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 | 1,516,825,221 | I_kwDOHa8MBc5aaOqF | 1,892 | https://github.com/huggingface/diffusers/issues/1892 | https://api.github.com/repos/huggingface/diffusers/issues/1892 | Memory-efficient attention (without xformers) | I implemented sub-quadratic attention (as described in https://arxiv.org/abs/2112.05682v2):
https://twitter.com/Birchlabs/status/1607503573906063362
https://github.com/Birch-san/diffusers/pull/1
https://github.com/Birch-san/diffusers-play/commit/a573e3d9ea4fdacfdee7ddd5eecdac29b236fc00
is this worth upstrea... | open | reopened | false | 31 | [
"wip"
] | [] | 2023-01-03T00:50:10Z | 2026-01-29T16:10:32Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Birch-san | 6,141,784 | MDQ6VXNlcjYxNDE3ODQ= | User | false |
huggingface/diffusers | 1,517,098,638 | I_kwDOHa8MBc5abRaO | 1,895 | https://github.com/huggingface/diffusers/issues/1895 | https://api.github.com/repos/huggingface/diffusers/issues/1895 | EMA model is not updated properly with multi-gpu configuration | ### Describe the bug
The training code for unconditional image generation at
```
examples/unconditional_image_generation/train_unconditional.py
```
is not working properly when it is trained via `accelerate` with multi-gpu training.
Specifically, the EMAModel is not updated while it works without `accelerate`. ... | closed | completed | false | 2 | [
"bug",
"stale"
] | [
"anton-l"
] | 2023-01-03T08:59:48Z | 2023-02-03T17:46:24Z | 2023-02-03T17:46:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | SuwoongHeo | 30,944,168 | MDQ6VXNlcjMwOTQ0MTY4 | User | false |
huggingface/diffusers | 1,517,182,327 | I_kwDOHa8MBc5abl13 | 1,896 | https://github.com/huggingface/diffusers/issues/1896 | https://api.github.com/repos/huggingface/diffusers/issues/1896 | Running CLIP Guided Stable Diffusion fails with Size should be int or sequence. Got <class 'dict'> | ### Describe the bug
I'm trying to run the example of this pipeline [examples/community#clip-guided-stable-diffusion](https://github.com/huggingface/diffusers/tree/main/examples/community#clip-guided-stable-diffusion). But this seems to fails with following error
```python
TypeError: Size should be int or sequence... | closed | completed | false | 1 | [
"bug"
] | [
"patil-suraj"
] | 2023-01-03T10:15:02Z | 2023-01-11T09:52:50Z | 2023-01-11T09:52:50Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dzlab | 1,645,304 | MDQ6VXNlcjE2NDUzMDQ= | User | false |
huggingface/diffusers | 1,517,242,069 | I_kwDOHa8MBc5ab0bV | 1,899 | https://github.com/huggingface/diffusers/issues/1899 | https://api.github.com/repos/huggingface/diffusers/issues/1899 | generation of multi person in one image is bad. | I train two person at the same time. one is 'a sks woman', the other is 'a zwx man'.
It is ok when using single person, for instance 'a sks woman'. But when I want to create photo of couple, using prompt 'a sks woman and a zwx man', it came bad generation, most time only one person is just so so, the other one is al... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-01-03T11:09:51Z | 2023-02-10T15:03:14Z | 2023-02-10T15:03:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tengshaofeng | 9,246,739 | MDQ6VXNlcjkyNDY3Mzk= | User | false |
huggingface/diffusers | 1,517,592,309 | I_kwDOHa8MBc5adJ71 | 1,904 | https://github.com/huggingface/diffusers/issues/1904 | https://api.github.com/repos/huggingface/diffusers/issues/1904 | Load Automatic1111 trained embeddings file to HF | Is it possible to load the Automatic1111 trained embeddings file to HF diffusers?
For example: https://huggingface.co/joachimsallstrom/Double-Exposure-Embedding
@patil-suraj @patrickvonplaten @anton-l | closed | completed | false | 18 | [
"stale"
] | [
"patil-suraj"
] | 2023-01-03T15:56:05Z | 2023-02-17T15:03:27Z | 2023-02-17T15:03:27Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kamalkraj | 17,096,858 | MDQ6VXNlcjE3MDk2ODU4 | User | false |
huggingface/diffusers | 1,517,629,468 | I_kwDOHa8MBc5adTAc | 1,905 | https://github.com/huggingface/diffusers/issues/1905 | https://api.github.com/repos/huggingface/diffusers/issues/1905 | StableDiffusionPipeline not freeing all GPU memory | ### Describe the bug
I have a simple inference server that upon request load a stable diffusion model, run the inference, then returns the images and clears all the memory cache.
In theory, the GPU usage should go back to 0% between each request, but in practice, after the first request, the GPU memory usage stays ... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-01-03T16:23:40Z | 2023-02-12T15:02:44Z | 2023-02-12T15:02:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tzvc | 14,275,989 | MDQ6VXNlcjE0Mjc1OTg5 | User | false |
huggingface/diffusers | 1,518,395,806 | I_kwDOHa8MBc5agOGe | 1,909 | https://github.com/huggingface/diffusers/issues/1909 | https://api.github.com/repos/huggingface/diffusers/issues/1909 | pipelines take a list of torch generators but schedulers only accept a single generator | ### Describe the bug
We have code to handle a list of generators in pipeline here
https://github.com/huggingface/diffusers/blob/d67c30512048e667cbf10c298a04c73f4eb5027c/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py#L392-L419
However, the `generator` argument is passed directly to the sch... | closed | completed | false | 3 | [
"bug"
] | [
"williamberman"
] | 2023-01-04T06:24:31Z | 2023-01-13T11:15:13Z | 2023-01-13T11:15:12Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,518,650,083 | I_kwDOHa8MBc5ahMLj | 1,911 | https://github.com/huggingface/diffusers/issues/1911 | https://api.github.com/repos/huggingface/diffusers/issues/1911 | use. pipline.from_pretrained error, It was ok until now, whats happen? | requests.exceptions.ConnectTimeout: HTTPSConnectionPool(host='s3-proxy.huggingface.tech', port=443): Max retries exceeded with url: /lfs.huggingface.co/repos/6b/20/6b201da5f0f5c60524535ebb7deac2eef68605655d3bbacfee9cce0087f3b3f5/770a47a9ffdcfda0b05506a7888ed714d06131d60267e6cf52765d61cf59fd67?X-Amz-Algorithm=AWS4-HMAC-... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-01-04T09:43:58Z | 2023-02-17T15:03:26Z | 2023-02-17T15:03:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ssxxx1a | 26,184,253 | MDQ6VXNlcjI2MTg0MjUz | User | false |
huggingface/diffusers | 1,519,389,578 | I_kwDOHa8MBc5akAuK | 1,915 | https://github.com/huggingface/diffusers/issues/1915 | https://api.github.com/repos/huggingface/diffusers/issues/1915 | Unify and upload hub weights for unclip text to image and image variation | **Is your feature request related to a problem? Please describe.**
We don't currently have a canonical hub upload under the kakaobrain org for the image variation weight i.e. the tests currently use `fusing/karlo-image-variations-diffusers`.
**Describe the solution you'd like**
Update the https://huggingface.co/ka... | closed | completed | false | 6 | [] | [
"williamberman"
] | 2023-01-04T18:15:31Z | 2023-01-31T18:11:56Z | 2023-01-31T18:11:56Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,519,487,147 | I_kwDOHa8MBc5akYir | 1,917 | https://github.com/huggingface/diffusers/issues/1917 | https://api.github.com/repos/huggingface/diffusers/issues/1917 | Add additional integration testing around training scripts i.e. dreambooth | **Is your feature request related to a problem? Please describe.**
We have some integration tests for testing training scripts. These were added pre-dreambooth, so we could add some more!
Existing tests: https://github.com/huggingface/diffusers/blob/main/examples/test_examples.py
CI Call: https://github.com/huggi... | closed | completed | false | 6 | [
"stale",
"training"
] | [
"williamberman"
] | 2023-01-04T19:45:06Z | 2023-03-21T18:01:48Z | 2023-03-17T15:04:06Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,519,709,511 | I_kwDOHa8MBc5alO1H | 1,918 | https://github.com/huggingface/diffusers/issues/1918 | https://api.github.com/repos/huggingface/diffusers/issues/1918 | issue with DDIM scheduler for google/ddpm-celebahq-256 example | ### Describe the bug
Referring to the example "Unconditional Diffusion with discrete scheduler" at https://github.com/huggingface/diffusers#running-code
The comment says you can use `DDIMPipeline` but this fails to run on
(got an unexpected keyword argument `eta`).
Also, the `PNDMPipeline` runs but just returns n... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-01-04T23:08:08Z | 2023-01-16T14:35:23Z | 2023-01-16T14:35:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | murphyk | 4,632,336 | MDQ6VXNlcjQ2MzIzMzY= | User | false |
huggingface/diffusers | 1,519,776,537 | I_kwDOHa8MBc5alfMZ | 1,919 | https://github.com/huggingface/diffusers/issues/1919 | https://api.github.com/repos/huggingface/diffusers/issues/1919 | dreambooth example missing datasets from requirements.txt | ### Describe the bug
In the example https://github.com/huggingface/diffusers/tree/main/examples/dreambooth#dog-toy-example,
the script fails since `datasets` is missing from requirements.txt
### Reproduction
```
git clone https://github.com/huggingface/diffusers
cd diffusers
pip install -e .
cd examples/drea... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-01-05T00:23:18Z | 2023-01-07T03:35:44Z | 2023-01-07T03:35:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | murphyk | 4,632,336 | MDQ6VXNlcjQ2MzIzMzY= | User | false |
huggingface/diffusers | 1,519,850,023 | I_kwDOHa8MBc5alxIn | 1,921 | https://github.com/huggingface/diffusers/issues/1921 | https://api.github.com/repos/huggingface/diffusers/issues/1921 | How to finetune inpainting model for object removal? What is the input prompt for object removal for both training and inference? |
Hi team,
Thanks for your great work!
