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 | 1,646,804,446 | I_kwDOHa8MBc5iKD3e | 2,895 | https://github.com/huggingface/diffusers/issues/2895 | https://api.github.com/repos/huggingface/diffusers/issues/2895 | UNet3DConditionModel not using Condition | ### Describe the bug
I could be entirely lost here, but it seems `UNet3DConditionModel` doesn't properly use `class_labels`. In fact, it doesn't use it at all. It seems odd to me that a "ConditionModel" doesn't use its condition.
### Reproduction
Doesn't need reproduction. The variable is unused in the source.
###... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-03-30T02:27:48Z | 2023-05-08T15:03:24Z | 2023-05-08T15:03:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Reichenbachian | 8,087,473 | MDQ6VXNlcjgwODc0NzM= | User | false |
huggingface/diffusers | 1,646,956,317 | I_kwDOHa8MBc5iKo8d | 2,898 | https://github.com/huggingface/diffusers/issues/2898 | https://api.github.com/repos/huggingface/diffusers/issues/2898 | Is it possible to run text2video-zero with less than 15GB GPU? | ### Describe the bug
Hi 👋 everybody, I have a question is it possible to use `xformers` with `text2video-zero`
when I add `self.pipe.enable_xformers_memory_efficient_attention()` I am getting a temporal consistency problem
| with xformers | without xformers (needs more than 20GB vram)
| --- | --- |
![image_5... | closed | completed | false | 28 | [
"bug"
] | [] | 2023-03-30T05:46:18Z | 2023-04-18T21:46:18Z | 2023-04-18T21:46:18Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | camenduru | 54,370,274 | MDQ6VXNlcjU0MzcwMjc0 | User | false |
huggingface/diffusers | 1,279,084,562 | I_kwDOHa8MBc5MPUgS | 29 | https://github.com/huggingface/diffusers/issues/29 | https://api.github.com/repos/huggingface/diffusers/issues/29 | In which config (or in None) should the `predict_epsilon` config be saved? | We're discussing where certain properties should be saved in this repository. For example, [this `predict_epsilon`](https://github.com/huggingface/diffusers/blob/320506c75ad7d40c7d9fc6db5a45955906c5f729/src/diffusers/schedulers/scheduling_ddpm.py#L125) property is used to distinguish if a model is trained to directly r... | closed | completed | false | 4 | [
"stale"
] | [
"natolambert",
"patrickvonplaten",
"anton-l",
"patil-suraj"
] | 2022-06-21T21:03:31Z | 2022-10-27T08:28:58Z | 2022-10-27T08:28:58Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | natolambert | 10,695,622 | MDQ6VXNlcjEwNjk1NjIy | User | false |
huggingface/diffusers | 1,357,094,775 | I_kwDOHa8MBc5Q4593 | 290 | https://github.com/huggingface/diffusers/issues/290 | https://api.github.com/repos/huggingface/diffusers/issues/290 | Training examples missing | Hello, in the Readme file you mentioned that the training examples can be found in `examples/training` directory, but it seems that it is not there. Maybe because it is under developing but I am really looking forward to see it. Big hug to this love community! :) | closed | completed | false | 2 | [] | [] | 2022-08-31T09:54:23Z | 2022-08-31T16:34:56Z | 2022-08-31T16:34:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | henrydylan | 104,773,765 | U_kgDOBj64hQ | User | false |
huggingface/diffusers | 1,646,991,522 | I_kwDOHa8MBc5iKxii | 2,900 | https://github.com/huggingface/diffusers/issues/2900 | https://api.github.com/repos/huggingface/diffusers/issues/2900 | How to merge lora weights into the unet and save the unet? | I want to merge my lora weights into the unet and then save the whole model together. When I load the model weight, I got an error:
```
from diffusers import StableDiffusionImg2ImgPipeline,DiffusionPipeline
import torch,time
pipe = DiffusionPipeline.from_pretrained("./dreamlike-photoreal-2.0/", torch_dtype=torch.fl... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-03-30T06:24:30Z | 2023-05-07T15:02:46Z | 2023-05-07T15:02:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | youyuge34 | 15,784,256 | MDQ6VXNlcjE1Nzg0MjU2 | User | false |
huggingface/diffusers | 1,647,814,025 | I_kwDOHa8MBc5iN6WJ | 2,904 | https://github.com/huggingface/diffusers/issues/2904 | https://api.github.com/repos/huggingface/diffusers/issues/2904 | AudioLDM Pipeline doesn't run: AttributeError: 'NoneType' object has no attribute 'scheduler' | ### Describe the bug
Downloaded weights from "cvssp/audioldm" to local folder,
running an example from docs (trying), receiving AttributeError: NoneType after initializing of the pipeline
### Reproduction
```
from diffusers import AudioLDMPipeline, DPMSolverMultistepScheduler
import torch
import sys
imp... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-03-30T15:05:05Z | 2023-05-16T23:22:08Z | 2023-05-16T23:22:07Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Pirog17000 | 3,022,852 | MDQ6VXNlcjMwMjI4NTI= | User | false |
huggingface/diffusers | 1,647,962,183 | I_kwDOHa8MBc5iOehH | 2,905 | https://github.com/huggingface/diffusers/issues/2905 | https://api.github.com/repos/huggingface/diffusers/issues/2905 | Add Karras Sigmas to all schedulers | **Is your feature request related to a problem? Please describe.**
As stated in multiple issue:
- https://github.com/huggingface/diffusers/issues/1887
- https://github.com/huggingface/diffusers/issues/2801
- https://github.com/huggingface/diffusers/issues/2676
**Describe the solution you'd like**
It would... | closed | completed | false | 7 | [
"Good second issue"
] | [
"sayakpaul",
"patrickvonplaten"
] | 2023-03-30T16:42:51Z | 2023-06-20T11:13:42Z | 2023-06-20T11:13:41Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,648,307,318 | I_kwDOHa8MBc5iPyx2 | 2,906 | https://github.com/huggingface/diffusers/issues/2906 | https://api.github.com/repos/huggingface/diffusers/issues/2906 | NameError: name 'PreTrainedTokenizer' is not defined - TextualInversionLoaderMixin | ### Describe the bug
On the main branch (a937e1b5) of diffusers, when `is_transformers_available` returns False, PreTrainedTokenizer is not defined. This would be fine, but PreTrainedTokenizer is used as a type annotation for some methods. Anything that references diffusers/loaders/TextualInversionLoaderMixin will... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-03-30T20:57:07Z | 2023-04-04T13:06:10Z | 2023-03-31T16:41:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JosephCatrambone | 2,160,055 | MDQ6VXNlcjIxNjAwNTU= | User | false |
huggingface/diffusers | 1,648,439,924 | I_kwDOHa8MBc5iQTJ0 | 2,907 | https://github.com/huggingface/diffusers/issues/2907 | https://api.github.com/repos/huggingface/diffusers/issues/2907 | Pipelines fail if enable_model_cpu_offload is called twice | ### Describe the bug
`enable_model_cpu_offload` seem to be tripping over its feet if it's called a second time. See repro case.
From what I can see, `pipe._execution_device` is `cuda:0` after the first call, but switches to `cpu` after the second call. I'd expect the second call to be pretty much a no-op.
### Repr... | closed | completed | false | 11 | [
"bug"
] | [
"williamberman"
] | 2023-03-30T23:08:15Z | 2023-04-27T18:07:33Z | 2023-04-27T18:07:33Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | antoche | 1,627,384 | MDQ6VXNlcjE2MjczODQ= | User | false |
huggingface/diffusers | 1,648,484,880 | I_kwDOHa8MBc5iQeIQ | 2,908 | https://github.com/huggingface/diffusers/issues/2908 | https://api.github.com/repos/huggingface/diffusers/issues/2908 | Array shape and Mixed Precision errors in examples/controlnet/train_controlnet.py script | ### Describe the bug
2 bugs:
Array shape:
- During the `log_validation` function, when [logging validation images](https://github.com/huggingface/diffusers/blob/a937e1b594da34b35ea9a090dc3ada57df12df49/examples/controlnet/train_controlnet.py#L134) to tensorboard, the call to `np.array` on the PIL validation imag... | closed | completed | false | 7 | [
"bug",
"stale"
] | [
"williamberman",
"yiyixuxu"
] | 2023-03-31T00:14:09Z | 2023-05-29T15:03:53Z | 2023-05-29T15:03:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jeromeku | 2,455,711 | MDQ6VXNlcjI0NTU3MTE= | User | false |
huggingface/diffusers | 1,648,496,707 | I_kwDOHa8MBc5iQhBD | 2,909 | https://github.com/huggingface/diffusers/issues/2909 | https://api.github.com/repos/huggingface/diffusers/issues/2909 | Controlnet training with cycle loss | Many times, I want to train my custom controlnet but find it very difficult to find paired data.
So, I was wondering if it is possible to modify controlnet training procedure to add cycle loss? That will make 100s of more controls possible which are not possible currently because of paired data limitations. | closed | completed | false | 2 | [
"stale"
] | [] | 2023-03-31T00:35:07Z | 2023-05-08T15:03:20Z | 2023-05-08T15:03:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dmangla3 | 28,572,201 | MDQ6VXNlcjI4NTcyMjAx | User | false |
huggingface/diffusers | 1,649,080,075 | I_kwDOHa8MBc5iSvcL | 2,916 | https://github.com/huggingface/diffusers/issues/2916 | https://api.github.com/repos/huggingface/diffusers/issues/2916 | How to support multiple concurrency? | I used gradio to make diffusers available on website, but the result were messed up when two people using it at the same time.
