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 | 1,599,723,852 | I_kwDOHa8MBc5fWdlM | 2,491 | https://github.com/huggingface/diffusers/issues/2491 | https://api.github.com/repos/huggingface/diffusers/issues/2491 | Stable Diffusion Text-to-image Finetuning is Broken | ### Describe the bug
I'm using https://github.com/huggingface/diffusers/tree/main/examples/text_to_image.py.
The training creates several checkpoints every 500 epochs, checkpoint-500, checkpoint-1000, etc. I load the diffusion_pytorch_model.bin from two checkpoints folders, and the weight values are equal after I... | closed | completed | false | 7 | [
"bug",
"stale"
] | [
"williamberman"
] | 2023-02-25T14:22:30Z | 2023-05-02T18:50:21Z | 2023-05-02T18:50:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xiankgx | 4,113,258 | MDQ6VXNlcjQxMTMyNTg= | User | false |
huggingface/diffusers | 1,599,770,808 | I_kwDOHa8MBc5fWpC4 | 2,493 | https://github.com/huggingface/diffusers/issues/2493 | https://api.github.com/repos/huggingface/diffusers/issues/2493 | [ONNX] Is it possible to load LORAs to be used in the ONNX pipelines? | A while ago I've asked about Textual Inversion embeddings and the answer was basically 'no'. There was a way around it but it wasn't an actual solution. However this was some time ago and things might have changed.
So, is it possible now? And if so is there any documentation on this matter I can read?
Sorry this is ... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-02-25T17:11:30Z | 2023-05-01T15:03:32Z | 2023-05-01T15:03:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | GreenLandisaLie | 80,653,235 | MDQ6VXNlcjgwNjUzMjM1 | User | false |
huggingface/diffusers | 1,599,852,578 | I_kwDOHa8MBc5fW9Ai | 2,495 | https://github.com/huggingface/diffusers/issues/2495 | https://api.github.com/repos/huggingface/diffusers/issues/2495 | MPS: StableDiffusionUpscalePipeline produces a black image | ### Describe the bug
Running the `StableDiffusionUpscalePipeline` on MPS produces a black (or sometimes gray) upscaled image.
### Reproduction
```python
from diffusers import StableDiffusionUpscalePipeline
from PIL import Image
# Name of the upscaling model to use
model_id = 'stabilityai/stable-diffusion-x4-up... | closed | completed | false | 5 | [
"bug",
"stale"
] | [
"pcuenca"
] | 2023-02-25T21:51:05Z | 2023-05-09T15:03:37Z | 2023-05-09T15:03:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Radagaisus | 550,061 | MDQ6VXNlcjU1MDA2MQ== | User | false |
huggingface/diffusers | 1,276,387,718 | I_kwDOHa8MBc5MFCGG | 25 | https://github.com/huggingface/diffusers/issues/25 | https://api.github.com/repos/huggingface/diffusers/issues/25 | Support batches in noise scheduler | Firstly, kudos to the HF team for building this awesome library!
I would like to propose some modifications to the DDPM noise scheduler (and potentially other schedulers) to support batching. Based on [`training_ddpm.py`](https://github.com/huggingface/diffusers/blob/main/examples/train_ddpm.py), it appears that the... | closed | completed | false | 2 | [] | [
"anton-l"
] | 2022-06-20T05:31:14Z | 2022-06-22T11:39:56Z | 2022-06-22T11:39:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jaketae | 25,360,440 | MDQ6VXNlcjI1MzYwNDQw | User | false |
huggingface/diffusers | 1,600,218,925 | I_kwDOHa8MBc5fYWct | 2,500 | https://github.com/huggingface/diffusers/issues/2500 | https://api.github.com/repos/huggingface/diffusers/issues/2500 | Error "NotImplementedError: Cannot copy out of meta tensor; no data!" on some tests | ### Describe the bug
I am hitting this exceptions on various tests:
```
self = <tests.pipelines.unclip.test_unclip.UnCLIPPipelineFastTests testMethod=test_cpu_offload_forward_pass>
@unittest.skipIf(
torch_device != "cuda" or not is_accelerate_available(),
reason="CPU offload is only availa... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-02-26T22:05:24Z | 2024-04-15T19:46:30Z | 2023-02-28T19:34:38Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | antoche | 1,627,384 | MDQ6VXNlcjE2MjczODQ= | User | false |
huggingface/diffusers | 1,600,228,822 | I_kwDOHa8MBc5fYY3W | 2,501 | https://github.com/huggingface/diffusers/issues/2501 | https://api.github.com/repos/huggingface/diffusers/issues/2501 | Multiple tests failing on v0.13.1 | ### Describe the bug
The following tests are failing on my system due to pipeline output differences:
- tests/models/test_models_unet_2d_condition.py::UNet2DConditionModelTests::test_lora_xformers_on_off
- tests/pipelines/stable_diffusion_2/test_stable_diffusion_depth.py::StableDiffusionDepth2ImgPipelineFastTests:... | closed | completed | false | 11 | [
"bug",
"wip"
] | [] | 2023-02-26T22:39:17Z | 2024-11-21T07:26:55Z | 2024-11-21T07:26:55Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | antoche | 1,627,384 | MDQ6VXNlcjE2MjczODQ= | User | false |
huggingface/diffusers | 1,601,139,626 | I_kwDOHa8MBc5fb3Oq | 2,505 | https://github.com/huggingface/diffusers/issues/2505 | https://api.github.com/repos/huggingface/diffusers/issues/2505 | First-time "warmup" pass for Apple Silicon (M1/M2) | As per the document "warm-up" pass is required for Apple silicon inference. Is this really needed in the diffuser 0.13.1? | closed | completed | false | 5 | [] | [] | 2023-02-27T12:59:46Z | 2023-03-05T11:15:14Z | 2023-03-05T11:15:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rupeshs | 3,255,994 | MDQ6VXNlcjMyNTU5OTQ= | User | false |
huggingface/diffusers | 1,601,451,991 | I_kwDOHa8MBc5fdDfX | 2,507 | https://github.com/huggingface/diffusers/issues/2507 | https://api.github.com/repos/huggingface/diffusers/issues/2507 | Loading safetensors on GPU is slower than when loading on CPU | ### Describe the bug
Loading models directly on GPU with `safetensors` is slower than when first loading on CPU and then on GPU.