I am trying to get object removal functionality from inpainting of SD.
How to finetune inpainting model for object removal?
What is the input prompt for object removal for both training and inference?
Thanks | closed | completed | false | 4 | [
"stale"
] | [] | 2023-01-05T01:23:20Z | 2023-04-03T14:50:38Z | 2023-02-12T15:02:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hdjsjyl | 7,590,139 | MDQ6VXNlcjc1OTAxMzk= | User | false |
huggingface/diffusers | 1,519,855,306 | I_kwDOHa8MBc5alybK | 1,922 | https://github.com/huggingface/diffusers/issues/1922 | https://api.github.com/repos/huggingface/diffusers/issues/1922 | diffusers_training_example.ipynb uses `init_git_repo` which is not defined | ### Describe the bug
https://colab.sandbox.google.com/github/huggingface/notebooks/blob/main/diffusers/training_example.ipynb#scrollTo=b11ba8b7-eb8f-4e8e-88ae-6e4a2b68433e
contains the line
```
from diffusers.hub_utils import init_git_repo, push_to_hub
```
which fails ('init_git_repo` not defined).
[Also, `c... | closed | completed | false | 1 | [
"bug"
] | [
"anton-l"
] | 2023-01-05T01:26:19Z | 2023-01-17T10:18:21Z | 2023-01-17T10:18:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | murphyk | 4,632,336 | MDQ6VXNlcjQ2MzIzMzY= | User | false |
huggingface/diffusers | 1,520,271,236 | I_kwDOHa8MBc5anX-E | 1,923 | https://github.com/huggingface/diffusers/issues/1923 | https://api.github.com/repos/huggingface/diffusers/issues/1923 | Missing CUDA libraries when using bitsandbyes with Dockerfile | ### Describe the bug
I am trying to run diffusers in Docker. I am using the diffusers-pytorch-cuda Dockerfile provided in the repo. The only change I made is to add diffusers and bitsandbytes to the pip install list in the Dockerfile.
When I enable --use_8bit_adam for Dreambooth training, I get this error:
000... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-01-05T07:57:36Z | 2023-02-12T15:02:42Z | 2023-02-12T15:02:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tobrien6 | 449,169 | MDQ6VXNlcjQ0OTE2OQ== | User | false |
huggingface/diffusers | 1,520,376,929 | I_kwDOHa8MBc5anxxh | 1,925 | https://github.com/huggingface/diffusers/issues/1925 | https://api.github.com/repos/huggingface/diffusers/issues/1925 | stable-diffusion-2-inpainting half precision not working | ### Describe the bug
I'm trying to use stabilityai/stable-diffusion-2-inpainting but it gives an error due to half precision `RuntimeError: "LayerNormKernelImpl" not implemented for 'Half'`
### Reproduction
I'm using the example described in the [guide](https://huggingface.co/stabilityai/stable-diffusion-2-inp... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-01-05T09:12:42Z | 2023-04-28T10:29:25Z | 2023-02-12T15:02:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Actis92 | 46,601,193 | MDQ6VXNlcjQ2NjAxMTkz | User | false |
huggingface/diffusers | 1,520,846,129 | I_kwDOHa8MBc5apkUx | 1,928 | https://github.com/huggingface/diffusers/issues/1928 | https://api.github.com/repos/huggingface/diffusers/issues/1928 | Converting SafeTensor files to diffusers | There's a way to convert ckpt files to diffusers, but do not seem to see a way to do so for SafeTensor files.
Additionally, is there a way to run just the SafeTensor files by itself? If it needs additional files, maybe we could introduce a way to apply some default files off of them. I am assuming that most current ... | closed | completed | false | 5 | [] | [] | 2023-01-05T14:32:37Z | 2023-02-11T17:53:04Z | 2023-02-11T17:53:04Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dblunk88 | 39,381,389 | MDQ6VXNlcjM5MzgxMzg5 | User | false |
huggingface/diffusers | 1,520,868,579 | I_kwDOHa8MBc5appzj | 1,929 | https://github.com/huggingface/diffusers/issues/1929 | https://api.github.com/repos/huggingface/diffusers/issues/1929 | CLIP Guided Image2Image | The CLIP guided text2img is pretty good. Would be nice to be able to use it for img2img | closed | completed | false | 15 | [
"community-examples",
"Good second issue"
] | [
"akshay-babbar"
] | 2023-01-05T14:48:06Z | 2023-03-21T22:58:12Z | 2023-03-21T22:58:12Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dblunk88 | 39,381,389 | MDQ6VXNlcjM5MzgxMzg5 | User | false |
huggingface/diffusers | 1,341,326,907 | I_kwDOHa8MBc5P8wY7 | 193 | https://github.com/huggingface/diffusers/issues/193 | https://api.github.com/repos/huggingface/diffusers/issues/193 | Pipeline error | ### Describe the bug
I am copying some code from Patrick but I keep getting errors, here I copy
```python
from diffusers import StableDiffusionPipeline
# make sure you're logged in with `huggingface-cli login`
pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-3-diffusers", use_auth... | closed | completed | false | 2 | [
"bug"
] | [] | 2022-08-17T07:44:41Z | 2022-09-13T15:52:22Z | 2022-09-13T15:52:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jpbonino | 81,916,559 | MDQ6VXNlcjgxOTE2NTU5 | User | false |
huggingface/diffusers | 1,521,190,104 | I_kwDOHa8MBc5aq4TY | 1,931 | https://github.com/huggingface/diffusers/issues/1931 | https://api.github.com/repos/huggingface/diffusers/issues/1931 | stable diffusion 2 float16 mode not working: expected scalar type Half but found Float | ### Describe the bug
https://huggingface.co/stabilityai/stable-diffusion-2-1#examples does not run in stated form, but does work if you omit `torch_dtype=torch.float16`.
### Reproduction
```
import torch
from diffusers import StableDiffusionPipeline, DPMSolverMultistepScheduler
model_id = "stabilityai/stable-di... | closed | completed | false | 11 | [
"bug",
"stale"
] | [] | 2023-01-05T18:26:01Z | 2023-03-15T15:04:20Z | 2023-03-15T15:04:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | murphyk | 4,632,336 | MDQ6VXNlcjQ2MzIzMzY= | User | false |
huggingface/diffusers | 1,521,368,642 | I_kwDOHa8MBc5arj5C | 1,933 | https://github.com/huggingface/diffusers/issues/1933 | https://api.github.com/repos/huggingface/diffusers/issues/1933 | Fix PNDMScheduler docstring | In the PNDM docstring it is saying that the `prediction_type` argument could be `epsilon` , `sample` or `v_prediction`.
https://github.com/huggingface/diffusers/blob/247b5feea125c6933384c59dcfd6d286c71dd8ba/src/diffusers/schedulers/scheduling_pndm.py#L87
But in the step method, it raises an error if we use `sampl... | closed | completed | false | 1 | [] | [] | 2023-01-05T20:29:53Z | 2023-01-12T20:01:41Z | 2023-01-12T20:01:41Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Warvito | 4,493,399 | MDQ6VXNlcjQ0OTMzOTk= | User | false |
huggingface/diffusers | 1,521,834,248 | I_kwDOHa8MBc5atVkI | 1,934 | https://github.com/huggingface/diffusers/issues/1934 | https://api.github.com/repos/huggingface/diffusers/issues/1934 | cpu_offload vRAM memory consumption large than 4GB | ### Describe the bug
I am using the code from https://huggingface.co/docs/diffusers/optimization/fp16#offloading-to-cpu-with-accelerate-for-memory-savings to test `cpu_offload`, but the vRAM memory consumption is large than 4GB
|GPU| cpu_offload enabled| vRAM cost|
|----|---------------------|------------|
| 1... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-01-06T03:23:31Z | 2023-02-05T23:06:21Z | 2023-02-05T23:06:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Sanster | 3,998,421 | MDQ6VXNlcjM5OTg0MjE= | User | false |
huggingface/diffusers | 1,523,146,453 | I_kwDOHa8MBc5ayV7V | 1,941 | https://github.com/huggingface/diffusers/issues/1941 | https://api.github.com/repos/huggingface/diffusers/issues/1941 | Unable to load custom pipeline for UnCLIP pipeline | ### Describe the bug
had been frequently using a custom pipeline for Karlo that now fails to load as of yesterday.
using: pipe = UnCLIPPipeline.from_pretrained("kakaobrain/karlo-v1-alpha",custom_pipeline="./karlo")
which gives
---------------------------------------------------------------------------
HTT... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-01-06T20:47:44Z | 2023-01-12T18:02:25Z | 2023-01-12T18:02:25Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ethansmith2000 | 98,723,285 | U_kgDOBeJl1Q | User | false |
huggingface/diffusers | 1,523,413,634 | I_kwDOHa8MBc5azXKC | 1,943 | https://github.com/huggingface/diffusers/issues/1943 | https://api.github.com/repos/huggingface/diffusers/issues/1943 | Prompt weighting would need better documentation | Hello,
I found this in the documentation: https://huggingface.co/docs/diffusers/v0.11.0/en/using-diffusers/custom_pipeline_examples#long-prompt-weighting-stable-diffusion
But it's not very clear to me whether prompt weighting is actually supported for Stable Diffusion 2.1 and 1.5? Is it?
I have also seen various fla... | closed | completed | false | 1 | [] | [] | 2023-01-07T02:31:46Z | 2023-01-09T02:29:35Z | 2023-01-09T02:29:35Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexisrolland | 13,064,696 | MDQ6VXNlcjEzMDY0Njk2 | User | false |
huggingface/diffusers | 1,523,513,400 | I_kwDOHa8MBc5azvg4 | 1,944 | https://github.com/huggingface/diffusers/issues/1944 | https://api.github.com/repos/huggingface/diffusers/issues/1944 | Dreambooth on SD1.5/SD2.1-768 - use prompt "an apple" to generate but got random results | I used dreambooth to fine-tune the models based on SD1.5 and SD2.1-768 with my photos. When I use my token (`xxx man`) in a prompt to generate, it only gives me something similar to my training images (but very ugly) and somehow the entire prompt was ignored.