How can I handle this problem? | closed | completed | false | 3 | [
"stale"
] | [] | 2023-03-31T10:08:59Z | 2023-05-08T15:03:19Z | 2023-05-08T15:03:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | blx0102 | 17,761,044 | MDQ6VXNlcjE3NzYxMDQ0 | User | false |
huggingface/diffusers | 1,357,162,682 | I_kwDOHa8MBc5Q5Ki6 | 292 | https://github.com/huggingface/diffusers/issues/292 | https://api.github.com/repos/huggingface/diffusers/issues/292 | Add inference support for `mps` device (Apple Silicon) | There is a lot of community interest in running diffusers on Apple Silicon. A first step could be to introduce support for the `mps` device, which currently requires PyTorch nightly builds. Another step down the line could be to convert to Core ML and optimize to make sure that the use of the ANE (Neural Engine) is max... | closed | completed | false | 5 | [] | [
"pcuenca"
] | 2022-08-31T10:55:34Z | 2023-05-02T07:46:45Z | 2022-09-08T11:37:37Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pcuenca | 1,177,582 | MDQ6VXNlcjExNzc1ODI= | User | false |
huggingface/diffusers | 1,357,272,232 | I_kwDOHa8MBc5Q5lSo | 293 | https://github.com/huggingface/diffusers/issues/293 | https://api.github.com/repos/huggingface/diffusers/issues/293 | [Docs] Doc Sprint on Wednesday Sep 7 | ** There is still more TODO, to see open options, see [this comment](https://github.com/huggingface/diffusers/issues/293#issuecomment-1239822401)**
----
Let's add them docs.
- Create a minimum toctree with mostly blank pages to get it working
- Merge: https://github.com/huggingface/moon-landing/pull/3415
- Do ... | closed | completed | false | 22 | [
"wip"
] | [] | 2022-08-31T12:34:13Z | 2024-11-21T11:24:02Z | 2024-11-21T11:24:02Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,649,739,280 | I_kwDOHa8MBc5iVQYQ | 2,940 | https://github.com/huggingface/diffusers/issues/2940 | https://api.github.com/repos/huggingface/diffusers/issues/2940 | Token Merging (ToMe) for Stable Diffusion | ToMe for SD speeds up diffusion by merging redundant tokens. by @dbolya
<img width="813" alt="2023-04-01 0 06 41" src="https://user-images.githubusercontent.com/10776/229186302-f8454de9-94c6-4a29-be0f-4cc88254f785.png">
>Token Merging (ToMe) speeds up transformers by merging redundant tokens, which means the tran... | closed | completed | false | 8 | [] | [
"sayakpaul"
] | 2023-03-31T17:18:30Z | 2024-03-12T08:35:12Z | 2023-04-26T10:56:10Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | takuma104 | 10,776 | MDQ6VXNlcjEwNzc2 | User | false |
huggingface/diffusers | 1,649,847,350 | I_kwDOHa8MBc5iVqw2 | 2,941 | https://github.com/huggingface/diffusers/issues/2941 | https://api.github.com/repos/huggingface/diffusers/issues/2941 | EulerAncestralDiscreteScheduler produces undesired results with Stable Diffusion pipeline | ### Describe the bug
The generated images are not natural and too blurry.
### Reproduction
scheduler = EulerAncestralDiscreteScheduler()
pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", scheduler=scheduler)
pipe = pipe.to('cuda')
# example 1
pipe('A dog eat on the pizza').images[... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-03-31T18:41:44Z | 2023-05-09T15:03:21Z | 2023-05-09T15:03:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zxydi1992 | 9,076,986 | MDQ6VXNlcjkwNzY5ODY= | User | false |
huggingface/diffusers | 1,650,587,718 | I_kwDOHa8MBc5iYfhG | 2,946 | https://github.com/huggingface/diffusers/issues/2946 | https://api.github.com/repos/huggingface/diffusers/issues/2946 | Optimizer selection | I'm wondering if it's possible to integrate the [Lion optimizer](https://github.com/lucidrains/lion-pytorch) in train_dreambooth.py in place of AdamW if we choose, like optimizer args. I know this would conflict with people who have low VRAM and use 8bit-Adam. So Just a store_true option maybe.
I'm going to have a ... | closed | completed | false | 3 | [] | [] | 2023-04-01T18:14:29Z | 2023-04-08T05:44:50Z | 2023-04-08T05:44:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jmaccall316 | 120,006,038 | U_kgDOBycllg | User | false |
huggingface/diffusers | 1,650,810,862 | I_kwDOHa8MBc5iZV_u | 2,947 | https://github.com/huggingface/diffusers/issues/2947 | https://api.github.com/repos/huggingface/diffusers/issues/2947 | AttributeError: 'UNet2DConditionModel' object has no attribute 'load_attn_procs' | ### Describe the bug
I am trying to follow [SD LORA training tutorial](https://huggingface.co/blog/lora )
After training on my custom dataset, the following function
`pipe.unet.load_attn_procs `
Returns this error
`AttributeError: 'UNet2DConditionModel' object has no attribute 'load_attn_procs'`
### Reproduct... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-04-02T05:19:32Z | 2023-04-03T08:32:49Z | 2023-04-03T08:32:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dokluch | 135,181 | MDQ6VXNlcjEzNTE4MQ== | User | false |
huggingface/diffusers | 1,651,070,194 | I_kwDOHa8MBc5iaVTy | 2,949 | https://github.com/huggingface/diffusers/issues/2949 | https://api.github.com/repos/huggingface/diffusers/issues/2949 | PyTorch Error in Line 251 of LMS | diffusers/schedulers/scheduling_lms_discrete.py:251
```
if isinstance(timestep, torch.Tensor):
timestep = timestep.to(self.timesteps.device)
step_index = (self.timesteps == timestep).nonzero().item()
sigma = self.sigmas[step_index]
```
When running this I get an error from... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-04-02T19:43:28Z | 2023-05-11T15:04:06Z | 2023-05-11T15:04:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Mark-777-0 | 83,738,282 | MDQ6VXNlcjgzNzM4Mjgy | User | false |
huggingface/diffusers | 1,357,496,137 | I_kwDOHa8MBc5Q6b9J | 295 | https://github.com/huggingface/diffusers/issues/295 | https://api.github.com/repos/huggingface/diffusers/issues/295 | Having "ImportError:" | I got this error
'''
ImportError: cannot import name 'PROTOCOL_TLS' from 'urllib3.util.ssl_'
'''
after installing diffusers, PyTorch, and transformers.
| closed | completed | false | 2 | [] | [] | 2022-08-31T15:19:52Z | 2022-09-03T10:09:09Z | 2022-09-03T10:08:51Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AhmedUZaki | 79,209,221 | MDQ6VXNlcjc5MjA5MjIx | User | false |
huggingface/diffusers | 1,651,158,495 | I_kwDOHa8MBc5iaq3f | 2,950 | https://github.com/huggingface/diffusers/issues/2950 | https://api.github.com/repos/huggingface/diffusers/issues/2950 | Why img2img as a separate pipeline? | **What API design would you like to have changed or added to the library? Why?**
I find it strange that img2img is a separate pipeline from `StableDiffusionPipeline`. The problem is that I have very limited VRAM (6GB) and I can't afford to load two separate pipelines for both txt2img and img2img. In general it seems... | closed | completed | false | 8 | [] | [] | 2023-04-03T00:42:30Z | 2023-12-01T10:13:46Z | 2023-04-11T14:34:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | oxysoft | 3,945,277 | MDQ6VXNlcjM5NDUyNzc= | User | false |
huggingface/diffusers | 1,652,314,117 | I_kwDOHa8MBc5ifFAF | 2,955 | https://github.com/huggingface/diffusers/issues/2955 | https://api.github.com/repos/huggingface/diffusers/issues/2955 | I built it with AMD6700XT video card | ### Describe the bug
python networks/merge_lora.py --sd_model F:\Programming language\Stable-Diffusion-ONNX-FP16\chilloutmix_Ni.safetensors --save_to F:\Programming language\Stable-Diffusion-ONNX-FP16\sd-model\chilloutmix-lora-mix VirtualGirl-Rin.safetensors --models F:\Programming language\Stable-Diffusion-ONNX-FP1... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-04-03T15:40:35Z | 2023-05-13T15:03:18Z | 2023-05-13T15:03:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FranklinBBai | 129,773,438 | U_kgDOB7wvfg | User | false |
huggingface/diffusers | 1,652,791,398 | I_kwDOHa8MBc5ig5hm | 2,957 | https://github.com/huggingface/diffusers/issues/2957 | https://api.github.com/repos/huggingface/diffusers/issues/2957 | local variable 'train_dataset' referenced before assignment | ### Describe the bug
Hi team,
Recently, we are running Diffusion model on JAX, and noticed an `UnboundLocalError` issue for local variable `train_dataset`. It could be a simple [fix](https://github.com/RissyRan/diffusers/blob/main/examples/text_to_image/train_text_to_image_flax.py#L396) as this forked repo.
#... | closed | completed | false | 7 | [
"bug",
"stale"
] | [
"yiyixuxu"
] | 2023-04-03T21:20:37Z | 2023-05-10T23:58:17Z | 2023-05-10T23:58:17Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | RissyRan | 20,385,466 | MDQ6VXNlcjIwMzg1NDY2 | User | false |
huggingface/diffusers | 1,653,123,962 | I_kwDOHa8MBc5iiKt6 | 2,960 | https://github.com/huggingface/diffusers/issues/2960 | https://api.github.com/repos/huggingface/diffusers/issues/2960 | StableDiffusionInpaintPipeline fails to preserve black pixels | Hi
I'm using StableDiffusionInpaintPipeline with runwayml/stable-diffusion-inpainting model and following the official instruction to do some inpainting work.
According to https://huggingface.co/docs/diffusers/api/pipelines/stable_diffusion/inpaint, black pixels would be preserved.