Note that `diffusers` depends on both `transformers` and `accelerate` for fast loading so the problem might very well be there.
**Note**:
Also please make sure to use this branch: ht... | closed | completed | false | 9 | [
"bug",
"stale"
] | [] | 2023-02-27T15:51:20Z | 2024-11-16T00:55:08Z | 2023-04-10T15:03:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,351,013,068 | I_kwDOHa8MBc5QhtLM | 251 | https://github.com/huggingface/diffusers/issues/251 | https://api.github.com/repos/huggingface/diffusers/issues/251 | Extreme fragility in latents distribution (solid color image results) | ### Describe the bug
To reproduce, one can replace lines 102-106 of [stable diffusers pipeline](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py) with a tensor of non-noise data.
More succinctly, this works as expected (hard coding the defaults... | closed | completed | false | 2 | [
"bug"
] | [] | 2022-08-25T14:42:07Z | 2022-09-13T15:55:58Z | 2022-09-13T15:55:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rseger | 2,055,031 | MDQ6VXNlcjIwNTUwMzE= | User | false |
huggingface/diffusers | 1,601,678,174 | I_kwDOHa8MBc5fd6te | 2,510 | https://github.com/huggingface/diffusers/issues/2510 | https://api.github.com/repos/huggingface/diffusers/issues/2510 | Standardizing batching logic for pipelines which take both images and text prompts | In the stable unclip img2img pipeline, I follow the standard batching logic for the prompts. If a single image is passed, I duplicate batch_size * num_images_per_prompt times. If multiple images are passed, I only support the same number of prompts and images passed, so I duplicate each image num_images_per_prompt time... | closed | not_planned | false | 3 | [
"stale"
] | [] | 2023-02-27T17:58:42Z | 2023-04-05T17:48:08Z | 2023-04-05T17:47:59Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,602,087,417 | I_kwDOHa8MBc5ffen5 | 2,512 | https://github.com/huggingface/diffusers/issues/2512 | https://api.github.com/repos/huggingface/diffusers/issues/2512 | Ft | null | closed | completed | false | 1 | [] | [] | 2023-02-27T23:03:37Z | 2023-02-28T09:54:44Z | 2023-02-28T09:54:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | KeazyKeys | 91,432,971 | MDQ6VXNlcjkxNDMyOTcx | User | false |
huggingface/diffusers | 1,602,683,404 | I_kwDOHa8MBc5fhwIM | 2,513 | https://github.com/huggingface/diffusers/issues/2513 | https://api.github.com/repos/huggingface/diffusers/issues/2513 | Is there exists a method to use unet.set_attn_processor from a config file with its network definitions ? | Huggingface transformers use auto_map in config file to allow users define their own network(by script)
and load them by Huggingface's interface. Because models such as "ControlNet" or "ControlLoRA" and many so on are an edit version by change their attention block of Unet.
Load from config and use "unet.set_attn_pr... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-02-28T09:39:40Z | 2023-04-09T15:03:04Z | 2023-04-09T15:03:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | svjack | 27,874,014 | MDQ6VXNlcjI3ODc0MDE0 | User | false |
huggingface/diffusers | 1,602,757,900 | I_kwDOHa8MBc5fiCUM | 2,514 | https://github.com/huggingface/diffusers/issues/2514 | https://api.github.com/repos/huggingface/diffusers/issues/2514 | Loading cached pipeline takes too long | ### Describe the bug
When loading a pipeline that is cached lot of time is wasted by calling the Hub unnecessarily. We should try to avoid this by trying to call the Hub only once if the model is cached.