The weirdest thing is that when I try to use `an apple`, ... | closed | completed | false | 10 | [
"stale"
] | [] | 2023-01-07T06:08:15Z | 2023-02-20T15:03:05Z | 2023-02-20T15:03:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | qiaoshun8888 | 1,053,081 | MDQ6VXNlcjEwNTMwODE= | User | false |
huggingface/diffusers | 1,523,789,090 | I_kwDOHa8MBc5a0y0i | 1,946 | https://github.com/huggingface/diffusers/issues/1946 | https://api.github.com/repos/huggingface/diffusers/issues/1946 | RuntimeError: CUDA error: invalid argument when using xformers | ### Describe the bug
When trying to run train_dreambooth.py with --enable_xformers_memory_efficient_attention the process exits with this error:
```
RuntimeError: CUDA error: invalid argument
CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect.
... | closed | completed | false | 19 | [
"bug",
"stale"
] | [
"patil-suraj"
] | 2023-01-07T12:38:55Z | 2023-03-30T15:03:58Z | 2023-03-30T15:03:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vmajor | 5,185,905 | MDQ6VXNlcjUxODU5MDU= | User | false |
huggingface/diffusers | 1,524,493,202 | I_kwDOHa8MBc5a3euS | 1,949 | https://github.com/huggingface/diffusers/issues/1949 | https://api.github.com/repos/huggingface/diffusers/issues/1949 | Errors while trying to train a custom model using the tutorial within the diffusers>examples>dreambooth | ### Describe the bug
NameError: name 'str2 optimizer 8bit blockwise' is not defined
Li... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-01-08T12:45:19Z | 2023-02-15T15:03:10Z | 2023-02-15T15:03:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tushar-31093 | 40,886,260 | MDQ6VXNlcjQwODg2MjYw | User | false |
huggingface/diffusers | 1,341,357,999 | I_kwDOHa8MBc5P83-v | 195 | https://github.com/huggingface/diffusers/issues/195 | https://api.github.com/repos/huggingface/diffusers/issues/195 | Implement `pipeline.to(device)` | Currently, pipeline modules are moved to the preferred compute device during `__call__`. This is reasonable, as they stay there as long as the user keeps passing the same `torch_device` across calls.
However, in multi-GPU model-serving scenarios, it could be useful to move each pipeline to a dedicated device during ... | closed | completed | false | 5 | [] | [
"pcuenca"
] | 2022-08-17T08:13:26Z | 2022-08-19T16:39:13Z | 2022-08-19T16:39:13Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pcuenca | 1,177,582 | MDQ6VXNlcjExNzc1ODI= | User | false |
huggingface/diffusers | 1,525,148,153 | I_kwDOHa8MBc5a5-n5 | 1,952 | https://github.com/huggingface/diffusers/issues/1952 | https://api.github.com/repos/huggingface/diffusers/issues/1952 | stable diffusion inference | Here is the command when testing the stable diffusion inference:
python3 scripts/txt2img.py --prompt "a photograph of an astronaut riding a horse" --plms --config configs/Inference/v2-inference.yaml --ckpt 768-v-ema.ckpt
768-v-ema.ckpt is downloaded from https://huggingface.co/stabilityai/stable-diffusion-2/tree/main... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-01-09T08:33:17Z | 2023-02-16T15:03:39Z | 2023-02-16T15:03:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | westnight | 32,773,215 | MDQ6VXNlcjMyNzczMjE1 | User | false |
huggingface/diffusers | 1,525,149,291 | I_kwDOHa8MBc5a5-5r | 1,953 | https://github.com/huggingface/diffusers/issues/1953 | https://api.github.com/repos/huggingface/diffusers/issues/1953 | Performance of diffusers | I trained a model with diffusers. But I found the testing performance is not stable for the trained model. Sometimes, it can generate high-quality images, but sometimes, the images does not make sense or may be weird or distorted. What factors may affect the permformance? How can I improve the model performance? I nee... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-01-09T08:34:15Z | 2023-02-27T15:03:50Z | 2023-02-27T15:03:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | StrugglingForBetter | 20,771,340 | MDQ6VXNlcjIwNzcxMzQw | User | false |
huggingface/diffusers | 1,525,361,824 | I_kwDOHa8MBc5a6yyg | 1,954 | https://github.com/huggingface/diffusers/issues/1954 | https://api.github.com/repos/huggingface/diffusers/issues/1954 | Dreambooth inference result is not correct when training with xformer | I train dreambooth with xformer but the result is not correct (It’s correct without xformer). The model initialize from runwayml/stable-diffusion-v1-5, but the inference result is the same as the runwayml/stable-diffusion-v1-5. Maybe there are some errors when saving the model with xformer? I also add the pipe.enable_... | closed | completed | false | 16 | [
"stale"
] | [
"patil-suraj"
] | 2023-01-09T11:04:06Z | 2023-03-26T15:03:12Z | 2023-03-26T15:03:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mingqizhang | 28,778,038 | MDQ6VXNlcjI4Nzc4MDM4 | User | false |
huggingface/diffusers | 1,525,884,900 | I_kwDOHa8MBc5a8yfk | 1,955 | https://github.com/huggingface/diffusers/issues/1955 | https://api.github.com/repos/huggingface/diffusers/issues/1955 | Custom pipelines are downloaded even though present locally | ### Describe the bug
When adding a custom pipeline to the `examples/community` folder and installing my local diffusers via `pip3 install -e .` diffusers still tries to download the pipeline from HF hub.
### Reproduction
see above
### Logs
_No response_
### System Info
diffusers@aba2a65d6ab47c0d1c1... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-01-09T16:08:05Z | 2023-02-19T15:02:43Z | 2023-02-19T15:02:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chris-aeviator | 11,522,213 | MDQ6VXNlcjExNTIyMjEz | User | false |
huggingface/diffusers | 1,526,128,140 | I_kwDOHa8MBc5a9t4M | 1,957 | https://github.com/huggingface/diffusers/issues/1957 | https://api.github.com/repos/huggingface/diffusers/issues/1957 | Train Unconditional - LR Scheduler init fix | ### Learning Rate Scheduler with gradient accumulation
Hi,
There seems to be a small error in the [train_unconditional.py](https://github.com/huggingface/diffusers/blob/main/examples/unconditional_image_generation/train_unconditional.py) script regarding the initialization of the learning rate scheduler (`lr_sch... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-01-09T18:47:32Z | 2023-01-17T09:11:48Z | 2023-01-17T09:11:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mathieujouffroy | 45,208,116 | MDQ6VXNlcjQ1MjA4MTE2 | User | false |
huggingface/diffusers | 1,526,963,135 | I_kwDOHa8MBc5bA5u_ | 1,961 | https://github.com/huggingface/diffusers/issues/1961 | https://api.github.com/repos/huggingface/diffusers/issues/1961 | colossalai error | ### Describe the bug
OSError: stable-diffusion-v1-5 does not appear to have a file named tokenizer/config.json. Checkout
### Reproduction
export MODEL_NAME="stable-diffusion-v1-5"
export INSTANCE_DIR="img"
export OUTPUT_DIR="idp-ysl"
torchrun --nproc_per_node 2 train_dreambooth_colossalai.py \
--pretraine... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-01-10T08:24:00Z | 2023-02-17T15:03:11Z | 2023-02-17T15:03:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | huanchengying | 17,674,609 | MDQ6VXNlcjE3Njc0NjA5 | User | false |
huggingface/diffusers | 1,527,150,809 | I_kwDOHa8MBc5bBnjZ | 1,962 | https://github.com/huggingface/diffusers/issues/1962 | https://api.github.com/repos/huggingface/diffusers/issues/1962 | Video generation with stable diffusion | ### Model/Pipeline/Scheduler description
Hey,
Thanks for sharing.
Please check out my modified version of video generation with stable diffusion:
https://github.com/feizc/Video-Stable-Diffusion
### Open source status
- [X] The model implementation is available
- [X] The model weights are available (Onl... | closed | completed | false | 14 | [
"New pipeline/model",
"community-examples",
"Good second issue"
] | [] | 2023-01-10T10:39:54Z | 2024-11-21T07:29:45Z | 2024-11-21T07:29:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | feizc | 37,614,046 | MDQ6VXNlcjM3NjE0MDQ2 | User | false |
huggingface/diffusers | 1,527,227,251 | I_kwDOHa8MBc5bB6Nz | 1,963 | https://github.com/huggingface/diffusers/issues/1963 | https://api.github.com/repos/huggingface/diffusers/issues/1963 | Latent unconditionnal generation | Hi,
in "https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/latent_diffusion_uncond/pipeline_latent_diffusion_uncond.py" a VQVAE is used so that diffusion is performed on a lower dimensional latent space. I'd like to generate my own data however I found no script to train a personal VQVAE and it... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-01-10T11:28:11Z | 2023-02-17T15:03:10Z | 2023-02-17T15:03:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Kageyoshi317 | 122,362,383 | U_kgDOB0saDw | User | false |
huggingface/diffusers | 1,527,298,571 | I_kwDOHa8MBc5bCLoL | 1,964 | https://github.com/huggingface/diffusers/issues/1964 | https://api.github.com/repos/huggingface/diffusers/issues/1964 | Support plan for Pytorch 2.0 | **Is your feature request related to a problem? Please describe.**
Although the stable version of Pytorch 2.0 has not been released yet, its new features do improve efficiency with only small modification on training code. Surely, I notice that some functions are re-named in the newer Pytorch.