But when I check the results it ... | closed | completed | false | 5 | [] | [] | 2023-04-04T03:48:59Z | 2023-04-04T08:10:31Z | 2023-04-04T08:09:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zihaog0724 | 63,634,669 | MDQ6VXNlcjYzNjM0NjY5 | User | false |
huggingface/diffusers | 1,653,565,895 | I_kwDOHa8MBc5ij2nH | 2,961 | https://github.com/huggingface/diffusers/issues/2961 | https://api.github.com/repos/huggingface/diffusers/issues/2961 | Training example for StableUnCLIP | The [pipeline_stable_unclip_img2img.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py) is great, but how do I finetune the model just as done in [train_text_to_image.py](https://github.com/huggingface/diffusers/blob/main/examples/text_to_im... | closed | completed | false | 8 | [
"stale"
] | [] | 2023-04-04T09:58:35Z | 2024-09-06T05:04:07Z | 2023-05-13T15:03:17Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | haofanwang | 18,741,068 | MDQ6VXNlcjE4NzQxMDY4 | User | false |
huggingface/diffusers | 1,653,659,208 | I_kwDOHa8MBc5ikNZI | 2,964 | https://github.com/huggingface/diffusers/issues/2964 | https://api.github.com/repos/huggingface/diffusers/issues/2964 | LoRA for clip and vae? | LoRA now only add parameters to the attention in unet. In my experiment, the performance is not good. I hope we can support lora layer for unet and clip. best wishes | closed | completed | false | 5 | [
"stale"
] | [] | 2023-04-04T10:57:34Z | 2024-04-29T13:05:33Z | 2023-05-13T15:03:15Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Zyriix | 40,727,904 | MDQ6VXNlcjQwNzI3OTA0 | User | false |
huggingface/diffusers | 1,653,799,635 | I_kwDOHa8MBc5ikvrT | 2,965 | https://github.com/huggingface/diffusers/issues/2965 | https://api.github.com/repos/huggingface/diffusers/issues/2965 | Is there a straightforward way to initialize the DDPMpipeline with an openai pretrained model ? | Hi,
I want to use the following imagenet DDPM pretrained model from the openai repo https://openaipublic.blob.core.windows.net/diffusion/jul-2021/256x256_diffusion_uncond.pt.
I wonder if there is a straightforward way to initialize the DDPMpipeline using the pretrained model?
Thanks. | closed | completed | false | 4 | [
"stale"
] | [] | 2023-04-04T12:30:33Z | 2023-05-24T15:03:46Z | 2023-05-24T15:03:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hamadichihaoui | 13,422,842 | MDQ6VXNlcjEzNDIyODQy | User | false |
huggingface/diffusers | 1,654,037,407 | I_kwDOHa8MBc5ilpuf | 2,966 | https://github.com/huggingface/diffusers/issues/2966 | https://api.github.com/repos/huggingface/diffusers/issues/2966 | train_instruct_pix2pix.py in example has problem when try train with mutil gpus? | ### Describe the bug
When I try to use accelerate launch train_instruct_pix2pix.py with mutil gpus, it report the error as below:
----------------------------------
Traceback (most recent call last):
File "train_instruct_pix2pix.py", line 1002, in <module>
main()
File "train_instruct_pix2pix.py", line 7... | closed | completed | false | 9 | [
"bug"
] | [] | 2023-04-04T14:46:07Z | 2023-04-12T09:13:56Z | 2023-04-12T09:13:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | whbzju | 1,175,651 | MDQ6VXNlcjExNzU2NTE= | User | false |
huggingface/diffusers | 1,654,147,704 | I_kwDOHa8MBc5imEp4 | 2,967 | https://github.com/huggingface/diffusers/issues/2967 | https://api.github.com/repos/huggingface/diffusers/issues/2967 | Cannot get simple UNet1D to run | This is using the latest huggingface diffusers:
```
from diffusers.models.unet_1d import UNet1DModel
import torch
net = UNet1DModel().cuda()
input_data = torch.rand(1, 2, 16000).cuda()
net(input_data, 1)
```
gives the following:
```
╭─────────────────────────────── Traceback (most recent call last) ──... | closed | completed | false | 8 | [
"stale"
] | [] | 2023-04-04T15:50:24Z | 2023-08-30T05:32:12Z | 2023-05-14T15:03:11Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | turian | 65,918 | MDQ6VXNlcjY1OTE4 | User | false |
huggingface/diffusers | 1,654,278,759 | I_kwDOHa8MBc5imkpn | 2,971 | https://github.com/huggingface/diffusers/issues/2971 | https://api.github.com/repos/huggingface/diffusers/issues/2971 | Supporting Guess Mode with StableDiffusionControlNetPipeline | The ControlNet model supports [Guess Mode](https://github.com/lllyasviel/ControlNet#guess-mode--non-prompt-mode) that helps generate results without any text prompts. It recognizes the contents of the control map quite well and generates amazing outputs without any text prompt input.
I would want to place request fo... | closed | completed | false | 9 | [] | [] | 2023-04-04T17:22:57Z | 2023-11-25T19:44:41Z | 2023-04-27T16:12:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hireshgupta1997 | 16,414,557 | MDQ6VXNlcjE2NDE0NTU3 | User | false |
huggingface/diffusers | 1,654,351,378 | I_kwDOHa8MBc5im2YS | 2,972 | https://github.com/huggingface/diffusers/issues/2972 | https://api.github.com/repos/huggingface/diffusers/issues/2972 | The adjustment assigned to 0.01 in the loop and never re-assigned to 0 | https://github.com/huggingface/diffusers/blob/0c63c3839a8dbaf336f640db3ddc8462d4f6711a/src/diffusers/pipelines/stable_diffusion/safety_checker.py#L73 | closed | completed | false | 6 | [
"stale"
] | [] | 2023-04-04T18:17:37Z | 2023-05-14T15:03:10Z | 2023-05-14T15:03:10Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zetyquickly | 25,350,960 | MDQ6VXNlcjI1MzUwOTYw | User | false |
huggingface/diffusers | 1,654,356,639 | I_kwDOHa8MBc5im3qf | 2,973 | https://github.com/huggingface/diffusers/issues/2973 | https://api.github.com/repos/huggingface/diffusers/issues/2973 | Stable diffusion 1.5: Different results with enable_model_cpu_offload and without | ### Describe the bug
`StableDiffusionPipeline` pipeline with and without `enable_model_cpu_offload` give different results with same parameters.
# Examples
![test2](https://user-images.githubuserco... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-04-04T18:21:06Z | 2023-04-07T13:28:54Z | 2023-04-07T13:28:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | KernelA | 17,554,646 | MDQ6VXNlcjE3NTU0NjQ2 | User | false |
huggingface/diffusers | 1,654,685,698 | I_kwDOHa8MBc5ioIAC | 2,975 | https://github.com/huggingface/diffusers/issues/2975 | https://api.github.com/repos/huggingface/diffusers/issues/2975 | Kandinsky 2.0 and 2.1 support | Another VQGAN based model, but with multi-lingual support and an UNCLIP style prior that allows image variations and image mixing.
https://github.com/ai-forever/Kandinsky-2
The FID is reportedly good (8.21) and since it's token based things like inpainting/outpainting are easy.
Unfortunately it is Russian stat... | closed | completed | false | 4 | [] | [] | 2023-04-04T23:00:46Z | 2023-05-14T15:03:09Z | 2023-05-14T15:03:09Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AmericanPresidentJimmyCarter | 110,263,573 | U_kgDOBpJ9FQ | User | false |
huggingface/diffusers | 1,654,784,854 | I_kwDOHa8MBc5iogNW | 2,977 | https://github.com/huggingface/diffusers/issues/2977 | https://api.github.com/repos/huggingface/diffusers/issues/2977 | Is it possible to inference using multiple GPUs? | Hi, Thanks for sharing this library for using stable diffusion.
There is one questions I want to ask.
Like title, Is it possible to inference using multiple GPUs? If possible, how?
do you share doc about inference using multiple GPUs?
```
OS: Linux Ubuntu 20.04
GPU: RTX 4090 (24GB) * n
RAM: 72GB
Python: 3.9... | closed | completed | false | 18 | [
"stale"
] | [] | 2023-04-05T01:24:32Z | 2023-12-27T14:05:52Z | 2023-06-09T15:03:15Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DevJunghun | 28,581,778 | MDQ6VXNlcjI4NTgxNzc4 | User | false |
huggingface/diffusers | 1,655,306,856 | I_kwDOHa8MBc5iqfpo | 2,979 | https://github.com/huggingface/diffusers/issues/2979 | https://api.github.com/repos/huggingface/diffusers/issues/2979 | TypeError: forward() got an unexpected keyword argument 'encoder_hidden_states' | Hello developers! Thank you for your contribution! I'm having problems with a github library provided by google (this library is mainly based on huggingface diffusers ): some variable names may be changed in newer versions of diffusers. Therefore, you may encounter some problems with variable names when using third-par... | closed | completed | false | 2 | [] | [] | 2023-04-05T09:59:11Z | 2024-01-15T08:58:47Z | 2023-04-06T14:33:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | HiddenGalaxy | 128,576,899 | U_kgDOB6ntgw | User | false |
huggingface/diffusers | 1,357,628,325 | I_kwDOHa8MBc5Q68Ol | 298 | https://github.com/huggingface/diffusers/issues/298 | https://api.github.com/repos/huggingface/diffusers/issues/298 | PNDM bug | ### Describe the bug
A PNDM bug got introduced in commit b64c5227595f5eed10f6ff3ac7953de0bb07ab2d. I get IndexErrors and a highly increased step count
### Reproduction
pipe.scheduler = PNDMScheduler(beta_start=0.00085, beta_end=0.012, beta_schedule="scaled_linear", num_train_timesteps=1000)
### Logs
_No response_
... | closed | completed | false | 4 | [
"bug"
] | [] | 2022-08-31T17:11:06Z | 2022-08-31T17:28:22Z | 2022-08-31T17:27:50Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dblunk88 | 39,381,389 | MDQ6VXNlcjM5MzgxMzg5 | User | false |
huggingface/diffusers | 1,655,648,091 | I_kwDOHa8MBc5iry9b | 2,981 | https://github.com/huggingface/diffusers/issues/2981 | https://api.github.com/repos/huggingface/diffusers/issues/2981 | Training text_to_img with multi gpu setup | Hi,
I really like the text_to_image training script and had really good results with it. I want to train a model now with multiple GPUs.