See [internal slack thread](https://huggingface.slack.com/archives/C03V11RNS7P/p1677521547429829) for more con... | closed | completed | false | 15 | [
"bug"
] | [] | 2023-02-28T10:22:43Z | 2023-04-11T05:26:25Z | 2023-04-10T09:34:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,603,663,256 | I_kwDOHa8MBc5flfWY | 2,518 | https://github.com/huggingface/diffusers/issues/2518 | https://api.github.com/repos/huggingface/diffusers/issues/2518 | `examples/unconditional_image_generation/train_unconditional.py` missing `xformers` support | **Is your feature request related to a problem? Please describe.**
I want the ability to use `xformers` in the unconditional generation script.
**Describe the solution you'd like**
Just a copy paste job from the other scripts, to use a flag to turn `xformers` on. The UNet in the file inherits from `ModelMixin` so ... | closed | completed | false | 2 | [] | [] | 2023-02-28T19:14:45Z | 2023-03-05T21:20:18Z | 2023-03-05T21:20:18Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vvvm23 | 44,398,246 | MDQ6VXNlcjQ0Mzk4MjQ2 | User | false |
huggingface/diffusers | 1,603,672,430 | I_kwDOHa8MBc5flhlu | 2,519 | https://github.com/huggingface/diffusers/issues/2519 | https://api.github.com/repos/huggingface/diffusers/issues/2519 | v_prediction high loss | Hello, i recently tried dreambooth fine-tuning using v_prediction instead of epsilon. I noticed that the loss was much higher with this method. With epsilon the loss is usually steady between 0.25 - 0.3 but with v_prediction it went up to 1.5 - 2, eventually dropping but not any lower than 0.8. Just wondering if this i... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-02-28T19:22:04Z | 2023-04-08T15:02:56Z | 2023-04-08T15:02:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rmac85 | 90,567,763 | MDQ6VXNlcjkwNTY3NzYz | User | false |
huggingface/diffusers | 1,603,858,787 | I_kwDOHa8MBc5fmPFj | 2,521 | https://github.com/huggingface/diffusers/issues/2521 | https://api.github.com/repos/huggingface/diffusers/issues/2521 | MPS fails with float16 and PyTorch 2.0 | ### Describe the bug
I don't know if this can be fixed in diffusers or if it's a PyTorch or MPS issue, but if you take the basic diffusers example from https://huggingface.co/docs/diffusers/optimization/mps and change the type to float16, like this:
```
# make sure you're logged in with `huggingface-cli login`
... | closed | completed | false | 8 | [
"bug",
"stale"
] | [
"pcuenca"
] | 2023-02-28T21:40:26Z | 2023-04-12T15:03:22Z | 2023-04-12T15:03:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mannby | 486,592 | MDQ6VXNlcjQ4NjU5Mg== | User | false |
huggingface/diffusers | 1,603,960,554 | I_kwDOHa8MBc5fmn7q | 2,522 | https://github.com/huggingface/diffusers/issues/2522 | https://api.github.com/repos/huggingface/diffusers/issues/2522 | [Models] [Pipelines] Patch Diffusion | ### Model/Pipeline/Scheduler description
The core idea of the paper is to insert a ViT-style patching operation at the beginning of the U-Net, letting it operate on data with smaller height and width. We show in our paper that the optimal prediction for x is quite blurry for most timesteps, and therefore convolutions ... | closed | completed | false | 2 | [
"New pipeline/model"
] | [] | 2023-02-28T23:16:04Z | 2024-11-21T07:27:31Z | 2024-11-21T07:27:30Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | clarencechen | 8,482,341 | MDQ6VXNlcjg0ODIzNDE= | User | false |
huggingface/diffusers | 1,604,305,580 | I_kwDOHa8MBc5fn8Ks | 2,523 | https://github.com/huggingface/diffusers/issues/2523 | https://api.github.com/repos/huggingface/diffusers/issues/2523 | Unconditioned latent diffusion model training on my own datasets | Recently I found diffusers repo, which is amzing.
I am intend to use this repo to realize my own training model using unconditioned latent diffusion model.
However, as a freshman, I have no idea where I should start with. I have checked the codes in the examples, but I don't find a suitable one.
Much apprecia... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-03-01T06:02:22Z | 2023-04-08T15:02:54Z | 2023-04-08T15:02:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhi-xuan-chen | 69,074,063 | MDQ6VXNlcjY5MDc0MDYz | User | false |
huggingface/diffusers | 1,604,997,986 | I_kwDOHa8MBc5fqlNi | 2,525 | https://github.com/huggingface/diffusers/issues/2525 | https://api.github.com/repos/huggingface/diffusers/issues/2525 | Initialisation of models takes long time | Hey,
I am running a web service based on Stable Diffusion. Somehow, when i launch a new instance, initialisation of stable diffusion takes about 15 minutes. For me this is critical, as I am offering 5 different models and thus start-up time becomes significant.