**Describe the soluti... | closed | completed | false | 4 | [
"stale"
] | [
"patil-suraj"
] | 2023-01-10T12:20:50Z | 2023-03-17T15:04:02Z | 2023-03-17T15:04:02Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | haofanwang | 18,741,068 | MDQ6VXNlcjE4NzQxMDY4 | User | false |
huggingface/diffusers | 1,527,778,886 | I_kwDOHa8MBc5bEA5G | 1,969 | https://github.com/huggingface/diffusers/issues/1969 | https://api.github.com/repos/huggingface/diffusers/issues/1969 | RuntimeError: view size is not compatible with input tensor's size and stride | ### Describe the bug
There seems to be a new error when using the .view(...) operation during forward pass for train unconditional task. This issue didn't arise until 1/10 and not sure if this affects any other models/tasks.
Potential solution found on stack overflow (https://stackoverflow.com/questions/66750391/... | closed | completed | false | 4 | [
"bug",
"stale"
] | [
"anton-l"
] | 2023-01-10T17:43:49Z | 2023-02-19T15:02:42Z | 2023-02-19T15:02:42Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | prathikr | 31,260,940 | MDQ6VXNlcjMxMjYwOTQw | User | false |
huggingface/diffusers | 1,341,461,180 | I_kwDOHa8MBc5P9RK8 | 197 | https://github.com/huggingface/diffusers/issues/197 | https://api.github.com/repos/huggingface/diffusers/issues/197 | Resnet - Delete dead code | There is a lot of dead
code in the file: https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/resnet.py which should be cleaned up, namely:
- Pretty much all code after: https://github.com/huggingface/diffusers/blob/80e0c8ba9ebd7b3ddd82f3e443715d4e7f08f2bb/src/diffusers/models/resnet.py#L389
- A... | closed | completed | false | 2 | [
"enhancement",
"good first issue"
] | [] | 2022-08-17T09:31:16Z | 2022-08-23T06:38:38Z | 2022-08-23T06:38:38Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,527,792,955 | I_kwDOHa8MBc5bEEU7 | 1,970 | https://github.com/huggingface/diffusers/issues/1970 | https://api.github.com/repos/huggingface/diffusers/issues/1970 | Problem with `local_files_only` | ### Describe the bug
There is a bug in `DiffusionPIpeline.from_pretrained()`.
I run `from_pretrained('repo-id')` on a machine with internet connection, then disable the internet and run from_pretrained() with the option local_files_only=True. However, the call fails with a ProxyError('Cannot connect to proxy')
W... | closed | completed | false | 3 | [
"bug"
] | [
"patrickvonplaten"
] | 2023-01-10T17:55:50Z | 2023-01-27T07:24:37Z | 2023-01-27T07:24:37Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,528,323,114 | I_kwDOHa8MBc5bGFwq | 1,973 | https://github.com/huggingface/diffusers/issues/1973 | https://api.github.com/repos/huggingface/diffusers/issues/1973 | text_to_image and image to images | I trained finetuned a stable diffusion model with my owndataset. The training scipt is like
https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image.py
The text_to_image pipleline output images with styles out of my expections. But if I use image to image pipeline, and feed it... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-01-11T02:44:24Z | 2023-02-24T15:03:41Z | 2023-02-24T15:03:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | StrugglingForBetter | 20,771,340 | MDQ6VXNlcjIwNzcxMzQw | User | false |
huggingface/diffusers | 1,528,350,692 | I_kwDOHa8MBc5bGMfk | 1,974 | https://github.com/huggingface/diffusers/issues/1974 | https://api.github.com/repos/huggingface/diffusers/issues/1974 | Euler schedulers don't work when passing a list of generators for reproducible batch generation | ### Describe the bug
With diffusers 0.11, I can now pass in a list of generators for the `generator` param when calling the pipeline, to supply a generator per-image when generating multiple images at once. This lets me control the seeds of all the generated images - great!
But for schedulers that accept a genera... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-01-11T03:09:55Z | 2023-01-11T04:31:28Z | 2023-01-11T04:31:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nhnt11 | 485,488 | MDQ6VXNlcjQ4NTQ4OA== | User | false |
huggingface/diffusers | 1,528,419,906 | I_kwDOHa8MBc5bGdZC | 1,975 | https://github.com/huggingface/diffusers/issues/1975 | https://api.github.com/repos/huggingface/diffusers/issues/1975 | textual inversion for inpainting | Is there an easy way to use word embedding extracted from textual inversion for inpainting? | closed | completed | false | 2 | [
"stale"
] | [] | 2023-01-11T04:40:08Z | 2023-02-19T15:02:41Z | 2023-02-19T15:02:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | SOTAMak1r | 113,093,634 | U_kgDOBr2sAg | User | false |
huggingface/diffusers | 1,528,474,666 | I_kwDOHa8MBc5bGqwq | 1,976 | https://github.com/huggingface/diffusers/issues/1976 | https://api.github.com/repos/huggingface/diffusers/issues/1976 | transformers looking for an incorrect file OSError: runwayml/stable-diffusion-v1-5 does not appear to have a file named tokenizer/config.json. | ### Describe the bug
transformers 4.24.0 is insisting on trying to find `OSError: runwayml/stable-diffusion-v1-5 does not appear to have a file named tokenizer/config.json.`
However this will never happen because the config file is actually `tokenizer_config.json`
### Reproduction
Attempt to run:
`export M... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-01-11T05:54:55Z | 2023-02-19T15:02:40Z | 2023-02-19T15:02:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vmajor | 5,185,905 | MDQ6VXNlcjUxODU5MDU= | User | false |
huggingface/diffusers | 1,529,404,701 | I_kwDOHa8MBc5bKN0d | 1,979 | https://github.com/huggingface/diffusers/issues/1979 | https://api.github.com/repos/huggingface/diffusers/issues/1979 | Cannot convert SD to diffusers | ### Describe the bug
Trying to convert a ckpt fine tuned from dreambooth (2.1 768) to diffusers yield the following error:
KeyError: 'cond_stage_model.model.text_projection'
This happens either locally or using a space from hugging face:
https://huggingface.co/spaces/diffusers/sd-to-diffusers
### Reproduction... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2023-01-11T17:15:41Z | 2023-03-10T22:12:54Z | 2023-03-10T22:12:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | brurpo | 24,902,022 | MDQ6VXNlcjI0OTAyMDIy | User | false |
huggingface/diffusers | 1,341,468,202 | I_kwDOHa8MBc5P9S4q | 198 | https://github.com/huggingface/diffusers/issues/198 | https://api.github.com/repos/huggingface/diffusers/issues/198 | Rename variables from single letter to better naming | Pretty much all code in https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/resnet.py still corresponds to the copy-pasted version from the beginning.
While weight names shouldn't be renamed, we should rename all single letter variables to more explicit variable naming. E.g. `h` -> `hidden_states... | closed | completed | false | 21 | [
"enhancement",
"good first issue"
] | [] | 2022-08-17T09:34:02Z | 2023-06-28T11:31:34Z | 2023-06-28T11:31:34Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,530,166,929 | I_kwDOHa8MBc5bNH6R | 1,981 | https://github.com/huggingface/diffusers/issues/1981 | https://api.github.com/repos/huggingface/diffusers/issues/1981 | The diffusers-pytorch-cuda/Dockerfile uses wrong python3 because of missing ln -s link | ### Describe the bug
The diffusers-pytorch-cuda/Dockerfile uses wrong python3 because of missing ln -s link.
As a result, python3 is python3.6.9 of nvidia/cuda:11.7.1-cudnn8-runtime-ubuntu20.04, but not python3.8.
### Solved
```shell
RUN rm /usr/bin/python3 && \
ln -s /usr/bin/python3.8 /usr/bin/python3
... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-01-12T06:13:18Z | 2023-01-17T02:54:36Z | 2023-01-13T01:19:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kenny-chen | 68,576,845 | MDQ6VXNlcjY4NTc2ODQ1 | User | false |
huggingface/diffusers | 1,531,268,379 | I_kwDOHa8MBc5bRU0b | 1,984 | https://github.com/huggingface/diffusers/issues/1984 | https://api.github.com/repos/huggingface/diffusers/issues/1984 | Allow passing a `CACHE_MAPPING` of previously downloaded components and their sha256 to avoid deduplicated downloading | A common problem of diffusers models/pipelines is that many components of different pipelines share the exact same underlying weights, but it's hard to avoid not downloading them twice. We could solve this problem by providing a `cache_mapping: Dict[str, path]` to `DiffusionPipeline.from_pretrained(...)` that would che... | open | null | false | 10 | [
"wip"
] | [] | 2023-01-12T19:50:50Z | 2023-04-11T18:18:44Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,531,437,275 | I_kwDOHa8MBc5bR-Db | 1,985 | https://github.com/huggingface/diffusers/issues/1985 | https://api.github.com/repos/huggingface/diffusers/issues/1985 | Please create simple way to load new embeddings on Stable Diffusion Pipeline | **Is your feature request related to a problem? Please describe.**
It is frustrating to load Textual Inversion embeddings on SD pipelines, adds bloat to the code. It would be good if we could just `pipeline.load_embeddings` and add new embeddings ad-hoc to the pipeline.