My current understanding:
I want to use a data paralell setup thus each GPU gets its batches and at the end they sync the gradients. This should not result in a significant incre... | closed | completed | false | 3 | [] | [] | 2023-04-05T13:43:02Z | 2023-04-18T03:08:44Z | 2023-04-18T03:08:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | flixmk | 66,331,106 | MDQ6VXNlcjY2MzMxMTA2 | User | false |
huggingface/diffusers | 1,655,988,791 | I_kwDOHa8MBc5itGI3 | 2,984 | https://github.com/huggingface/diffusers/issues/2984 | https://api.github.com/repos/huggingface/diffusers/issues/2984 | Latent Video Diffusion Models for High-Fidelity Long Video Generation | ### Model/Pipeline/Scheduler description
VideoCrafter is an open-source video generation and editing toolbox for crafting video content.
GitHub: https://github.com/VideoCrafter/VideoCrafter
### Open source status
- [X] The model implementation is available
- [X] The model weights are available (Only relevant if a... | closed | completed | false | 2 | [
"New pipeline/model"
] | [] | 2023-04-05T17:00:40Z | 2024-11-21T15:37:01Z | 2024-11-21T15:37:01Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,655,991,579 | I_kwDOHa8MBc5itG0b | 2,985 | https://github.com/huggingface/diffusers/issues/2985 | https://api.github.com/repos/huggingface/diffusers/issues/2985 | Kandinsky 2.1 | ### Model/Pipeline/Scheduler description
Kandinsky 2.1 inherits best practicies from Dall-E 2 and Latent diffusion, while introducing some new ideas.
As text and image encoder it uses CLIP model and diffusion image prior (mapping) between latent spaces of CLIP modalities. This approach increases the visual performa... | closed | completed | false | 19 | [
"New pipeline/model"
] | [
"yiyixuxu"
] | 2023-04-05T17:02:46Z | 2023-10-02T16:45:25Z | 2023-10-02T16:45:25Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,656,145,160 | I_kwDOHa8MBc5itsUI | 2,988 | https://github.com/huggingface/diffusers/issues/2988 | https://api.github.com/repos/huggingface/diffusers/issues/2988 | Scheduler crashes when num_training_steps == num_inference_steps | ### Describe the bug
`scheduling_dpmsolver_multistep.set_timesteps()` computes a `timesteps` array, which the scheduler accesses during forward steps, the array is rounded to the nearest int which will give duplicate entries in certain cases (e.g. when `num_train_timesteps` is equal to `num_inference_steps`). This cau... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-04-05T19:06:31Z | 2023-04-10T20:20:30Z | 2023-04-10T20:20:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lucala | 14,397,281 | MDQ6VXNlcjE0Mzk3Mjgx | User | false |
huggingface/diffusers | 1,656,557,485 | I_kwDOHa8MBc5ivQ-t | 2,990 | https://github.com/huggingface/diffusers/issues/2990 | https://api.github.com/repos/huggingface/diffusers/issues/2990 | How to preserve gradient after vae? | Currently when passing latents to the VAE the latents.requires_grad become false, I wonder is there any method to perseve this graidents for the applications like the Clip guidance? | closed | completed | false | 3 | [] | [] | 2023-04-06T02:12:58Z | 2023-04-06T14:52:55Z | 2023-04-06T14:52:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zzhang2816 | 77,660,497 | MDQ6VXNlcjc3NjYwNDk3 | User | false |
huggingface/diffusers | 1,656,559,343 | I_kwDOHa8MBc5ivRbv | 2,991 | https://github.com/huggingface/diffusers/issues/2991 | https://api.github.com/repos/huggingface/diffusers/issues/2991 | [Examples] ControlNet (PyTorch) example fails when using mixed-precision | ### Describe the bug
The ControlNet training [example](https://github.com/huggingface/diffusers/tree/main/examples/controlnet) (PyTorch variant) is failing when used with mixed-precision.
Here's the command I used:
```bash
accelerate launch train_controlnet.py \
--pretrained_model_name_or_path=runwayml/stab... | closed | completed | false | 2 | [
"bug"
] | [
"williamberman",
"yiyixuxu"
] | 2023-04-06T02:16:13Z | 2023-04-12T07:23:31Z | 2023-04-12T07:23:23Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,656,668,910 | I_kwDOHa8MBc5ivsLu | 2,995 | https://github.com/huggingface/diffusers/issues/2995 | https://api.github.com/repos/huggingface/diffusers/issues/2995 | RuntimeError: mat1 and mat2 shapes cannot be multiplied (154x1024 and 768x320) | ### Describe the bug
With controlnet pipeline, if `'runwayml/stable-diffusion-v1-5'` is used, it works but with `'stabilityai/stable-diffusion-2-1'`, it doesn't. It looks like controlnet weight for sd 2 for `lllyasviel/sd-controlnet-canny` doesn't exist. Is there any support?
### Reproduction
```
canny = ControlNet... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-04-06T05:02:35Z | 2023-11-07T01:32:11Z | 2023-05-28T15:03:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pure-rgb | 45,315,076 | MDQ6VXNlcjQ1MzE1MDc2 | User | false |
huggingface/diffusers | 1,657,445,651 | I_kwDOHa8MBc5iyp0T | 2,997 | https://github.com/huggingface/diffusers/issues/2997 | https://api.github.com/repos/huggingface/diffusers/issues/2997 | [Examples] Test and update READMEs on multi-GPU support | We keep getting issues where community members want to know how to do multi-GPU training with our training examples. An example: https://github.com/huggingface/diffusers/issues/2981.
I think we should test the following examples on a priority (as for others, multi-GPU is probably overkill) and ensure their respectiv... | closed | completed | false | 3 | [] | [
"williamberman",
"yiyixuxu",
"sayakpaul"
] | 2023-04-06T13:36:55Z | 2023-04-18T03:06:15Z | 2023-04-18T03:06:15Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,657,668,436 | I_kwDOHa8MBc5izgNU | 2,999 | https://github.com/huggingface/diffusers/issues/2999 | https://api.github.com/repos/huggingface/diffusers/issues/2999 | vq-reg model | Since vq-reg model has been metioned by stable diffusion original paper, is there any implementation about that? ty
| closed | completed | false | 5 | [
"stale"
] | [] | 2023-04-06T15:48:13Z | 2024-01-22T02:34:19Z | 2023-05-15T15:03:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rese1f | 58,205,475 | MDQ6VXNlcjU4MjA1NDc1 | User | false |
huggingface/diffusers | 1,279,579,176 | I_kwDOHa8MBc5MRNQo | 30 | https://github.com/huggingface/diffusers/issues/30 | https://api.github.com/repos/huggingface/diffusers/issues/30 | example script doesn't work with pip install | The pip package doesn't include any of these:
from diffusers.modeling_utils import unwrap_model
from diffusers.optimization import get_scheduler
from diffusers.utils import logging
so "train_unconditional.py" fails:
`python3 train_unconditional.py --dataset="cifar10" --resolution=64 --output_path="cifar... | closed | completed | false | 3 | [] | [] | 2022-06-22T05:46:50Z | 2022-06-30T22:19:00Z | 2022-06-30T22:19:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | richardrl | 5,181,990 | MDQ6VXNlcjUxODE5OTA= | User | false |
huggingface/diffusers | 1,357,718,737 | I_kwDOHa8MBc5Q7STR | 300 | https://github.com/huggingface/diffusers/issues/300 | https://api.github.com/repos/huggingface/diffusers/issues/300 | F401 quality warnings | Hello, I noticed your comment in the top `__init__.py` module:
> There's no way to ignore "F401 '...' imported but unused" warnings in this
module, but to preserve other warnings. So, don't check this module at all.
There is a way: it's to add the objects you want to export to the `__all__` variable. Flake8 won'... | closed | completed | false | 1 | [] | [] | 2022-08-31T18:43:55Z | 2022-09-01T12:56:26Z | 2022-09-01T12:56:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pawamoy | 3,999,221 | MDQ6VXNlcjM5OTkyMjE= | User | false |
huggingface/diffusers | 1,658,297,216 | I_kwDOHa8MBc5i15uA | 3,004 | https://github.com/huggingface/diffusers/issues/3004 | https://api.github.com/repos/huggingface/diffusers/issues/3004 | weight auto saving location | ### Describe the bug
Using
```
pipe = StableDiffusionPipeline.from_pretrained(
model_id, torch_dtype=torch.float16
)
```
the weight is saved to some location (ie.e ./cache/huggingface/...). But they're not there. How to get those files?
### Reproduction
```
pipe = StableDiffusionPipeline.from... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-04-07T03:02:41Z | 2023-04-09T06:39:08Z | 2023-04-09T06:39:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pure-rgb | 45,315,076 | MDQ6VXNlcjQ1MzE1MDc2 | User | false |
huggingface/diffusers | 1,658,697,481 | I_kwDOHa8MBc5i3bcJ | 3,008 | https://github.com/huggingface/diffusers/issues/3008 | https://api.github.com/repos/huggingface/diffusers/issues/3008 | lpw_stable_diffusion community pipeline passes invalid combination of arguments to torch.randn | ### Describe the bug
When using `lpw_stable_diffusion` pipeline, it passes an invalid combination of arguments into `torch.randn`. Problem still persists while using `cpu`.