Basically i am running:
```
pipe = StableDiff... | closed | completed | false | 4 | [] | [] | 2023-03-01T13:42:21Z | 2023-03-05T12:27:05Z | 2023-03-05T12:27:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nicollegah | 55,440,280 | MDQ6VXNlcjU1NDQwMjgw | User | false |
huggingface/diffusers | 1,606,589,602 | I_kwDOHa8MBc5fwpyi | 2,531 | https://github.com/huggingface/diffusers/issues/2531 | https://api.github.com/repos/huggingface/diffusers/issues/2531 | Using lpw_stable_diffusion with sequential_cpu_offload gives 'NotImplementedError: Cannot copy out of meta tensor; no data!' | ### Describe the bug
Trying to use the example code from https://huggingface.co/docs/diffusers/using-diffusers/custom_pipeline_examples
with cpu offloading as suggested at https://huggingface.co/docs/diffusers/optimization/fp16#offloading-to-cpu-with-accelerate-for-memory-savings
I have removed the `pipe.to("cuda")`... | closed | completed | false | 11 | [
"bug",
"stale"
] | [] | 2023-03-02T11:02:52Z | 2023-04-24T15:03:45Z | 2023-04-24T15:03:45Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | d1g1t | 661,828 | MDQ6VXNlcjY2MTgyOA== | User | false |
huggingface/diffusers | 1,606,670,116 | I_kwDOHa8MBc5fw9ck | 2,532 | https://github.com/huggingface/diffusers/issues/2532 | https://api.github.com/repos/huggingface/diffusers/issues/2532 | Bug with `get_views` method of `StableDiffusionPanoramaPipeline` | ### Describe the bug
The `get_views` method of `StableDiffusionPanoramaPipeline` lacks proper bounds checking when `window_size` is large.
I think I can make time to submit a PR for this.
### Reproduction
```
# Create model
model_name_or_path = "runwayml/stable-diffusion-v1-5"
model = StableDiffusionPano... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-03-02T11:59:21Z | 2023-04-10T15:03:13Z | 2023-04-10T15:03:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lukemelas | 13,307,440 | MDQ6VXNlcjEzMzA3NDQw | User | false |
huggingface/diffusers | 1,607,375,956 | I_kwDOHa8MBc5fzpxU | 2,534 | https://github.com/huggingface/diffusers/issues/2534 | https://api.github.com/repos/huggingface/diffusers/issues/2534 | Changing the scheduler from UniPC or DEIS to DPM or DPM multistep results in error | ### Describe the bug
Hello there!
Whenever I try to change the scheduler from UniPC or DEIS to DPM or DPM multistep I get the following error:
UnboundLocalError: local variable 'x_t' referenced before assignment
full error:
```
Traceback (most recent call last):
File "E:\sd\output_gui2\sdiff_GUI.py", line ... | closed | completed | false | 8 | [
"bug"
] | [] | 2023-03-02T19:31:59Z | 2023-03-19T14:10:53Z | 2023-03-09T13:17:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | brurpo | 24,902,022 | MDQ6VXNlcjI0OTAyMDIy | User | false |
huggingface/diffusers | 1,607,867,314 | I_kwDOHa8MBc5f1huy | 2,538 | https://github.com/huggingface/diffusers/issues/2538 | https://api.github.com/repos/huggingface/diffusers/issues/2538 | enable train_text_encoder option in train_dreambooth_lora.py | **Is your feature request related to a problem? Please describe.**
The current [train_dreambooth_lora.py](https://github.com/huggingface/diffusers/blob/8f21a9f0e29cc5e3d139095785d52ff474ebda0e/examples/dreambooth/train_dreambooth_lora.py#L667) does not support train_text_encoder option
**Describe the solution you'... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-03-03T03:07:29Z | 2023-04-10T15:03:12Z | 2023-04-10T15:03:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | harrywang | 595,772 | MDQ6VXNlcjU5NTc3Mg== | User | false |
huggingface/diffusers | 1,609,401,366 | I_kwDOHa8MBc5f7YQW | 2,548 | https://github.com/huggingface/diffusers/issues/2548 | https://api.github.com/repos/huggingface/diffusers/issues/2548 | Saving error finetuning stable diffusion LoRA | ### Describe the bug
After the stable diffusion model is fully trained, an error occurs:
```
Traceback (most recent call last):
File "train_text_to_image_lora.py", line 872, in <module>
main()
File "train_text_to_image_lora.py", line 825, in main
unet.save_attn_procs(args.output_dir)
File "/usr/... | closed | completed | false | 17 | [
"bug"
] | [
"sayakpaul"
] | 2023-03-03T23:26:07Z | 2023-03-17T21:40:30Z | 2023-03-17T21:40:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sr5434 | 118,690,585 | U_kgDOBxMTGQ | User | false |
huggingface/diffusers | 1,609,543,832 | I_kwDOHa8MBc5f77CY | 2,549 | https://github.com/huggingface/diffusers/issues/2549 | https://api.github.com/repos/huggingface/diffusers/issues/2549 | SD 2.0 Based UNet2DConditionModel doesn't work | ### Describe the bug
I am trying to use a SD 2.x based model by loading it as UNet2DConditionModel just like the pipeline does. However upon inference I am met with the following error:
```python
Traceback (most recent call last):
File "/workspace/node/threads/base.py", line 215, in image_generator
noise_p... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-03-04T02:21:38Z | 2023-04-11T15:03:32Z | 2023-04-11T15:03:32Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chavinlo | 85,657,083 | MDQ6VXNlcjg1NjU3MDgz | User | false |
huggingface/diffusers | 1,352,089,190 | I_kwDOHa8MBc5Qlz5m | 255 | https://github.com/huggingface/diffusers/issues/255 | https://api.github.com/repos/huggingface/diffusers/issues/255 | [StableDiffusionPipeline] Image height and width both divisible by 8, throws error in unet_blocks | ### Describe the bug
Dear diffusers team,
I encountered a bug where the image width and height are both divisible by 8, but the prompt results in the following runtime error:
RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 18 but got size 17 for tensor number 1 in the list.