**Describe the solution you'd like**
I would... | closed | completed | false | 11 | [] | [] | 2023-01-12T22:16:22Z | 2023-03-30T17:08:42Z | 2023-03-30T17:08:42Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | piEsposito | 47,679,710 | MDQ6VXNlcjQ3Njc5NzEw | User | false |
huggingface/diffusers | 1,531,485,067 | I_kwDOHa8MBc5bSJuL | 1,986 | https://github.com/huggingface/diffusers/issues/1986 | https://api.github.com/repos/huggingface/diffusers/issues/1986 | Is this correct?? | In the convert_stable_diffusion_checkpoint_to_onnx.py the inputs for the unet are defined as:
```
model_args=(
torch.randn(2, unet_in_channels, unet_sample_size, unet_sample_size).to(device=device, dtype=dtype),
torch.randn(2).to(device=device, dtype=dtype),
torch.rand... | closed | completed | false | 4 | [] | [
"anton-l"
] | 2023-01-12T23:28:04Z | 2023-01-16T13:44:01Z | 2023-01-16T13:41:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | elephantpanda | 33,497,043 | MDQ6VXNlcjMzNDk3MDQz | User | false |
huggingface/diffusers | 1,531,534,827 | I_kwDOHa8MBc5bSV3r | 1,987 | https://github.com/huggingface/diffusers/issues/1987 | https://api.github.com/repos/huggingface/diffusers/issues/1987 | from_flax only working with safety_checker=None | ### Describe the bug
when I try converting without `safety_checker=None`
`loaded_sub_model` pointing to transformers/modeling_flax_pytorch_utils.py 😐
https://github.com/huggingface/diffusers/blob/135567f18edc0bf02d515d5c76cc736d1ebddad3/src/diffusers/pipelines/pipeline_utils.py#L735
### Reproduction
working
... | closed | completed | false | 3 | [
"bug"
] | [
"pcuenca"
] | 2023-01-13T00:33:03Z | 2023-02-04T19:56:17Z | 2023-02-04T19:56:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | camenduru | 54,370,274 | MDQ6VXNlcjU0MzcwMjc0 | User | false |
huggingface/diffusers | 1,531,697,032 | I_kwDOHa8MBc5bS9eI | 1,988 | https://github.com/huggingface/diffusers/issues/1988 | https://api.github.com/repos/huggingface/diffusers/issues/1988 | Generating images in text to image trainer example w/ fp16/bf16: RuntimeError Input type and bias type should be the same | ### Describe the bug
I added a simple function train_text_to_image.py to generate a sample at a checkpointing step:
```python3
def save_in_progress():
nonlocal unet
if accelerator.is_main_process:
unet = accelerator.unwrap_model(unet)
if args.use_ema:
ema_unet.copy_to(une... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-01-13T04:36:46Z | 2023-01-19T00:02:00Z | 2023-01-19T00:01:22Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Cyberes | 64,224,601 | MDQ6VXNlcjY0MjI0NjAx | User | false |
huggingface/diffusers | 1,341,480,686 | I_kwDOHa8MBc5P9V7u | 199 | https://github.com/huggingface/diffusers/issues/199 | https://api.github.com/repos/huggingface/diffusers/issues/199 | Attention - Delete dead code | The attention file: https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention.py also contains still quite some dead code that we should clean up.
This class is never used and should be deleted
https://github.com/huggingface/diffusers/blob/80e0c8ba9ebd7b3ddd82f3e443715d4e7f08f2bb/src/diffuse... | closed | completed | false | 1 | [
"enhancement",
"good first issue"
] | [] | 2022-08-17T09:43:48Z | 2022-08-25T15:55:24Z | 2022-08-25T15:54:49Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,532,667,971 | I_kwDOHa8MBc5bWqhD | 1,990 | https://github.com/huggingface/diffusers/issues/1990 | https://api.github.com/repos/huggingface/diffusers/issues/1990 | Pipeline proposal: Depth2ImgVariation | ### Model/Pipeline/Scheduler description
It would be nice to have a pipeline that combines the `StableDiffusionImageVariationPipeline` with the `StableDiffusionDepth2ImgPipeline`, i.e. a pipeline that creates an image from a depth map, but guided by image embeddings rather than a prompt.
### Open source status
... | closed | completed | false | 8 | [
"community-examples",
"stale"
] | [] | 2023-01-13T17:23:28Z | 2023-02-24T15:03:39Z | 2023-02-24T15:03:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | maxfrei750 | 36,265,931 | MDQ6VXNlcjM2MjY1OTMx | User | false |
huggingface/diffusers | 1,532,831,037 | I_kwDOHa8MBc5bXSU9 | 1,991 | https://github.com/huggingface/diffusers/issues/1991 | https://api.github.com/repos/huggingface/diffusers/issues/1991 | Checkpoint Merger Pipeline giving error: "Expected all tensors to be on the same device" | ### Describe the bug
Hi,
I'm following the three-way merge example given at https://github.com/huggingface/diffusers/tree/main/examples/community#checkpoint-merger-pipeline, and am getting the following reproducible error with `diffusers 0.11.1`:
```
RuntimeError: Expected all tensors to be on the same device... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-01-13T19:56:45Z | 2023-01-16T13:25:28Z | 2023-01-16T13:25:28Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lstein | 111,189 | MDQ6VXNlcjExMTE4OQ== | User | false |
huggingface/diffusers | 1,532,900,064 | I_kwDOHa8MBc5bXjLg | 1,992 | https://github.com/huggingface/diffusers/issues/1992 | https://api.github.com/repos/huggingface/diffusers/issues/1992 | JAX Attention Refactoring? | At the moment, pytorch version of most attention had been refactored to use attention processor (except a few being deprecated). Is there plans to do the same for flax version?
Since AttentionBlock is deprecated, should the same be done to its flax counterpart?
https://github.com/huggingface/diffusers/blob/9b37ed... | closed | completed | false | 8 | [] | [
"pcuenca"
] | 2023-01-13T20:52:25Z | 2023-03-05T17:40:47Z | 2023-03-05T17:40:47Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Lime-Cakes | 91,322,985 | MDQ6VXNlcjkxMzIyOTg1 | User | false |
huggingface/diffusers | 1,533,102,760 | I_kwDOHa8MBc5bYUqo | 1,993 | https://github.com/huggingface/diffusers/issues/1993 | https://api.github.com/repos/huggingface/diffusers/issues/1993 | Save sample images while training Dreambooth & Textual Inversion | During Dreambooth or Textual Inversion training, I want to see periodic sample images to see how the training is going. There should be an option to save every X steps (or epochs).
I've been using the following code, which I adapted from the ShivamShrirao fork, but it does not seem to be very effective at showing an... | closed | completed | false | 4 | [
"stale"
] | [
"patrickvonplaten",
"patil-suraj"
] | 2023-01-14T02:24:53Z | 2023-03-14T21:06:12Z | 2023-03-14T21:06:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | David-Hari | 7,530,738 | MDQ6VXNlcjc1MzA3Mzg= | User | false |
huggingface/diffusers | 1,533,105,872 | I_kwDOHa8MBc5bYVbQ | 1,994 | https://github.com/huggingface/diffusers/issues/1994 | https://api.github.com/repos/huggingface/diffusers/issues/1994 | from_pretrained , save_directory support aws 3 | **Is your feature request related to a problem? Please describe.**
I wan use s3 store my model , if diffusers from_pretrained, save_directory can load/save model to s3 , I think that's cool
| closed | completed | false | 2 | [
"stale"
] | [] | 2023-01-14T02:38:38Z | 2023-02-21T15:03:39Z | 2023-02-21T15:03:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | stevensu1977 | 6,694,421 | MDQ6VXNlcjY2OTQ0MjE= | User | false |
huggingface/diffusers | 1,533,679,603 | I_kwDOHa8MBc5bahfz | 1,997 | https://github.com/huggingface/diffusers/issues/1997 | https://api.github.com/repos/huggingface/diffusers/issues/1997 | Reproducibility improvement of SD generation with xFormers | I wrote a patch related to #1339. This patch makes xFormers use Flash Attention as a priority. More details to see: [Generating (almost) reproducible pictures using Diffusers with xFormers](https://gist.github.com/takuma104/9d25bb87ae3b52e41e0132aa737c0b03)
![fig_xformers_flash_attention_seed_7](https://user-images.... | closed | completed | false | 4 | [
"stale"
] | [
"pcuenca",
"patil-suraj"
] | 2023-01-15T06:21:24Z | 2023-02-23T15:03:21Z | 2023-02-23T15:03:21Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | takuma104 | 10,776 | MDQ6VXNlcjEwNzc2 | User | false |
huggingface/diffusers | 1,533,946,383 | I_kwDOHa8MBc5bbioP | 1,999 | https://github.com/huggingface/diffusers/issues/1999 | https://api.github.com/repos/huggingface/diffusers/issues/1999 | [Community Pipeline] Paint with Words | Just saw this https://github.com/cloneofsimo/paint-with-words-sd which is based on Nvidia's paper of paint with words. If it's not yet implemented, a community pipeline would be pretty cool.
Essentially it's img2img, but with more control as a description of what each color should be is supplied | open | null | false | 9 | [
"community-examples",
"wip"
] | [
"cybergeekgyan"
] | 2023-01-15T18:28:24Z | 2023-03-23T12:43:32Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dblunk88 | 39,381,389 | MDQ6VXNlcjM5MzgxMzg5 | User | false |
huggingface/diffusers | 1,341,484,752 | I_kwDOHa8MBc5P9W7Q | 200 | https://github.com/huggingface/diffusers/issues/200 | https://api.github.com/repos/huggingface/diffusers/issues/200 | [Test] Test the resnet class | We should write tests for the resnet class: https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/resnet.py | closed | completed | false | 10 | [
"enhancement",
"Good second issue"
] | [] | 2022-08-17T09:46:37Z | 2024-11-21T11:20:46Z | 2024-11-21T11:20:46Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,534,065,941 | I_kwDOHa8MBc5bb_0V | 2,000 | https://github.com/huggingface/diffusers/issues/2000 | https://api.github.com/repos/huggingface/diffusers/issues/2000 | Loading checkpoint dreambooth fine tuned sd from local disk | hi team, Im making many experiments fine tuning SD with dreambooth and I want to load the models directly from my local disk,
im using the code from : https://github.com/nateraw/stable-diffusion-videos
So I download the entire SD 1.5, repo doing:
git lfs install
git clone https://huggingface.co/runwayml/stable-di... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-01-16T00:11:53Z | 2023-02-23T15:03:19Z | 2023-02-23T15:03:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | javismiles | 1,000,452 | MDQ6VXNlcjEwMDA0NTI= | User | false |
huggingface/diffusers | 1,534,997,110 | I_kwDOHa8MBc5bfjJ2 | 2,008 | https://github.com/huggingface/diffusers/issues/2008 | https://api.github.com/repos/huggingface/diffusers/issues/2008 | Does text_encoder work in float16? | I recently converted stable diffusion to float16 onnx with the convert tool.
Everything works great. Except the text_encoder.onnx doesn't work (as in it doesn't produce any images just random patterns). So I am still using the float32 version of the text_econder.