### Reproduction
```py
from diffusers import DiffusionPipeline
import torch
pipe = DiffusionPipeline.from_pretrained('Linaqruf/anything... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-04-07T11:18:03Z | 2023-04-21T13:56:05Z | 2023-04-13T06:29:51Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sertdfyguhi | 71,869,572 | MDQ6VXNlcjcxODY5NTcy | User | false |
huggingface/diffusers | 1,658,731,423 | I_kwDOHa8MBc5i3juf | 3,009 | https://github.com/huggingface/diffusers/issues/3009 | https://api.github.com/repos/huggingface/diffusers/issues/3009 | How to improve the training of controlnet | Hi, thanks for the implementation of contorlnet in diffusers [https://github.com/huggingface/diffusers/blob/main/examples/controlnet/train_controlnet.py](url). I have tried the example with batch size 4 on A100 40G but I cannot produce good samples. The following is one of the generated sample
`cyan circle with brow... | closed | completed | false | 3 | [
"stale"
] | [
"yiyixuxu"
] | 2023-04-07T12:00:52Z | 2023-05-15T15:03:17Z | 2023-05-15T15:03:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | garychan22 | 108,175,311 | U_kgDOBnKfzw | User | false |
huggingface/diffusers | 1,658,796,962 | I_kwDOHa8MBc5i3zui | 3,010 | https://github.com/huggingface/diffusers/issues/3010 | https://api.github.com/repos/huggingface/diffusers/issues/3010 | [Docs] Distributed inference with multiple GPUs | Originally posted by @patrickvonplaten in https://github.com/huggingface/diffusers/issues/2977#issuecomment-1498832883
| closed | completed | false | 3 | [] | [
"stevhliu"
] | 2023-04-07T13:16:46Z | 2023-05-19T17:07:34Z | 2023-05-19T17:07:34Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,659,179,967 | I_kwDOHa8MBc5i5RO_ | 3,013 | https://github.com/huggingface/diffusers/issues/3013 | https://api.github.com/repos/huggingface/diffusers/issues/3013 | Replacing text encoder | Hello, I would like to replace the text encoder with a custom one for some experiments. In this case it would be ("cambridgeltl/SapBERT-from-PubMedBERT-fulltext").
How would i go about that when I want to use StableDiffusionPipeline and not LDMTextToImagePipeline or AltDiffusionPipeline?
I tried initializing them ... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-04-07T20:45:20Z | 2023-05-22T15:03:22Z | 2023-05-22T15:03:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | leabrix | 130,184,959 | U_kgDOB8J2_w | User | false |
huggingface/diffusers | 1,659,346,769 | I_kwDOHa8MBc5i559R | 3,015 | https://github.com/huggingface/diffusers/issues/3015 | https://api.github.com/repos/huggingface/diffusers/issues/3015 | Convert ControlNet trained with diffusers into original CompVis format used in WebUI | Hi, I have followed [blog/train-your-controlnet](https://huggingface.co/blog/train-your-controlnet) and got my own ControlNet model in diffusers format. Now I want to convert that into `.pth` format which can be used in [sd-webui-controlnet](https://github.com/Mikubill/sd-webui-controlnet). Anyone helps?
There are c... | closed | completed | false | 4 | [] | [] | 2023-04-08T02:46:32Z | 2023-05-30T17:48:48Z | 2023-04-08T16:36:55Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | huanngzh | 78,398,294 | MDQ6VXNlcjc4Mzk4Mjk0 | User | false |
huggingface/diffusers | 1,659,396,357 | I_kwDOHa8MBc5i6GEF | 3,017 | https://github.com/huggingface/diffusers/issues/3017 | https://api.github.com/repos/huggingface/diffusers/issues/3017 | [Docs] Write one on running how to perform distributed inference using multiple GPUs in Diffusers | Originally initiated here:
https://github.com/huggingface/diffusers/issues/2977 | closed | completed | false | 1 | [] | [
"sayakpaul"
] | 2023-04-08T05:13:28Z | 2023-04-08T05:14:30Z | 2023-04-08T05:14:30Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,659,412,305 | I_kwDOHa8MBc5i6J9R | 3,018 | https://github.com/huggingface/diffusers/issues/3018 | https://api.github.com/repos/huggingface/diffusers/issues/3018 | [Latent Upscaling] Restore noise | **What API design would you like to have changed or added to the library? Why?**
The `StableDiffusionLatentUpscalePipeline` should have the `noise_level` parameter reinstated. It was removed based on the false assumption that noise is undesirable. The author of the original commit said that:
> this step [noise] t... | closed | completed | false | 8 | [
"stale"
] | [] | 2023-04-08T06:14:10Z | 2023-06-24T15:03:06Z | 2023-06-24T15:03:06Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chillpixelfun | 124,188,297 | U_kgDOB2b2iQ | User | false |
huggingface/diffusers | 1,659,425,792 | I_kwDOHa8MBc5i6NQA | 3,019 | https://github.com/huggingface/diffusers/issues/3019 | https://api.github.com/repos/huggingface/diffusers/issues/3019 | segment anything model for text-guided image editing | ### Model/Pipeline/Scheduler description
Use recently release segment anything model (SAM) to support image editing.
Specifically, SAM support click or text to segment the target region, which is used to create mask image for image editing.
Compared with plot-based, its interactive form is more intelligent and conv... | closed | completed | false | 8 | [
"stale"
] | [] | 2023-04-08T07:07:52Z | 2023-08-03T19:05:50Z | 2023-05-29T15:03:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | feizc | 37,614,046 | MDQ6VXNlcjM3NjE0MDQ2 | User | false |
huggingface/diffusers | 1,659,731,966 | I_kwDOHa8MBc5i7X_- | 3,026 | https://github.com/huggingface/diffusers/issues/3026 | https://api.github.com/repos/huggingface/diffusers/issues/3026 | [instruct_pix2pix]: When I try to use accelerate launch train_instruct_pix2pix.py with one gpus, it report the error as below: | ### Describe the bug
When I try to use accelerate launch train_instruct_pix2pix.py with one gpus, it report the error as below:
File "/home/xiangpeng.wan/miniconda3/envs/transformers/lib/python3.8/site-packages/accelerate/utils/dataclasses.py", line 836, in set_auto_wrap_policy
raise Exception("Could not fin... | closed | completed | false | 9 | [
"bug"
] | [
"sayakpaul"
] | 2023-04-09T03:19:57Z | 2025-07-07T06:58:10Z | 2023-04-13T17:43:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | elricwan | 33,402,371 | MDQ6VXNlcjMzNDAyMzcx | User | false |
huggingface/diffusers | 1,659,863,280 | I_kwDOHa8MBc5i74Dw | 3,028 | https://github.com/huggingface/diffusers/issues/3028 | https://api.github.com/repos/huggingface/diffusers/issues/3028 | Cannot import ClipProcessor and other bugs | ### Describe the bug
There is a constant import error:
`ImportError: cannot import name 'CLIPImageProcessor' from 'transformers'
(D:\Python\lib\site-packages\transformers\__init__.py)`
Somehow managed to fix it, but still another problem: no matter what I write, NSFW content is always determined
When disabli... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2023-04-09T12:41:16Z | 2025-07-26T14:10:23Z | 2023-06-13T15:03:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | OlegXio | 125,251,033 | U_kgDOB3ct2Q | User | false |
huggingface/diffusers | 1,659,908,379 | I_kwDOHa8MBc5i8DEb | 3,029 | https://github.com/huggingface/diffusers/issues/3029 | https://api.github.com/repos/huggingface/diffusers/issues/3029 | [Examples] ControlNet (PyTorch) - custom resolution? | On the tutorial example bellow by @pcuenca:
https://github.com/huggingface/diffusers/tree/main/examples/controlnet
https://huggingface.co/blog/train-your-controlnet
Is it possible to train with a custom resolution dataset, ex 1024 x 768?
I tried something like this, but didn't work:
accelerate launch train_con... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-04-09T15:13:39Z | 2023-05-18T15:03:27Z | 2023-05-18T15:03:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alelordelo | 17,708,751 | MDQ6VXNlcjE3NzA4NzUx | User | false |
huggingface/diffusers | 1,660,006,948 | I_kwDOHa8MBc5i8bIk | 3,032 | https://github.com/huggingface/diffusers/issues/3032 | https://api.github.com/repos/huggingface/diffusers/issues/3032 | StableDiffusionControlNetImg2ImgPipeline' has no attribute 'from_pretrained' | ### Describe the bug
https://github.com/huggingface/diffusers/blob/main/examples/community/stable_diffusion_controlnet_img2img.py
unable to launch this pipeline anyhow.
any tips?
### Reproduction
Running an example from the file itself:
```py
import numpy as np
import torch
from PIL import Image
from diffuser... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-04-09T21:19:32Z | 2023-06-25T07:58:51Z | 2023-05-05T00:52:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Pirog17000 | 3,022,852 | MDQ6VXNlcjMwMjI4NTI= | User | false |
huggingface/diffusers | 1,660,010,864 | I_kwDOHa8MBc5i8cFw | 3,033 | https://github.com/huggingface/diffusers/issues/3033 | https://api.github.com/repos/huggingface/diffusers/issues/3033 | may I ask how could i generate one single image with multiple gpu processing at the same time? | null | closed | completed | false | 2 | [
"stale"
] | [] | 2023-04-09T21:36:37Z | 2023-05-18T15:03:26Z | 2023-05-18T15:03:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chikiuso | 79,965 | MDQ6VXNlcjc5OTY1 | User | false |
huggingface/diffusers | 1,660,166,219 | I_kwDOHa8MBc5i9CBL | 3,035 | https://github.com/huggingface/diffusers/issues/3035 | https://api.github.com/repos/huggingface/diffusers/issues/3035 | how to use two placeholder_tokens in textual_inversion? | I separately trained two image sets with two placeholder_tokens, using <*dog> and <@cat>, I'd like to ask how to merge these placeholder_tokens together and generate an image such as "a <*dog> playing with a <@cat>" | closed | completed | false | 6 | [
"stale"
] | [] | 2023-04-10T03:14:14Z | 2023-06-01T15:03:10Z | 2023-06-01T15:03:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yuxu915 | 94,051,337 | U_kgDOBZscCQ | User | false |
huggingface/diffusers | 1,660,320,178 | I_kwDOHa8MBc5i9nmy | 3,036 | https://github.com/huggingface/diffusers/issues/3036 | https://api.github.com/repos/huggingface/diffusers/issues/3036 | [ControlNet] How to reproduce cartoon to realistic photo | How can I reproduce the ControlNet cartoon-> realistic photo example? Since both the content and color are preserved, I do not think the control signal is canny/HED/hough/depth/sketch/pose/norm etc?