Thi... | closed | completed | false | 2 | [
"bug"
] | [
"patil-suraj"
] | 2022-08-26T10:40:24Z | 2022-10-06T11:27:23Z | 2022-10-06T11:27:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | SebOchs | 48,994,875 | MDQ6VXNlcjQ4OTk0ODc1 | User | false |
huggingface/diffusers | 1,609,642,106 | I_kwDOHa8MBc5f8TB6 | 2,550 | https://github.com/huggingface/diffusers/issues/2550 | https://api.github.com/repos/huggingface/diffusers/issues/2550 | Can't convert from controlNet model to diffusers | ### Describe the bug
I was able to do so before the PR merge, when it was being developed.
now it throws the following error:
```
Traceback (most recent call last):
File "E:\BRP_STABLE_DIFFUSION\bin\convert_original_stable_diffusion_to_diffusers.py", line 129, in <module>
pipe = load_pipeline_from_origi... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-03-04T06:43:26Z | 2023-03-07T05:27:05Z | 2023-03-07T05:27:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | brurpo | 24,902,022 | MDQ6VXNlcjI0OTAyMDIy | User | false |
huggingface/diffusers | 1,609,844,084 | I_kwDOHa8MBc5f9EV0 | 2,552 | https://github.com/huggingface/diffusers/issues/2552 | https://api.github.com/repos/huggingface/diffusers/issues/2552 | About the StableDiffusionInpaintPipeline model tutorial | ### Describe the bug
There is no Stable Diffusion Inpaint Pipeline class in the model_index.json file.
### Reproduction
Train:
```
!accelerate launch train_dreambooth_inpaint.py \
--pretraine... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-03-04T16:13:47Z | 2023-04-13T15:03:23Z | 2023-04-13T15:03:23Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kadirnar | 36,204,372 | MDQ6VXNlcjM2MjA0Mzcy | User | false |
huggingface/diffusers | 1,609,967,484 | I_kwDOHa8MBc5f9id8 | 2,554 | https://github.com/huggingface/diffusers/issues/2554 | https://api.github.com/repos/huggingface/diffusers/issues/2554 | StableDiffusionPipeline requires unreleased accelerate version (0.17.0) | ### Describe the bug
`StableDiffusionPipeline.enable_model_cpu_offload()` requires accelerate **0.17.0**:
https://github.com/huggingface/diffusers/blob/2ea1da89ab42bde69caad909048925ca99873400/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py#L237-L240
However, right now, the latest release... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-03-04T22:14:40Z | 2023-06-19T10:38:52Z | 2023-03-10T20:03:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ernesto-elsaesser | 2,370,583 | MDQ6VXNlcjIzNzA1ODM= | User | false |
huggingface/diffusers | 1,610,088,529 | I_kwDOHa8MBc5f-ABR | 2,556 | https://github.com/huggingface/diffusers/issues/2556 | https://api.github.com/repos/huggingface/diffusers/issues/2556 | Add Multi-ControlNet pipeline | Discussed in #2331. I created a PoC that supports multiple ControlNets, called Multi-ControlNet, based on the `StableDiffusionControlNetPipeline`. Any feedback is appreciated.
https://github.com/takuma104/diffusers/tree/multi_controlnet
The idea of using multiple ControlNets and adding their outputs together was pr... | closed | completed | false | 27 | [] | [
"williamberman"
] | 2023-03-05T07:24:04Z | 2024-09-25T05:21:57Z | 2023-03-16T10:38:15Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | takuma104 | 10,776 | MDQ6VXNlcjEwNzc2 | User | false |
huggingface/diffusers | 1,610,714,688 | I_kwDOHa8MBc5gAY5A | 2,560 | https://github.com/huggingface/diffusers/issues/2560 | https://api.github.com/repos/huggingface/diffusers/issues/2560 | StableDiffusionImg2ImgPipeline creates different outputs when using a list of generators | ### Describe the bug
When passing a list of generators, `StableDiffusionImg2ImgPipeline` creates different images based on the position of the generator in the list.
The reproduction code generates 4 completely different images even though only two seeds were passed, but in the different order. It is however consis... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-03-06T05:48:46Z | 2023-03-07T14:18:32Z | 2023-03-07T14:18:31Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | d1g1t | 661,828 | MDQ6VXNlcjY2MTgyOA== | User | false |
huggingface/diffusers | 1,610,985,489 | I_kwDOHa8MBc5gBbAR | 2,562 | https://github.com/huggingface/diffusers/issues/2562 | https://api.github.com/repos/huggingface/diffusers/issues/2562 | Adding support for ControlNet pipeline for img2img flows | It would be nice to achieve a somewhat feature parity with automatic1111 in terms using controlnet with img2img flows.