My assumption is that because although the input i... | closed | completed | false | 5 | [
"stale"
] | [
"anton-l"
] | 2023-01-16T14:05:43Z | 2023-03-23T15:03:36Z | 2023-03-23T15:03:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | elephantpanda | 33,497,043 | MDQ6VXNlcjMzNDk3MDQz | User | false |
huggingface/diffusers | 1,341,485,857 | I_kwDOHa8MBc5P9XMh | 201 | https://github.com/huggingface/diffusers/issues/201 | https://api.github.com/repos/huggingface/diffusers/issues/201 | [Tests] Test attention class | Test attention class: https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention.py
We should test this class well | closed | completed | false | 5 | [
"enhancement",
"Good second issue"
] | [] | 2022-08-17T09:47:29Z | 2024-11-21T11:22:00Z | 2024-11-21T11:22:00Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,535,530,822 | I_kwDOHa8MBc5bhldG | 2,011 | https://github.com/huggingface/diffusers/issues/2011 | https://api.github.com/repos/huggingface/diffusers/issues/2011 | the `attention_head_dim` argument for `UNet2DConditionModel` | the `attention_head_dim` in `UNet2DConditionModel` seems to be passed down to `CrossAttnDownBlock2D` and `CrossAttnUpBlock2D` as the number of attention head, instead of the dimension of each attention head
```python
from diffusers import UNet2DConditionModel
unet = UNet2DConditionModel(
attention_head_dim... | closed | completed | false | 14 | [
"stale"
] | [] | 2023-01-16T21:59:09Z | 2024-07-11T01:56:43Z | 2023-07-18T15:04:05Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 1,535,606,143 | I_kwDOHa8MBc5bh31_ | 2,012 | https://github.com/huggingface/diffusers/issues/2012 | https://api.github.com/repos/huggingface/diffusers/issues/2012 | Reduce Imagic Pipeline Memory Consumption | I'm running the [Imagic Stable Diffusion community pipeline](https://github.com/huggingface/diffusers/blob/main/examples/community/imagic_stable_diffusion.py) and it's routinely allocating 25-38 GiB GPU vRAM which seems excessively high.
@MarkRich any ideas on how to reduce memory usage? Xformers and attention slic... | closed | completed | false | 1 | [
"question",
"stale"
] | [] | 2023-01-16T23:43:03Z | 2023-02-24T15:03:35Z | 2023-02-24T15:03:35Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andreemic | 27,695,722 | MDQ6VXNlcjI3Njk1NzIy | User | false |
huggingface/diffusers | 1,535,658,452 | I_kwDOHa8MBc5biEnU | 2,013 | https://github.com/huggingface/diffusers/issues/2013 | https://api.github.com/repos/huggingface/diffusers/issues/2013 | Check k-diffusion version to prevent ImportError | `k_diffusion.external` was added in `k-diffusion==0.0.12`, if an older version is installed, user will get an ImportError
```python
ImportError: cannot import name 'CompVisVDenoiser' from 'k_diffusion.external'
```
# Reproduce code
- diffusers version: 0.11.1
- k-diffusion==0.0.11
```python
import diffuse... | closed | completed | false | 0 | [] | [] | 2023-01-17T01:28:45Z | 2023-01-17T20:16:49Z | 2023-01-17T20:16:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Sanster | 3,998,421 | MDQ6VXNlcjM5OTg0MjE= | User | false |
huggingface/diffusers | 1,535,721,602 | I_kwDOHa8MBc5biUCC | 2,014 | https://github.com/huggingface/diffusers/issues/2014 | https://api.github.com/repos/huggingface/diffusers/issues/2014 | TorchScript compatible UNet | **Is your feature request related to a problem? Please describe.**
Currently UNet can be exported using `torch.jit.trace` but it's device dependent.
**Describe the solution you'd like**
It would be much better if I can use `torch.jit.script`, e.g. I can export the model on a cpu device then I can load and use the... | closed | completed | false | 6 | [
"stale"
] | [
"patrickvonplaten",
"patil-suraj"
] | 2023-01-17T03:22:37Z | 2023-07-14T14:37:00Z | 2023-02-24T15:03:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ChenchaoZhao | 35,147,961 | MDQ6VXNlcjM1MTQ3OTYx | User | false |
huggingface/diffusers | 1,535,723,028 | I_kwDOHa8MBc5biUYU | 2,015 | https://github.com/huggingface/diffusers/issues/2015 | https://api.github.com/repos/huggingface/diffusers/issues/2015 | Categorical diffusion pipeline | ### Model/Pipeline/Scheduler description
Are there any plans for diffusion models other than Gaussian, e.g. categorical diffusion https://arxiv.org/abs/2102.05379
### Open source status
- [X] The model implementation is available
- [X] The model weights are available (Only relevant if addition is not a schedu... | closed | completed | false | 1 | [
"stale"
] | [] | 2023-01-17T03:25:08Z | 2023-02-24T15:03:33Z | 2023-02-24T15:03:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ChenchaoZhao | 35,147,961 | MDQ6VXNlcjM1MTQ3OTYx | User | false |
huggingface/diffusers | 1,535,750,861 | I_kwDOHa8MBc5bibLN | 2,016 | https://github.com/huggingface/diffusers/issues/2016 | https://api.github.com/repos/huggingface/diffusers/issues/2016 | Sampling >= 1 image per device broke between v0.9.0 and v0.11.1? [flax] | ### Describe the bug
I am sampling from the `FlaxStableDiffusionPipeline` using a script almost identical to that in [the jax section of the README](https://github.com/huggingface/diffusers/tree/b2ea8a84e9db71c207d28687722d241447bde81e#jaxflax); the only modification is that I am sampling more than 1 image per devic... | closed | completed | false | 3 | [
"bug"
] | [
"pcuenca"
] | 2023-01-17T04:13:06Z | 2023-01-18T09:41:50Z | 2023-01-18T09:41:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jannerm | 10,070,626 | MDQ6VXNlcjEwMDcwNjI2 | User | false |
huggingface/diffusers | 1,535,772,017 | I_kwDOHa8MBc5bigVx | 2,017 | https://github.com/huggingface/diffusers/issues/2017 | https://api.github.com/repos/huggingface/diffusers/issues/2017 | How to do inpainting with vanilla stable diffusion? | Just taking the vanilla 1-4 stable diffusion model, can I get it to do good inpainting?
I can get it to do somewhat good inpainting, by taking the predicted latents and replacing the masked areas with noised latents from the original image at each step.
Also, it helps if I apply a guidance mask to the image so that... | closed | completed | false | 2 | [] | [] | 2023-01-17T04:48:33Z | 2023-01-23T00:46:47Z | 2023-01-23T00:46:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | elephantpanda | 33,497,043 | MDQ6VXNlcjMzNDk3MDQz | User | false |
huggingface/diffusers | 1,535,830,012 | I_kwDOHa8MBc5biuf8 | 2,018 | https://github.com/huggingface/diffusers/issues/2018 | https://api.github.com/repos/huggingface/diffusers/issues/2018 | How to do fp16 finetuning for dreambooth inpainting? | Hi @thedarkzeno and @patil-suraj
Thanks for your excellent work on dream booth inpainting.
My question is that how to perform fp16 dreambooth inpainting fine-tuning?
| closed | completed | false | 5 | [
"stale"
] | [] | 2023-01-17T06:10:10Z | 2023-02-27T15:03:43Z | 2023-02-27T15:03:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shileims | 104,052,591 | U_kgDOBjO3bw | User | false |
huggingface/diffusers | 1,341,486,941 | I_kwDOHa8MBc5P9Xdd | 202 | https://github.com/huggingface/diffusers/issues/202 | https://api.github.com/repos/huggingface/diffusers/issues/202 | [Test] Add tests for unet_blocks file | https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/unet_blocks.py needs tests | closed | completed | false | 6 | [
"enhancement",
"Good second issue"
] | [] | 2022-08-17T09:48:22Z | 2024-11-21T11:22:24Z | 2024-11-21T11:22:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,536,022,295 | I_kwDOHa8MBc5bjdcX | 2,020 | https://github.com/huggingface/diffusers/issues/2020 | https://api.github.com/repos/huggingface/diffusers/issues/2020 | Error running convert_dance_diffusion_to_diffusers.py | ### Describe the bug
When I try to convert a Dance Diffusion .ckpt file to use with my Diffusers implementation, I'm running into this error:
```
File "./convert_dance_diffusion_to_diffusers.py", line 10, in <module>
from audio_diffusion.models import DiffusionAttnUnet1D
ModuleNotFoundError: No module named 'audio... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-01-17T09:09:21Z | 2023-02-26T15:03:41Z | 2023-02-26T15:03:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Skquark | 3,192,935 | MDQ6VXNlcjMxOTI5MzU= | User | false |
huggingface/diffusers | 1,537,005,136 | I_kwDOHa8MBc5bnNZQ | 2,023 | https://github.com/huggingface/diffusers/issues/2023 | https://api.github.com/repos/huggingface/diffusers/issues/2023 | choice of prior preservation images when training a style with dreambooth | When using dreambooth, it is recommended that offering more regularization images to reduce overfitting and language drift. If you want to train a object, for example, a certain type of dog, you will easily find many images of dogs to be treated as prior preservation images.