https://github.com/huggingface/diffusers/releases
![image](https://user-images.githubusercontent.com/24302931/230841831... | closed | completed | false | 9 | [
"stale"
] | [
"patrickvonplaten"
] | 2023-04-10T06:36:46Z | 2023-08-14T15:04:29Z | 2023-08-14T15:04:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zechenghe | 24,302,931 | MDQ6VXNlcjI0MzAyOTMx | User | false |
huggingface/diffusers | 1,660,833,281 | I_kwDOHa8MBc5i_k4B | 3,038 | https://github.com/huggingface/diffusers/issues/3038 | https://api.github.com/repos/huggingface/diffusers/issues/3038 | Expand documentation for LORA with locally stored models | **Description of LORA fine tuning does not cover all the bases**
I am fairly new to Stable Diffusion, but have some experience using other ML projects, so in general, configuring a dataset and fine tuning is not an issue, however, I run into a block here trying to finetune LORA on a local model, as the `--pretrained_m... | closed | completed | false | 7 | [] | [] | 2023-04-10T14:05:33Z | 2023-04-21T18:03:46Z | 2023-04-21T18:03:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nicemicro | 63,175,285 | MDQ6VXNlcjYzMTc1Mjg1 | User | false |
huggingface/diffusers | 1,660,905,482 | I_kwDOHa8MBc5i_2gK | 3,039 | https://github.com/huggingface/diffusers/issues/3039 | https://api.github.com/repos/huggingface/diffusers/issues/3039 | bfloat16/float32 type mismatch in lax.select in scheduling_pndm_flax.py | ### Describe the bug
I'm following this [tutorial](https://huggingface.co/blog/stable_diffusion_jax) for Flax Stable Diffusion (https://huggingface.co/CompVis/stable-diffusion) on Cloud TPU VM (v4-8).
## Issue with newer version of diffusers
When making inferences with the newest version of diffusers/transformers ... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-04-10T14:56:14Z | 2023-04-13T19:13:30Z | 2023-04-12T13:48:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yeandy | 14,128,880 | MDQ6VXNlcjE0MTI4ODgw | User | false |
huggingface/diffusers | 1,357,770,323 | I_kwDOHa8MBc5Q7e5T | 304 | https://github.com/huggingface/diffusers/issues/304 | https://api.github.com/repos/huggingface/diffusers/issues/304 | Failed to import transformers.models.clip.feature_extraction_clip | ### Describe the bug
On a fresh install (after deleting Stable_Diffusion from my gdrive, and disconnecting and reconnecting for a fresh environment) I get the following error:
```
RuntimeError: Failed to import transformers.models.clip.feature_extraction_clip because of the following error (look up to see its trac... | closed | completed | false | 3 | [
"bug"
] | [] | 2022-08-31T19:30:54Z | 2022-09-20T06:08:12Z | 2022-09-20T06:08:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | WASasquatch | 1,151,589 | MDQ6VXNlcjExNTE1ODk= | User | false |
huggingface/diffusers | 1,660,958,801 | I_kwDOHa8MBc5jADhR | 3,040 | https://github.com/huggingface/diffusers/issues/3040 | https://api.github.com/repos/huggingface/diffusers/issues/3040 | AttributeError: 'list' object has no attribute '__module__' after copy pasting code from official docs | ### Describe the bug
`text_config_dict` is provided which will be used to initialize `CLIPTextConfig`. The value `text_config["id2label"]` will be overriden.
/usr/local/lib/python3.9/dist-packages/transformers/models/clip/feature_extraction_clip.py:28: FutureWarning: The class CLIPFeatureExtractor is deprecated and w... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-04-10T15:34:36Z | 2023-04-11T13:51:02Z | 2023-04-11T13:51:01Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,661,291,451 | I_kwDOHa8MBc5jBUu7 | 3,043 | https://github.com/huggingface/diffusers/issues/3043 | https://api.github.com/repos/huggingface/diffusers/issues/3043 | Add ODISE | ### Model/Pipeline/Scheduler description
Add ODISE, a model capable of zero-shot text-based image segmentation.
Supports text prompts out of the box, unlike Segment Anything. The model is based on the features of the frozen U-Net of Stable Diffusion
### Open source status
- [X] The model implementation is availab... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-04-10T20:06:51Z | 2023-05-20T15:02:45Z | 2023-05-20T15:02:45Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | NielsRogge | 48,327,001 | MDQ6VXNlcjQ4MzI3MDAx | User | false |
huggingface/diffusers | 1,661,306,510 | I_kwDOHa8MBc5jBYaO | 3,044 | https://github.com/huggingface/diffusers/issues/3044 | https://api.github.com/repos/huggingface/diffusers/issues/3044 | conditional Unet1D | **Is your feature request related to a problem? Please describe.**
I would like to work on text conditioned diffusion of 1d signals. There exists a variety of building blocks for conditional Unet2D but not for Unet1D.
**Describe the solution you'd like**
Support for conditional Unet1D (unet_1d_condition.py and res... | closed | completed | false | 18 | [
"stale"
] | [] | 2023-04-10T20:18:04Z | 2025-11-19T20:08:17Z | 2023-05-30T15:03:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lucala | 14,397,281 | MDQ6VXNlcjE0Mzk3Mjgx | User | false |
huggingface/diffusers | 1,661,373,184 | I_kwDOHa8MBc5jBosA | 3,045 | https://github.com/huggingface/diffusers/issues/3045 | https://api.github.com/repos/huggingface/diffusers/issues/3045 | Converting lora .safetensors to diffusers | ### Describe the bug
Hi there,
I'm trying to convert a civitai model to diffusers format using this script in your repository: https://github.com/huggingface/diffusers/blob/main/scripts/convert_lora_safetensor_to_diffusers.py
I'm trying to convert the file like this:
`python convert.py --checkpoint_path /Use... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-04-10T21:11:51Z | 2023-05-30T15:03:34Z | 2023-05-30T15:03:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tentothedollar1 | 130,407,072 | U_kgDOB8XaoA | User | false |
huggingface/diffusers | 1,358,234,962 | I_kwDOHa8MBc5Q9QVS | 305 | https://github.com/huggingface/diffusers/issues/305 | https://api.github.com/repos/huggingface/diffusers/issues/305 | ImportError from 'diffusers' |
[<ipython-input-4-20ae85d6bfb8>](https://localhost:8080/#) <module>
6 import PIL
7
----> 8 from diffusers import StableDiffusionInpaintPipeline
9
10 def download_image(url):
ImportError: cannot import name 'StableDiffusionInpaintPipeline' from 'diffusers' (/usr/local/lib/python3.7/di... | closed | completed | false | 27 | [
"stale"
] | [] | 2022-09-01T04:30:33Z | 2022-11-02T15:04:03Z | 2022-11-02T15:04:03Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xABSTRACTORx | 93,626,586 | U_kgDOBZSg2g | User | false |
huggingface/diffusers | 1,661,755,436 | I_kwDOHa8MBc5jDGAs | 3,050 | https://github.com/huggingface/diffusers/issues/3050 | https://api.github.com/repos/huggingface/diffusers/issues/3050 | What does "encoder_hidden_states" used for in DiT examples? | null | closed | completed | false | 3 | [
"stale"
] | [] | 2023-04-11T04:36:31Z | 2023-05-20T15:02:43Z | 2023-05-20T15:02:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JincanDeng | 26,006,156 | MDQ6VXNlcjI2MDA2MTU2 | User | false |
huggingface/diffusers | 1,661,760,775 | I_kwDOHa8MBc5jDHUH | 3,051 | https://github.com/huggingface/diffusers/issues/3051 | https://api.github.com/repos/huggingface/diffusers/issues/3051 | 👾 8-bit model converter | We start putting multiple models in a vram at the same time like this issue https://github.com/huggingface/diffusers/issues/2898 😐 If it is possible we need 8-bit model converter
like this for diffusers
4 bits quantization of [LLaMA](https://arxiv.org/abs/2302.13971) using [GPTQ](https://arxiv.org/abs/2210.17323)
... | closed | completed | false | 4 | [] | [] | 2023-04-11T04:44:33Z | 2023-04-12T17:30:02Z | 2023-04-12T16:46:59Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | camenduru | 54,370,274 | MDQ6VXNlcjU0MzcwMjc0 | User | false |
huggingface/diffusers | 1,661,789,445 | I_kwDOHa8MBc5jDOUF | 3,052 | https://github.com/huggingface/diffusers/issues/3052 | https://api.github.com/repos/huggingface/diffusers/issues/3052 | control for versatile diffusion | Hi, in original repo https://github.com/SHI-Labs/Versatile-Diffusion there is fidelity level and focus level to control the generated image variation. In diffusers pipeline, is there a similar parameter for that?. Thanks | closed | completed | false | 3 | [
"stale"
] | [] | 2023-04-11T05:21:23Z | 2023-05-20T15:02:42Z | 2023-05-20T15:02:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pribadihcr | 13,068,286 | MDQ6VXNlcjEzMDY4Mjg2 | User | false |
huggingface/diffusers | 1,661,960,873 | I_kwDOHa8MBc5jD4Kp | 3,055 | https://github.com/huggingface/diffusers/issues/3055 | https://api.github.com/repos/huggingface/diffusers/issues/3055 | How do I use multicontrolnet? The example in the docs doesn't work | Hi. I've been trying to use multicontrolnet by replicating the second example from the blog: https://huggingface.co/docs/diffusers/main/en/api/pipelines/stable_diffusion/controlnet
but I keep getting this error:
AttributeError: 'list' object has no attribute '__module__'
I believe that it gets confused when I pa... | closed | completed | false | 1 | [] | [] | 2023-04-11T07:57:47Z | 2023-04-11T09:58:02Z | 2023-04-11T09:58:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,661,977,621 | I_kwDOHa8MBc5jD8QV | 3,056 | https://github.com/huggingface/diffusers/issues/3056 | https://api.github.com/repos/huggingface/diffusers/issues/3056 | Why always --fp32 precision in schedulers? | Can anybody clarify why do we always use `--fp32` precision in schedulers such as [here] ?(https://github.com/huggingface/diffusers/blob/4f48476dd6336f35489378cf38c0852a48f92289/src/diffusers/schedulers/scheduling_pndm.py#L112).