Current ControlNet pipeline doesn't allow an input source image (if I understand correctly the image parameter goes to controlnet as condition and not as src image for diffusion).
I think I might t... | closed | completed | false | 6 | [] | [] | 2023-03-06T09:17:51Z | 2023-03-08T13:19:38Z | 2023-03-08T13:19:38Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mikegarts | 3,885,814 | MDQ6VXNlcjM4ODU4MTQ= | User | false |
huggingface/diffusers | 1,611,262,165 | I_kwDOHa8MBc5gCejV | 2,564 | https://github.com/huggingface/diffusers/issues/2564 | https://api.github.com/repos/huggingface/diffusers/issues/2564 | Unclear workflow for Textual Inversion | Hi All!
First of all, thank you so much for the amazing work with `diffusers`. I have literally been able to pickup Stable Diffusion and grasp different concepts to a great extent in under a week thanks to it!
## Cannot load in custom Textual Inversion models
I've been unable to make `diffusers` work with Text... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-03-06T12:03:12Z | 2023-03-07T12:29:40Z | 2023-03-07T12:29:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | init27 | 16,226,196 | MDQ6VXNlcjE2MjI2MTk2 | User | false |
huggingface/diffusers | 1,611,996,151 | I_kwDOHa8MBc5gFRv3 | 2,568 | https://github.com/huggingface/diffusers/issues/2568 | https://api.github.com/repos/huggingface/diffusers/issues/2568 | Confusion whether autocast(...) should be used or not? | Copy-paste the very nice comment/question from PrinceRay7 from here: https://github.com/huggingface/diffusers/issues/993#issuecomment-1446588265
> I'm quite confused. In Diffusers docs, it is strongly disencouraged to use torch.autocast, however, in accelerate scripts, when training with the flag fp16, it is still i... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-03-06T18:28:23Z | 2023-04-14T15:03:22Z | 2023-04-14T15:03:22Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,612,095,704 | I_kwDOHa8MBc5gFqDY | 2,571 | https://github.com/huggingface/diffusers/issues/2571 | https://api.github.com/repos/huggingface/diffusers/issues/2571 | ImportError: cannot import name 'ProjectConfiguration' from 'accelerate.utils' | ### Describe the bug
Train Dreambooth fails after upgrade with ImportError: cannot import name 'ProjectConfiguration' from 'accelerate.utils'
### Reproduction
Latest version
git clone https://github.com/huggingface/diffusers
cd diffusers
pip install -e .
### Logs
```shell
To avoid this warning pass in ... | closed | completed | false | 6 | [
"bug"
] | [
"williamberman"
] | 2023-03-06T19:45:07Z | 2023-04-11T17:13:00Z | 2023-04-11T17:12:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bluusun | 65,051,271 | MDQ6VXNlcjY1MDUxMjcx | User | false |
huggingface/diffusers | 1,612,103,324 | I_kwDOHa8MBc5gFr6c | 2,572 | https://github.com/huggingface/diffusers/issues/2572 | https://api.github.com/repos/huggingface/diffusers/issues/2572 | Non Huggingface models from Dreambooth (i.e. civit.ai) | What's the workflow (if it exists) for non Hugginface models?
I downloaded and used the local ckpt file but got this error (no JSON conf files are provided for most civit.ai models)
Traceback (most recent call last): ... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-03-06T19:51:25Z | 2023-04-14T15:03:21Z | 2023-04-14T15:03:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bluusun | 65,051,271 | MDQ6VXNlcjY1MDUxMjcx | User | false |
huggingface/diffusers | 1,612,148,623 | I_kwDOHa8MBc5gF2-P | 2,573 | https://github.com/huggingface/diffusers/issues/2573 | https://api.github.com/repos/huggingface/diffusers/issues/2573 | Precommit hooks for contributors | **Is your feature request related to a problem? Please describe.**
Not really a problem but just a quality of life suggestion.
I noticed PR #2524 which simply updated copyrights. It got me thinking that this could be quite easily automated to trigger for every new file added in commits using git pre-commit hooks. I... | closed | completed | false | 2 | [] | [] | 2023-03-06T20:24:07Z | 2023-03-07T15:10:28Z | 2023-03-07T15:10:28Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vvvm23 | 44,398,246 | MDQ6VXNlcjQ0Mzk4MjQ2 | User | false |
huggingface/diffusers | 1,612,352,526 | I_kwDOHa8MBc5gGowO | 2,575 | https://github.com/huggingface/diffusers/issues/2575 | https://api.github.com/repos/huggingface/diffusers/issues/2575 | dreambooth training issue | ### Describe the bug
I tried to run this example about dreambooth https://huggingface.co/docs/diffusers/main/en/training/dreambooth#finetuning
but for `triton` package, it crashes every time.