However, if you want to train a style, how ... | closed | completed | false | 3 | [] | [
"patil-suraj"
] | 2023-01-17T20:36:42Z | 2023-01-24T12:40:21Z | 2023-01-24T12:40:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chesha1 | 49,867,392 | MDQ6VXNlcjQ5ODY3Mzky | User | false |
huggingface/diffusers | 1,537,018,360 | I_kwDOHa8MBc5bnQn4 | 2,024 | https://github.com/huggingface/diffusers/issues/2024 | https://api.github.com/repos/huggingface/diffusers/issues/2024 | Textual Inversion Training notebook: "The file has been corrupted or is not a valid notebook file." | ### Describe the bug
I try to open the training notebook on linked at https://github.com/huggingface/diffusers/blob/main/examples/textual_inversion/README.md colab to https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/sd_textual_inversion_training.ipynb
```
SyntaxError: JSON.pa... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-01-17T20:50:05Z | 2023-01-18T18:28:43Z | 2023-01-18T16:34:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CptanPanic | 4,087,170 | MDQ6VXNlcjQwODcxNzA= | User | false |
huggingface/diffusers | 1,537,337,353 | I_kwDOHa8MBc5boegJ | 2,025 | https://github.com/huggingface/diffusers/issues/2025 | https://api.github.com/repos/huggingface/diffusers/issues/2025 | use a custom model in pipeline. unable to load directly by from_pretrained() | I define a new unet class (eg. CustomUnet)and a new pipeline(CustomPipeline) that is slightly different from StableDiffusionPipeline.
After training, I use `save_pretrained()` to save it and I find the `model_index` is like(`models` is where I put the unet I define, and but not in the system path? ):
```
CustomPipel... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-01-18T03:10:18Z | 2023-02-25T15:03:38Z | 2023-02-25T15:03:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Weifeng-Chen | 34,848,350 | MDQ6VXNlcjM0ODQ4MzUw | User | false |
huggingface/diffusers | 1,537,352,928 | I_kwDOHa8MBc5boiTg | 2,026 | https://github.com/huggingface/diffusers/issues/2026 | https://api.github.com/repos/huggingface/diffusers/issues/2026 | cpu_offload not work with PaintByExample pipeline | ### Describe the bug
According to https://github.com/huggingface/diffusers/issues/1934, when using `cpu_offload`, we should not manually call `.to("cuda")`. After remove `.to("cuda")`, paint_by_example pipeline yield following error:
<img width="1000" alt="image" src="https://user-images.githubusercontent.com/399... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-01-18T03:34:49Z | 2023-01-24T12:24:31Z | 2023-01-24T09:05:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Sanster | 3,998,421 | MDQ6VXNlcjM5OTg0MjE= | User | false |
huggingface/diffusers | 1,537,776,254 | I_kwDOHa8MBc5bqJp- | 2,028 | https://github.com/huggingface/diffusers/issues/2028 | https://api.github.com/repos/huggingface/diffusers/issues/2028 | Write a doc for centralizing our LoRA efforts | Currently, we have incoming support for LoRA to fine-tune with [DreamBooth](https://github.com/huggingface/diffusers/pull/1884) and [Text2Image](https://github.com/huggingface/diffusers/pull/2002).
These things enable the community to run fine-tuning on consumer GPUs which is a big enabling factor for many use case... | closed | completed | false | 2 | [] | [
"sayakpaul"
] | 2023-01-18T10:42:54Z | 2023-01-25T11:23:14Z | 2023-01-25T11:23:14Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,548,422,237 | I_kwDOHa8MBc5cSwxd | 2,035 | https://github.com/huggingface/diffusers/issues/2035 | https://api.github.com/repos/huggingface/diffusers/issues/2035 | tiled_upscaling outputs terrible artifacts, issues with launching it right | ### Describe the bug
I'm preparing pipeline for upscale this way:
```
pipe = StableDiffusionUpscalePipeline.from_pretrained("stabilityai/stable-diffusion-2-1",
use_auth_token="12345_token_4321",
local_... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-01-19T02:25:14Z | 2023-03-25T15:04:28Z | 2023-03-25T15:04:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Pirog17000 | 3,022,852 | MDQ6VXNlcjMwMjI4NTI= | User | false |
huggingface/diffusers | 1,548,444,937 | I_kwDOHa8MBc5cS2UJ | 2,036 | https://github.com/huggingface/diffusers/issues/2036 | https://api.github.com/repos/huggingface/diffusers/issues/2036 | OSError with stable-diffusion-v1-4 model | ### Describe the bug
When I load the stable diffusion v1-4 model, the following error occurs:
```
python my_script.py
2023-01-19 02:56:57.587154: W tensorflow/stream_executor/platform/default/dso_loader.cc:59] Could not load dynamic library 'libcudart.so.10.1'; dlerror: libcudart.so.10.1: cannot open shared o... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-01-19T03:04:11Z | 2023-11-24T00:05:19Z | 2023-01-19T14:46:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | liu00222 | 42,081,599 | MDQ6VXNlcjQyMDgxNTk5 | User | false |
huggingface/diffusers | 1,549,843,725 | I_kwDOHa8MBc5cYL0N | 2,041 | https://github.com/huggingface/diffusers/issues/2041 | https://api.github.com/repos/huggingface/diffusers/issues/2041 | Dreambooth example is not compatible with Windows | Same issue as described here: [https://github.com/huggingface/diffusers/issues/146](https://github.com/huggingface/diffusers/issues/1466) still occurring with main.
The fix from [https://github.com/huggingface/diffusers/pull/1547](https://github.com/huggingface/diffusers/pull/1547) does not completely fix it on Window... | closed | completed | false | 2 | [] | [
"patil-suraj"
] | 2023-01-19T19:47:12Z | 2023-05-03T03:05:44Z | 2023-01-23T09:58:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bbi-ai | 122,675,830 | U_kgDOB0_idg | User | false |
huggingface/diffusers | 1,550,353,902 | I_kwDOHa8MBc5caIXu | 2,044 | https://github.com/huggingface/diffusers/issues/2044 | https://api.github.com/repos/huggingface/diffusers/issues/2044 | AttributeError: 'DistributedDataParallel' object has no attribute 'dtype' | ### Describe the bug
While trying to launch the train script getting below error for the torch library
```
Traceback (most recent call last):
File "train_dreambooth.py", line 861, in <module>
main(args)
File "train_dreambooth.py", line 701, in main
if unet.dtype != torch.float32:
File "/home/ubu... | closed | completed | false | 2 | [
"bug"
] | [
"patil-suraj"
] | 2023-01-20T06:09:12Z | 2023-01-24T15:53:17Z | 2023-01-24T15:53:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chjose | 4,662,631 | MDQ6VXNlcjQ2NjI2MzE= | User | false |
huggingface/diffusers | 1,342,011,327 | I_kwDOHa8MBc5P_Xe_ | 205 | https://github.com/huggingface/diffusers/issues/205 | https://api.github.com/repos/huggingface/diffusers/issues/205 | Add noise interpolation and text intertpolation scripts to examples | Noise and text embedding interpolation scripts like @karpathy's [stable-diffusion-walk](https://gist.github.com/karpathy/00103b0037c5aaea32fe1da1af553355) are very nice to make artistic designs / videos and are relatively easy to add. We could add them under https://github.com/huggingface/diffusers/tree/main/examples t... | closed | completed | false | 5 | [
"enhancement"
] | [] | 2022-08-17T16:25:43Z | 2024-11-21T11:23:33Z | 2024-11-21T11:23:33Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,551,249,501 | I_kwDOHa8MBc5cdjBd | 2,052 | https://github.com/huggingface/diffusers/issues/2052 | https://api.github.com/repos/huggingface/diffusers/issues/2052 | Textual Inversion script does not resume from checkpoint | ### Describe the bug
The textual inversion script `textual_inversion.py` located in `examples/textual_inversion` does not seem to resume properly from a checkpoint. If I interrupt the script halfway through training and then try to resume with the `--resume_from_checkpoint` argument, the script runs to completion with... | closed | completed | false | 1 | [
"bug"
] | [
"pcuenca"
] | 2023-01-20T18:00:35Z | 2023-01-24T09:25:43Z | 2023-01-24T09:25:43Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lstein | 111,189 | MDQ6VXNlcjExMTE4OQ== | User | false |
huggingface/diffusers | 1,551,523,282 | I_kwDOHa8MBc5cel3S | 2,053 | https://github.com/huggingface/diffusers/issues/2053 | https://api.github.com/repos/huggingface/diffusers/issues/2053 | 0.12.0.dev0 consumes more VRAM than 0.10.2 | ### Describe the bug
0.10.2 handles resolutions up to 2500 on the long side of the image, while 0.12.0.dev0 can't handle even 1700 on the long side.
this is unexpected and I assume this is a bug. It began to happen as I updated on 0.11.1, and 0.12.0.dev0 didn't improve this.
### Reproduction
img2img pipeline, very ... | closed | completed | false | 11 | [
"bug"
] | [
"patrickvonplaten"
] | 2023-01-20T22:19:10Z | 2023-01-31T13:43:23Z | 2023-01-31T13:43:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Pirog17000 | 3,022,852 | MDQ6VXNlcjMwMjI4NTI= | User | false |
huggingface/diffusers | 1,551,786,783 | I_kwDOHa8MBc5cfmMf | 2,055 | https://github.com/huggingface/diffusers/issues/2055 | https://api.github.com/repos/huggingface/diffusers/issues/2055 | Multiple GPU problems? | I use U2Dmodel with nn.DataParallel
if torch.cuda.device_count() > 1:
model = nn.DataParallel(model)
Then I try to get the output.
model_output = model(noisy_images, timesteps)
The model _output is UNet2DOutput(sample=<generator object gather.<locals>.gather_map.<locals>.<genexpr> at 0x149714991620>)
mo... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-01-21T12:25:15Z | 2023-02-28T15:03:29Z | 2023-02-28T15:03:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | QishengL | 89,773,749 | MDQ6VXNlcjg5NzczNzQ5 | User | false |
huggingface/diffusers | 1,551,998,318 | I_kwDOHa8MBc5cgZ1u | 2,061 | https://github.com/huggingface/diffusers/issues/2061 | https://api.github.com/repos/huggingface/diffusers/issues/2061 | Add some documentation to the StableDiffusionImg2Img Pipeline tensor input format | So I was trying to feed the stablediffusionimg2imgpipeline with a video frame turned into torch tensor instead of a PIL image as is typical, and I was... struggling with garbled output, until I *eventually* found this bit in the code:
https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_d... | closed | completed | false | 9 | [
"wip"
] | [
"yiyixuxu"
] | 2023-01-22T04:08:45Z | 2024-11-28T12:59:43Z | 2024-11-24T03:28:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | brucethemoose | 8,422,224 | MDQ6VXNlcjg0MjIyMjQ= | User | false |
huggingface/diffusers | 1,552,362,061 | I_kwDOHa8MBc5chypN | 2,063 | https://github.com/huggingface/diffusers/issues/2063 | https://api.github.com/repos/huggingface/diffusers/issues/2063 | How to create videos with Stable Diffusion with high temporal cohesion? | I've been trying to work out how this is done?
https://www.youtube.com/watch?v=StpTUnYbxBg
https://www.youtube.com/watch?v=7ieodV1cvag
I can slowly change the latent space random noise to get slightly different images with a few jumps.