So when I load the model with `--fp16` precision, like this here below, it doesnt affect ... | closed | completed | false | 2 | [] | [] | 2023-04-11T08:08:32Z | 2023-04-27T02:46:25Z | 2023-04-27T02:45:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | himmetozcan | 44,242,024 | MDQ6VXNlcjQ0MjQyMDI0 | User | false |
huggingface/diffusers | 1,662,296,997 | I_kwDOHa8MBc5jFKOl | 3,057 | https://github.com/huggingface/diffusers/issues/3057 | https://api.github.com/repos/huggingface/diffusers/issues/3057 | conversion scripts fails to import from diffusers.pipelines.stable_diffusion.convert_from_ckpt | https://github.com/huggingface/diffusers/blob/8b451eb63b0f101e7fcc72365fe0d683808b22cd/scripts/convert_original_stable_diffusion_to_diffusers.py#L21
function seems to be renamed to `load_pipeline_from_original_stable_diffusion_ckpt`
Also, the `--half` flag is not working. | closed | completed | false | 5 | [] | [] | 2023-04-11T11:37:35Z | 2023-10-23T09:21:54Z | 2023-04-12T12:30:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rvorias | 21,060,408 | MDQ6VXNlcjIxMDYwNDA4 | User | false |
huggingface/diffusers | 1,662,555,760 | I_kwDOHa8MBc5jGJZw | 3,059 | https://github.com/huggingface/diffusers/issues/3059 | https://api.github.com/repos/huggingface/diffusers/issues/3059 | Trying to load my Lora but it doesn't work for some reason | ```
import os
files = os.listdir('.')
for file in files:
print(file)
if 'mngstle.safetensors' in files:
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(333)
pipe.unet.load_attn_procs('mngstle.safetensors').to(device)
# Use the loaded UNet module here
else:
... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-04-11T14:00:28Z | 2023-05-28T15:03:11Z | 2023-05-28T15:03:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,662,877,905 | I_kwDOHa8MBc5jHYDR | 3,061 | https://github.com/huggingface/diffusers/issues/3061 | https://api.github.com/repos/huggingface/diffusers/issues/3061 | Stable Diffusion pipeline reproducibility issue | ### Describe the bug
Making `StableDiffusionPipeline` inference two times in a row with fixed seed gives different results.
The source of difference seems to come from the `AttnProcessor2_0` class, as the issue dissapears when using `CrossAttnProcessor` instead of `AttnProcessor2_0`.
### Reproduction
```pyth... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-04-11T16:43:17Z | 2023-04-21T17:38:36Z | 2023-04-21T17:38:36Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | 19and99 | 48,154,088 | MDQ6VXNlcjQ4MTU0MDg4 | User | false |
huggingface/diffusers | 1,663,517,588 | I_kwDOHa8MBc5jJ0OU | 3,062 | https://github.com/huggingface/diffusers/issues/3062 | https://api.github.com/repos/huggingface/diffusers/issues/3062 | Is this docstring of `encoder_hidden_states` argument in `Transformer2DModel` correct? | https://github.com/huggingface/diffusers/blob/e607a582cfaa7dfaf7913fc3bb54c35eceee583c/src/diffusers/models/transformer_2d.py#L228-L230
The shape is `(batch_size, seq_length, embedding_dim)`, isn't it? Also it's supposed to be float tensor. Or maybe it supports both long and float? | closed | completed | false | 2 | [] | [] | 2023-04-12T01:39:43Z | 2023-04-19T16:52:00Z | 2023-04-19T16:52:00Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | offchan42 | 15,215,732 | MDQ6VXNlcjE1MjE1NzMy | User | false |
huggingface/diffusers | 1,663,810,602 | I_kwDOHa8MBc5jK7wq | 3,064 | https://github.com/huggingface/diffusers/issues/3064 | https://api.github.com/repos/huggingface/diffusers/issues/3064 | Loading .safetensors Lora | ### Describe the bug
I have downloaded lora from civitai which is in .safetensor format.
When i load it using below code,
pipe = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16)
pipe.unet.load_attn_procs("lora.safetensors")
It throws error : KeyError: 'to_k... | closed | completed | false | 85 | [
"bug",
"stale"
] | [] | 2023-04-12T05:21:35Z | 2023-12-19T02:58:31Z | 2023-06-16T23:32:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | adhikjoshi | 11,740,719 | MDQ6VXNlcjExNzQwNzE5 | User | false |
huggingface/diffusers | 1,663,971,182 | I_kwDOHa8MBc5jLi9u | 3,065 | https://github.com/huggingface/diffusers/issues/3065 | https://api.github.com/repos/huggingface/diffusers/issues/3065 | [Examples - DreamBooth] Add an option to tune the text encoder with LoRA | We recently added support for fine-tuning the text encoder with LoRA here: https://github.com/huggingface/diffusers/pull/2918.
Next up, we need to modify our [DreamBooth training example](https://github.com/huggingface/diffusers/tree/main/examples/dreambooth) to incorporate this support. | closed | completed | false | 1 | [] | [
"sayakpaul"
] | 2023-04-12T07:04:38Z | 2023-04-20T11:55:19Z | 2023-04-20T11:55:19Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,358,365,928 | I_kwDOHa8MBc5Q9wTo | 307 | https://github.com/huggingface/diffusers/issues/307 | https://api.github.com/repos/huggingface/diffusers/issues/307 | Merge Stable Diffusion pipelines into one | **Is your feature request related to a problem? Please describe.**
It would be nice to be able to use text2img & img2img & inpaint with the same instanced pipeline
**Describe the solution you'd like**
`pipe.generate(...)`
`pipe.image_to_image(...)`
`pipe.inpaint(...)`
**Describe alternatives you've considered... | closed | completed | false | 10 | [
"community-examples"
] | [] | 2022-09-01T07:11:54Z | 2022-09-12T12:28:26Z | 2022-09-01T11:47:44Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | python273 | 3,097,956 | MDQ6VXNlcjMwOTc5NTY= | User | false |
huggingface/diffusers | 1,664,646,387 | I_kwDOHa8MBc5jOHzz | 3,071 | https://github.com/huggingface/diffusers/issues/3071 | https://api.github.com/repos/huggingface/diffusers/issues/3071 | Support "SVDiff: Compact Parameter Space for Diffusion Fine-Tuning" | ## Model/Pipeline/Scheduler description
### Summary
[SVDiff: Compact Parameter Space for Diffusion Fine-Tuning](https://arxiv.org/abs/2303.11305) is an effective domain-tuning method for text-to-image models like LoRA but the number of trainable extra parameters is less than LoRA and the result is better or compati... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-04-12T14:08:43Z | 2023-06-10T15:03:08Z | 2023-06-10T15:03:08Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mkshing | 33,302,880 | MDQ6VXNlcjMzMzAyODgw | User | false |
huggingface/diffusers | 1,664,806,773 | I_kwDOHa8MBc5jOu91 | 3,073 | https://github.com/huggingface/diffusers/issues/3073 | https://api.github.com/repos/huggingface/diffusers/issues/3073 | [Request] Add Consistency models | ### Model/Pipeline/Scheduler description
OpenAI recently announced "Consistency" models - a new family of generative models that support fast one-step generation (without adversarial training).
According to the paper, "Consistency models are heavily inspired by the theory of continuous-time) diffusion models", so i... | closed | completed | false | 13 | [
"New pipeline/model",
"Good second issue"
] | [] | 2023-04-12T15:40:46Z | 2023-07-07T17:32:59Z | 2023-07-07T17:32:59Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xenova | 26,504,141 | MDQ6VXNlcjI2NTA0MTQx | User | false |
huggingface/diffusers | 1,664,993,518 | I_kwDOHa8MBc5jPcju | 3,075 | https://github.com/huggingface/diffusers/issues/3075 | https://api.github.com/repos/huggingface/diffusers/issues/3075 | Create a Video ControlNet Pipeline | **Is your feature request related to a problem? Please describe.**
Stable Diffusion video generation lacks precise movement control and composition control. This is not surprising, since the model was not trained or fine-tuned with videos.
