1,948,406,716 | I_kwDOHa8MBc50IlO8 | 5,434 | https://github.com/huggingface/diffusers/issues/5434 | https://api.github.com/repos/huggingface/diffusers/issues/5434 | Partial diffusion for the stable diffusion pipelines | Hey!
I was searching the issues page for an answer for whether we can currently do inference for any pipeline where we want to start from an intermediate timestep (say I add noise to an original image corresponding to timestep 200, then I want to be able to get the SD model's output on this noised image by telling it ... | closed | completed | false | 2 | [] | [] | 2023-10-17T22:39:53Z | 2023-10-21T11:35:01Z | 2023-10-21T11:35:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | gunshi | 8,274,753 | MDQ6VXNlcjgyNzQ3NTM= | User | false |
huggingface/diffusers | 1,948,806,180 | I_kwDOHa8MBc50KGwk | 5,436 | https://github.com/huggingface/diffusers/issues/5436 | https://api.github.com/repos/huggingface/diffusers/issues/5436 | train controlnet xl error for example | ### Describe the bug
When I train controlnet_xl use fill50k dataset ,
`File "/app/SDXL_trainer/diffusers/examples/controlnet/train_controlnet_sdxl.py", line 742, in collate_fn
prompt_ids = torch.stack([torch.tensor(example["prompt_embeds"]) for example in examples])
File "/app/SDXL_trainer/diffusers/examples... | closed | completed | false | 1 | [
"bug",
"stale"
] | [] | 2023-10-18T04:47:34Z | 2023-12-04T15:05:19Z | 2023-12-04T15:05:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kelisiya | 23,169,766 | MDQ6VXNlcjIzMTY5NzY2 | User | false |
huggingface/diffusers | 1,949,560,993 | I_kwDOHa8MBc50M_Ch | 5,441 | https://github.com/huggingface/diffusers/issues/5441 | https://api.github.com/repos/huggingface/diffusers/issues/5441 | WuerstchenPrior no attn_processors in diffusers/examples/wuerstchen/text_to_image/train_text_to_image_lora_prior.py | ### Describe the bug
~/workspace/diffusers/examples/wuerstchen/text_to_image/train_text_to_image_lora_prior.py
user~/workspace/diffusers/examples/wuerstchen/text_to_image% python train_text_to_image_lora_prior.py
2023-10-19 16:05:03.614305: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You m... | closed | completed | false | 8 | [
"bug"
] | [
"kashif"
] | 2023-10-18T11:53:25Z | 2023-11-29T15:18:22Z | 2023-11-29T15:18:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zdxpan | 16,832,891 | MDQ6VXNlcjE2ODMyODkx | User | false |
huggingface/diffusers | 1,949,561,109 | I_kwDOHa8MBc50M_EV | 5,442 | https://github.com/huggingface/diffusers/issues/5442 | https://api.github.com/repos/huggingface/diffusers/issues/5442 | [Pipeline] ScaleCrafter: Tuning-free Higher-Resolution Visual Generation with Diffusion Models | ### Model/Pipeline/Scheduler description
ScaleCrafter: Tuning-free Higher-Resolution Visual Generation with Diffusion Models : https://yingqinghe.github.io/scalecrafter/
TLDR from github repo : ScaleCrafter is capable of generating images with a resolution of 4096 x 4096 and videos with a resolution of 2048 x 115... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-10-18T11:53:28Z | 2023-11-28T15:05:49Z | 2023-11-28T15:05:49Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ayushtues | 43,698,245 | MDQ6VXNlcjQzNjk4MjQ1 | User | false |
huggingface/diffusers | 1,949,623,269 | I_kwDOHa8MBc50NOPl | 5,443 | https://github.com/huggingface/diffusers/issues/5443 | https://api.github.com/repos/huggingface/diffusers/issues/5443 | DeferredCudaCallError: CUDA call failed lazily at initialization with error: device >= 0 && device < num_gpus INTERNAL ASSERT FAILED at "../aten/src/ATen/cuda/CUDAContext.cpp":50, please report a bug to PyTorch. |
### Describe the bug
```
import torch
from diffusers import StableDiffusionXLInpaintPipeline
from diffusers.utils import load_image
from diffusers import ControlNetModel, StableDiffusionXLControlNetImg2ImgPipeline, AutoencoderKL
# import os
# os.environ["CUDA_VISIBLE_DEVICES"] = "1"
device = "cuda"
pipe ... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-10-18T12:23:56Z | 2023-11-28T15:05:46Z | 2023-11-28T15:05:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | p-moon | 16,488,309 | MDQ6VXNlcjE2NDg4MzA5 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.