### Reproduction
```
import diffusers
or
import triton
```
### Logs
```shell
in <module> ... | closed | completed | false | 7 | [
"bug"
] | [
"yiyixuxu"
] | 2023-03-06T22:50:47Z | 2023-06-20T21:38:08Z | 2023-03-28T13:16:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pure-rgb | 45,315,076 | MDQ6VXNlcjQ1MzE1MDc2 | User | false |
huggingface/diffusers | 1,612,419,585 | I_kwDOHa8MBc5gG5IB | 2,577 | https://github.com/huggingface/diffusers/issues/2577 | https://api.github.com/repos/huggingface/diffusers/issues/2577 | I don't get DDIMInverseScheduler working | ### Describe the bug
I want to try the StableDiffusionPix2PixZeroPipeline with Stable Diffusion v2.1. However, when I run the code from the example (https://huggingface.co/docs/diffusers/api/pipelines/stable_diffusion/pix2pix_zero) I just get noisy images back.
So I tried something simpler and just want to get ... | closed | completed | false | 7 | [
"bug",
"good first issue"
] | [] | 2023-03-06T23:49:33Z | 2023-10-30T03:23:06Z | 2023-04-12T07:27:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kaibioinfo | 1,015,048 | MDQ6VXNlcjEwMTUwNDg= | User | false |
huggingface/diffusers | 1,612,640,389 | I_kwDOHa8MBc5gHvCF | 2,578 | https://github.com/huggingface/diffusers/issues/2578 | https://api.github.com/repos/huggingface/diffusers/issues/2578 | Installing diffusers in editable mode does not work | ### Describe the bug
I'm trying to install diffusers in editable mode in a Colab/Kaggle notebook, following https://huggingface.co/docs/diffusers/installation . It installs successfully, but then I cannot import the package unless I use import src.diffusers .
Only advise I could find on this was to add the package ... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-03-07T03:18:03Z | 2023-04-14T15:03:20Z | 2023-04-14T15:03:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | generatorman | 127,107,642 | U_kgDOB5OCOg | User | false |
huggingface/diffusers | 1,612,647,040 | I_kwDOHa8MBc5gHwqA | 2,579 | https://github.com/huggingface/diffusers/issues/2579 | https://api.github.com/repos/huggingface/diffusers/issues/2579 | possible xFormers bug or misuse | ### Describe the bug
providing a mask seems to produce weird errors unless, that's not how a mask is supposed to be passed?
### Reproduction
```python
from diffusers.models.cross_attention import XFormersCrossAttnProcessor, CrossAttention
hidden_states = torch.randn(32, 30, 16).cuda()
attn_mask = torch.zeros((3... | closed | completed | false | 3 | [
"bug",
"stale"
] | [
"pcuenca",
"patil-suraj"
] | 2023-03-07T03:25:44Z | 2023-04-14T15:03:18Z | 2023-04-14T15:03:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Infrared1029 | 60,873,139 | MDQ6VXNlcjYwODczMTM5 | User | false |
huggingface/diffusers | 1,612,674,813 | I_kwDOHa8MBc5gH3b9 | 2,580 | https://github.com/huggingface/diffusers/issues/2580 | https://api.github.com/repos/huggingface/diffusers/issues/2580 | How to use schedulers in numpy mode? | Initially you used to have a tensor format = numpy.
but now it looks like it only supports pytorch. why??
How to use it for numpy? | closed | completed | false | 6 | [
"stale"
] | [] | 2023-03-07T03:54:24Z | 2023-04-14T15:03:17Z | 2023-04-14T15:03:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghost | 10,137 | MDQ6VXNlcjEwMTM3 | User | false |
huggingface/diffusers | 1,612,994,714 | I_kwDOHa8MBc5gJFia | 2,581 | https://github.com/huggingface/diffusers/issues/2581 | https://api.github.com/repos/huggingface/diffusers/issues/2581 | Support to convert between ControlNet 2.1 checkpoints and diffusers | ### Describe the bug
I'm unable to convert a 2.1 trained ControlNet model using `scripts/convert_original_stable_diffusion_to_diffusers.py`
The ControlNet models in question are here:
https://huggingface.co/furusu/ControlNet
It would seem this ControlNet models have been created by subtracting them out from t... | closed | completed | false | 39 | [
"bug"
] | [
"williamberman"
] | 2023-03-07T08:54:05Z | 2023-07-24T02:48:28Z | 2023-03-10T16:25:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tonetechnician | 13,471,797 | MDQ6VXNlcjEzNDcxNzk3 | User | false |
huggingface/diffusers | 1,613,010,462 | I_kwDOHa8MBc5gJJYe | 2,582 | https://github.com/huggingface/diffusers/issues/2582 | https://api.github.com/repos/huggingface/diffusers/issues/2582 | Language support and contribution to docs | I would like to contribute to translating the documentation into zh-CN.