But I don't understand how this guy does the transformations as well such as... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-01-23T00:52:30Z | 2023-03-02T15:03:30Z | 2023-03-02T15:03:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | elephantpanda | 33,497,043 | MDQ6VXNlcjMzNDk3MDQz | User | false |
huggingface/diffusers | 1,552,372,305 | I_kwDOHa8MBc5ch1JR | 2,064 | https://github.com/huggingface/diffusers/issues/2064 | https://api.github.com/repos/huggingface/diffusers/issues/2064 | Switching to DPM++ SDE Karras sampling method | How can I change the scheduler in `StableDiffusionPipeline` to use what AUTOMATIC111's gui calls the "DPM++ SDE Karras" sampling method?
Also, whats the default sampling method being used by `StableDiffusionPipeline`? | closed | completed | false | 17 | [
"New scheduler"
] | [
"patrickvonplaten"
] | 2023-01-23T01:11:00Z | 2023-06-20T11:13:05Z | 2023-06-20T11:13:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | gameveloster | 103,383,490 | U_kgDOBimBwg | User | false |
huggingface/diffusers | 1,552,378,586 | I_kwDOHa8MBc5ch2ra | 2,065 | https://github.com/huggingface/diffusers/issues/2065 | https://api.github.com/repos/huggingface/diffusers/issues/2065 | Is num_inference_steps the same as Sampling Steps? | is `num_inference_steps` parameter of `StableDiffusionPipeline` refering to the same thing as in the "sampling steps" option in auto1111's?
I get much better results using 20 sampling steps in auto1111 than `num_inference_steps=20` in `StableDiffusionPipeline` | closed | completed | false | 2 | [
"stale"
] | [] | 2023-01-23T01:28:33Z | 2023-03-02T15:03:28Z | 2023-03-02T15:03:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | athenawisdoms | 6,615,285 | MDQ6VXNlcjY2MTUyODU= | User | false |
huggingface/diffusers | 1,552,424,747 | I_kwDOHa8MBc5ciB8r | 2,066 | https://github.com/huggingface/diffusers/issues/2066 | https://api.github.com/repos/huggingface/diffusers/issues/2066 | convert_stable_diffusion_checkpoint_to_onnx.py suddenly stops | ### Describe the bug
ONNX export script
convert_stable_diffusion_checkpoint_to_onnx.py
suddenly stops
### Reproduction
cd diffusers/src
#add src path
echo -e "import sys\nsys.path.append(\"../src\")" >tmp
cat ../scripts/convert_stable_diffusion_checkpoint_to_onnx.py >> tmp
mv tmp ../scripts/conver... | closed | completed | false | 3 | [
"bug",
"stale"
] | [
"anton-l"
] | 2023-01-23T02:44:07Z | 2023-03-07T15:03:42Z | 2023-03-07T15:03:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xiangze | 1,210,370 | MDQ6VXNlcjEyMTAzNzA= | User | false |
huggingface/diffusers | 1,552,510,914 | I_kwDOHa8MBc5ciW_C | 2,068 | https://github.com/huggingface/diffusers/issues/2068 | https://api.github.com/repos/huggingface/diffusers/issues/2068 | [Community] Specifying dtype in from_pretrained does not convert the model's datatype | ### Describe the bug
When I grab a model using "from_pretrained" and specifying the datatype, the returned model's parameters do not have that type.
This causes problems with flax text to image and flax dreambooth training scripts [here](https://github.com/huggingface/diffusers/blob/ce1c27adc85916a802db579df789d9... | open | null | false | 7 | [
"bug",
"Good second issue"
] | [] | 2023-01-23T04:45:37Z | 2024-11-21T15:34:00Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | roryoday | 51,141,557 | MDQ6VXNlcjUxMTQxNTU3 | User | false |
huggingface/diffusers | 1,553,087,704 | I_kwDOHa8MBc5ckjzY | 2,073 | https://github.com/huggingface/diffusers/issues/2073 | https://api.github.com/repos/huggingface/diffusers/issues/2073 | Using a Mask with the img2img Pipeline (perhaps as callback function?) | I am having a looped img2img pipeline to produce videos - similar to [deforum](https://github.com/deforum/deforum-stable-diffusion). For that, I would love some objects in the scene to remain unchanged. The way to do this, is to use a mask. With huggingface components, the only way I am aware of achieving this is the i... | closed | completed | false | 12 | [
"stale"
] | [
"patrickvonplaten"
] | 2023-01-23T13:19:42Z | 2023-03-19T10:02:05Z | 2023-03-06T15:03:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nicollegah | 55,440,280 | MDQ6VXNlcjU1NDQwMjgw | User | false |
huggingface/diffusers | 1,342,874,583 | I_kwDOHa8MBc5QCqPX | 208 | https://github.com/huggingface/diffusers/issues/208 | https://api.github.com/repos/huggingface/diffusers/issues/208 | Pipeline: obtain per-image reproducibility seeds when generating multiple images | People are interested in the following workflow: generate a few images from a prompt, select the one you like and tweak the prompt to steer it towards your goal. There are many examples in Twitter and the Discord bot, see context section below.
I managed to replicate this workflow in a demo Space generating one imag... | closed | completed | false | 8 | [] | [] | 2022-08-18T10:08:24Z | 2022-08-25T15:33:41Z | 2022-08-25T15:33:41Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pcuenca | 1,177,582 | MDQ6VXNlcjExNzc1ODI= | User | false |
huggingface/diffusers | 1,553,493,295 | I_kwDOHa8MBc5cmG0v | 2,081 | https://github.com/huggingface/diffusers/issues/2081 | https://api.github.com/repos/huggingface/diffusers/issues/2081 | Converting pytorch stable diffusion weights to jax doesn't always work | ### Describe the bug
I am trying to convert pytorch weights to jax as follows:
```python
from diffusers import FlaxStableDiffusionPipeline
model_name='riffusion/riffusion-model-v1'
pipeline, params = FlaxStableDiffusionPipeline.from_pretrained(model_name, from_pt=True)
pipeline.save_pretrained('riffusion_jax', ... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-01-23T17:35:18Z | 2023-01-24T19:07:33Z | 2023-01-24T19:07:32Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | entrpn | 6,580,675 | MDQ6VXNlcjY1ODA2NzU= | User | false |
huggingface/diffusers | 1,553,910,041 | I_kwDOHa8MBc5cnskZ | 2,082 | https://github.com/huggingface/diffusers/issues/2082 | https://api.github.com/repos/huggingface/diffusers/issues/2082 | Are all models fully versioned and deterministic? | Hello everyone,
I wanted to ask about some issues that have been popping up in libraries that use diffusers as a dependency. Specifically, there have been some unexpected changes to the output (check out this video: https://www.youtube.com/watch?v=usgqmQ0Mq7g) and I've also encountered similar problems when using a ... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-01-23T22:09:11Z | 2023-03-03T15:03:35Z | 2023-03-03T15:03:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | arpowers | 308,134 | MDQ6VXNlcjMwODEzNA== | User | false |
huggingface/diffusers | 1,553,974,627 | I_kwDOHa8MBc5cn8Vj | 2,083 | https://github.com/huggingface/diffusers/issues/2083 | https://api.github.com/repos/huggingface/diffusers/issues/2083 | Cannot use multiple generators and DDIMScheduler with eta > 0 | ### Describe the bug
As described in the call function of any pipeline (e.g., [pipeline_stable_diffusion_img2img.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py)), one may specify a single generator, or a list of generators:
https:... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-01-23T23:07:07Z | 2023-01-23T23:23:38Z | 2023-01-23T23:23:38Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xenova | 26,504,141 | MDQ6VXNlcjI2NTA0MTQx | User | false |
huggingface/diffusers | 1,554,195,177 | I_kwDOHa8MBc5coyLp | 2,084 | https://github.com/huggingface/diffusers/issues/2084 | https://api.github.com/repos/huggingface/diffusers/issues/2084 | [Dreambooth] ValueError: Unet loaded as datatype torch.float16 (should be torch.float32) | ### Describe the bug
Greetings,
I am following the Dog Toy example on an 8gb GPU,
(https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/README.md#training-on-a-8-gb-gpu)
I use `accelerate config` to configure DeepSpeed according to the example (see config below) then launch `train_dreamb... | closed | completed | false | 3 | [
"bug"
] | [
"patil-suraj"
] | 2023-01-24T02:34:02Z | 2023-01-25T19:08:47Z | 2023-01-25T19:06:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ArgeusDominguez | 42,524,141 | MDQ6VXNlcjQyNTI0MTQx | User | false |
huggingface/diffusers | 1,554,619,100 | I_kwDOHa8MBc5cqZrc | 2,087 | https://github.com/huggingface/diffusers/issues/2087 | https://api.github.com/repos/huggingface/diffusers/issues/2087 | Load safetensors directly on GPU | Currently, loading any diffusers pipeline with safetensors defaults to loading on CPU. I believe there is low hanging fruit in enabling direct loading to GPU (one of the key benefits of safetensors) in order to cut the step of moving of weights from CPU -> GPU. This will make pipeline initialization significantly faste... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-01-24T09:27:52Z | 2023-03-03T15:03:33Z | 2023-03-03T15:03:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jeanmichaeldiei | 14,225,675 | MDQ6VXNlcjE0MjI1Njc1 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.