**Describe the solution you'd like**
By following an analogous extension p... | closed | completed | false | 3 | [
"question"
] | [] | 2023-04-12T17:51:35Z | 2023-04-13T16:21:28Z | 2023-04-13T16:21:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jfischoff | 168,507 | MDQ6VXNlcjE2ODUwNw== | User | false |
huggingface/diffusers | 1,665,344,195 | I_kwDOHa8MBc5jQyLD | 3,081 | https://github.com/huggingface/diffusers/issues/3081 | https://api.github.com/repos/huggingface/diffusers/issues/3081 | Textual Inversion Colab is non-functional | ### Describe the bug
The "Create our Training Data" cell fails every time
### Reproduction
https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/sd_textual_inversion_training.ipynb
### Logs
```shell
╭─────────────────────────────── Traceback (most recent call last) ────────────────────... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-04-12T22:21:00Z | 2023-05-22T15:03:12Z | 2023-05-22T15:03:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | whatlurks | 107,269,235 | U_kgDOBmTMcw | User | false |
huggingface/diffusers | 1,665,347,006 | I_kwDOHa8MBc5jQy2- | 3,082 | https://github.com/huggingface/diffusers/issues/3082 | https://api.github.com/repos/huggingface/diffusers/issues/3082 | PFGM++ | ### Model/Pipeline/Scheduler description
The PFGM++ (https://github.com/Newbeeer/pfgmpp) is a new type of generative models subsuming PFGM / diffusion models. It's more robust than diffusion models, easier to train than PFGM, and achieve SOTA on several standard conditional/unconditional image generation benchmarks.... | closed | not_planned | false | 1 | [
"New pipeline/model",
"Good second issue"
] | [] | 2023-04-12T22:24:35Z | 2024-11-24T03:25:18Z | 2024-11-24T03:25:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Newbeeer | 33,647,781 | MDQ6VXNlcjMzNjQ3Nzgx | User | false |
huggingface/diffusers | 1,665,762,185 | I_kwDOHa8MBc5jSYOJ | 3,085 | https://github.com/huggingface/diffusers/issues/3085 | https://api.github.com/repos/huggingface/diffusers/issues/3085 | Return cache folder removed as argument in 0.15.0 | ### Describe the bug
`return_cached_folder` argument removed from `from_pretrained`
### Reproduction
```py
from diffusers import DiffusionPipeline
import torch
pipe, cache_folder = DiffusionPipeline.from_pretrained(
"runwayml/stable-diffusion-v1-5",
torch_dtype=torch.float16,
... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-04-13T05:59:13Z | 2023-04-13T12:33:13Z | 2023-04-13T12:33:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | laxuscullen | 49,696,073 | MDQ6VXNlcjQ5Njk2MDcz | User | false |
huggingface/diffusers | 1,665,809,537 | I_kwDOHa8MBc5jSjyB | 3,086 | https://github.com/huggingface/diffusers/issues/3086 | https://api.github.com/repos/huggingface/diffusers/issues/3086 | input audio output image | **Is your feature request related to a problem? Please describe.**
Can I input audio output image
Not speech recognition to text, then text as input
I means, For example, the input audio may be pop music, ambient sound, the output is the image cover of this audio
**Describe the solution you'd like**
support inpu... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-04-13T06:36:03Z | 2023-05-22T15:03:11Z | 2023-05-22T15:03:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | newcworld | 71,140,907 | MDQ6VXNlcjcxMTQwOTA3 | User | false |
huggingface/diffusers | 1,665,835,614 | I_kwDOHa8MBc5jSqJe | 3,087 | https://github.com/huggingface/diffusers/issues/3087 | https://api.github.com/repos/huggingface/diffusers/issues/3087 | Support of the LyCORIS (LoCon/LoHA) models | ### Model/Pipeline/Scheduler description
Hi everyone ! Thanks for your amazing work !
Some specific (optimized) version of Lora are developed (https://github.com/KohakuBlueleaf/LyCORIS) and are available (https://civitai.com/models/37053/michael-jordan) with pretty cool features (ability to make distinctions betwee... | closed | completed | false | 49 | [
"New pipeline/model",
"Good second issue"
] | [] | 2023-04-13T06:58:23Z | 2024-10-18T00:50:39Z | 2024-01-31T01:22:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mkhennoussi | 10,024,845 | MDQ6VXNlcjEwMDI0ODQ1 | User | false |
huggingface/diffusers | 1,665,904,198 | I_kwDOHa8MBc5jS65G | 3,089 | https://github.com/huggingface/diffusers/issues/3089 | https://api.github.com/repos/huggingface/diffusers/issues/3089 | bug in ddpm scheduler | ### Describe the bug
forget to convert self.alphas_cumprod to the same device with samples
https://github.com/huggingface/diffusers/blob/46c52f9b9607e6ecb29c782c052aea313e6487b7/src/diffusers/schedulers/scheduling_ddpm.py#L326, which will cause calculation error.
suggest add line self.alphas_cumprod = self.alphas... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-04-13T07:49:55Z | 2023-04-28T02:00:44Z | 2023-04-28T02:00:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LSQsjtu | 59,142,188 | MDQ6VXNlcjU5MTQyMTg4 | User | false |
huggingface/diffusers | 1,666,050,885 | I_kwDOHa8MBc5jTetF | 3,090 | https://github.com/huggingface/diffusers/issues/3090 | https://api.github.com/repos/huggingface/diffusers/issues/3090 | Can not load pretrained model properly | **Is your feature request related to a problem? Please describe.**
I use `pipeline.save_pretrained(config.output_dir, variant=str(epoch)) ` to save model every n epochs, this leads to there are multiple model in the unet folder:
![image](https://user-images.githubusercontent.com/43233772/231714552-521fcad7-b94b-4b57-... | closed | completed | false | 3 | [] | [] | 2023-04-13T09:23:39Z | 2024-11-17T09:48:02Z | 2023-04-13T14:28:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wwdok | 43,233,772 | MDQ6VXNlcjQzMjMzNzcy | User | false |
huggingface/diffusers | 1,666,122,067 | I_kwDOHa8MBc5jTwFT | 3,091 | https://github.com/huggingface/diffusers/issues/3091 | https://api.github.com/repos/huggingface/diffusers/issues/3091 | Text to audio with default code produces an unplayable file | ### Describe the bug
Great with the new Text to audio feature. However, when I use the default script, the resulting file doesn't contain any audio, there are no errors printed in the console:
The file properties looks ok:
![image](https://user-images.githubusercontent.com/1322593/231726379-5d14fbee-2b80-49f0-925... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-04-13T10:08:29Z | 2023-04-21T18:24:39Z | 2023-04-21T18:24:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 1,666,534,725 | I_kwDOHa8MBc5jVU1F | 3,093 | https://github.com/huggingface/diffusers/issues/3093 | https://api.github.com/repos/huggingface/diffusers/issues/3093 | Is it possible to modify the noise directly? | **Is your feature request related to a problem? Please describe.**
I want to do seed interpolation animation.
**Describe the solution you'd like**
We can already specify the seed as a 'generator', the next step would be to pass the noise directly instead. Then I can generate two noise tensors and interpolate them ... | closed | completed | false | 2 | [] | [] | 2023-04-13T14:26:25Z | 2023-04-17T12:31:35Z | 2023-04-17T12:31:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | oxysoft | 3,945,277 | MDQ6VXNlcjM5NDUyNzc= | User | false |
huggingface/diffusers | 1,666,647,623 | I_kwDOHa8MBc5jVwZH | 3,094 | https://github.com/huggingface/diffusers/issues/3094 | https://api.github.com/repos/huggingface/diffusers/issues/3094 | cuda out of memory when finetune pokemon example in train_text_to_image.py | this is my bash script, i just add enable_xformers_memory_efficient_attention compared to the original
```
export MODEL_NAME="CompVis/stable-diffusion-v1-4"
export dataset_name="lambdalabs/pokemon-blip-captions"
accelerate launch --mixed_precision="fp16" train_text_to_image.py \
--pretrained_model_name_or_pat... | closed | completed | false | 11 | [] | [] | 2023-04-13T15:25:42Z | 2023-04-24T02:28:02Z | 2023-04-24T02:28:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dora-lemon | 55,336,873 | MDQ6VXNlcjU1MzM2ODcz | User | false |
huggingface/diffusers | 1,666,769,626 | I_kwDOHa8MBc5jWOLa | 3,095 | https://github.com/huggingface/diffusers/issues/3095 | https://api.github.com/repos/huggingface/diffusers/issues/3095 | ControlNet v1.1 | Today, [ControlNet v1.1 was released](https://github.com/lllyasviel/ControlNet-v1-1-nightly). As for the current situation, it seems to be positioned as a preview, and they are particularly working on improving the annotator (image preprocessing) code. It is said that most of the model weights are already production-re... | closed | completed | false | 35 | [] | [] | 2023-04-13T16:47:12Z | 2023-07-13T06:59:47Z | 2023-04-15T03:22:44Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | takuma104 | 10,776 | MDQ6VXNlcjEwNzc2 | User | false |
huggingface/diffusers | 1,666,785,609 | I_kwDOHa8MBc5jWSFJ | 3,096 | https://github.com/huggingface/diffusers/issues/3096 | https://api.github.com/repos/huggingface/diffusers/issues/3096 | Suggestion: Bump the major release number when releasing a compatibility-breaking version | The InvokeAI team had a bit of a surprise yesterday when diffusers moved from 0.14 to 0.15 and changed the names of several documented methods that we were using. Because we had pinned diffusers using the "compatible release" indicator (`diffusers ~= 0.14`) we did not expect to get an upstream update to an incompatible... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-04-13T16:57:47Z | 2023-05-22T15:03:09Z | 2023-05-22T15:03:09Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lstein | 111,189 | MDQ6VXNlcjExMTE4OQ== | User | false |
huggingface/diffusers | 1,667,109,806 | I_kwDOHa8MBc5jXhOu | 3,097 | https://github.com/huggingface/diffusers/issues/3097 | https://api.github.com/repos/huggingface/diffusers/issues/3097 | Using load_textual_inversion appears to do nothing | ### Describe the bug
When using `load_textual_inversion` it does not affect inference in any way. I provided a version of the modified [sample code from the docs](https://huggingface.co/docs/diffusers/main/en/api/loaders#diffusers.TextualInversionLoaderMixin.load_textual_inversion). I am using the embedding from [Ci... | closed | completed | false | 10 | [
"bug",
"stale"
] | [] | 2023-04-13T20:52:07Z | 2023-05-17T15:51:20Z | 2023-05-17T15:51:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sakistriker | 115,505,099 | U_kgDOBuJ3yw | User | false |
huggingface/diffusers | 1,667,428,819 | I_kwDOHa8MBc5jYvHT | 3,098 | https://github.com/huggingface/diffusers/issues/3098 | https://api.github.com/repos/huggingface/diffusers/issues/3098 | Support assigning different optimizers with weights to multiple models during deepspeed.initialize in Accelerate | **Is your feature request related to a problem? Please describe.**
Currently, training multiple models with Deepspeed's ZeRO optimization does not work as intended. A detailed issue has been reported and can be found here: https://github.com/huggingface/transformers/issues/22705
sp
**Describe the solution you'd ... | closed | completed | false | 3 | [] | [] | 2023-04-14T02:56:03Z | 2023-04-17T13:13:03Z | 2023-04-17T13:13:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | uygnef | 13,539,441 | MDQ6VXNlcjEzNTM5NDQx | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.