I want to translate the chapter GET STARTED.🧐 | closed | completed | false | 2 | [
"stale"
] | [] | 2023-03-07T09:04:13Z | 2023-04-14T15:03:16Z | 2023-04-14T15:03:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LolitaSian | 34,176,474 | MDQ6VXNlcjM0MTc2NDc0 | User | false |
huggingface/diffusers | 1,613,331,133 | I_kwDOHa8MBc5gKXq9 | 2,585 | https://github.com/huggingface/diffusers/issues/2585 | https://api.github.com/repos/huggingface/diffusers/issues/2585 | DDIM timesteps function. | ### Describe the bug
DDIM scheduler uses wrong timesteps for sampling.
For example, given 50 steps and 1 steps offset gives timesteps [981, 961, ..., 1].
However, start of the diffusion process should be at timestep 999.
The same issue appears in [Stabilityai/Stab](https://github.com/Stability-AI/stablediffusion)... | closed | completed | false | 9 | [
"bug"
] | [] | 2023-03-07T12:19:54Z | 2023-06-02T13:26:28Z | 2023-04-27T01:01:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | GLivshits | 47,557,983 | MDQ6VXNlcjQ3NTU3OTgz | User | false |
huggingface/diffusers | 1,613,972,891 | I_kwDOHa8MBc5gM0Wb | 2,589 | https://github.com/huggingface/diffusers/issues/2589 | https://api.github.com/repos/huggingface/diffusers/issues/2589 | Add CI check for if deps table wasn't updated | Our CI should check for if the deps table wasn't updated after a dependency update in setup.py
```sh
$ make deps_table_update
$ git diff --exit-code
```
| closed | completed | false | 0 | [] | [
"williamberman"
] | 2023-03-07T18:17:27Z | 2023-03-07T23:24:52Z | 2023-03-07T23:24:52Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,352,447,683 | I_kwDOHa8MBc5QnLbD | 259 | https://github.com/huggingface/diffusers/issues/259 | https://api.github.com/repos/huggingface/diffusers/issues/259 | Fix progress bar in Stable Diffusion pipeline | ### Describe the bug
Generating images with StableDiffusionPipeline does not display the total number of iterations because of `tqdm` and `enumerate` being swapped in the [code](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py#L136).
This is a ... | closed | completed | false | 0 | [
"bug"
] | [] | 2022-08-26T15:54:45Z | 2022-08-30T10:30:07Z | 2022-08-30T10:30:07Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | neverix | 46,641,404 | MDQ6VXNlcjQ2NjQxNDA0 | User | false |
huggingface/diffusers | 1,614,144,924 | I_kwDOHa8MBc5gNeWc | 2,594 | https://github.com/huggingface/diffusers/issues/2594 | https://api.github.com/repos/huggingface/diffusers/issues/2594 | Diffusers GUI | **Is your feature request related to a problem? Please describe.**
No
**Describe the solution you'd like**
A gui to easily access the diffusers capabilities
**Describe alternatives you've considered**
InvokeAI have some drawbacks and do not leverage all that diffusers has to offer.
Automatic1111 is not diffus... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-03-07T20:16:55Z | 2023-04-15T15:03:31Z | 2023-04-15T15:03:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | brurpo | 24,902,022 | MDQ6VXNlcjI0OTAyMDIy | User | false |
huggingface/diffusers | 1,614,344,335 | I_kwDOHa8MBc5gOPCP | 2,596 | https://github.com/huggingface/diffusers/issues/2596 | https://api.github.com/repos/huggingface/diffusers/issues/2596 | Documentation contains non-existing load_image in diffusers.utils | ### Describe the bug
The documentation contains the code which doesn't work.
https://huggingface.co/docs/diffusers/api/pipelines/stable_diffusion/controlnet
### Reproduction
from diffusers.utils import load_image
### Logs
```shell
----------------------------------------------------------------------... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-03-07T23:01:33Z | 2023-04-15T15:03:30Z | 2023-04-15T15:03:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | belyak | 2,033,399 | MDQ6VXNlcjIwMzMzOTk= | User | false |
huggingface/diffusers | 1,614,642,099 | I_kwDOHa8MBc5gPXuz | 2,599 | https://github.com/huggingface/diffusers/issues/2599 | https://api.github.com/repos/huggingface/diffusers/issues/2599 | Dreambooth multi-subjects training makes the results quality worse | ### Overview
**I finetuned from stable diffusion v1.5 with dreambooth method, it worked fine when I trained single subject: car/dog. But the results quality decreased when I trained both categories at the same time with script: [train_multi_subject_dreambooth.py](https://github.com/huggingface/diffusers/blob/main/exam... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-03-08T04:38:12Z | 2023-04-15T15:03:28Z | 2023-04-15T15:03:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lifang1996 | 46,549,740 | MDQ6VXNlcjQ2NTQ5NzQw | User | false |
huggingface/diffusers | 1,614,725,719 | I_kwDOHa8MBc5gPsJX | 2,601 | https://github.com/huggingface/diffusers/issues/2601 | https://api.github.com/repos/huggingface/diffusers/issues/2601 | runwayml/stable-diffusion-v1-5 is missing fp16 files | When loading sd v1.5 with the command below, I received the warning letting me submit this issue. Thanks.
```
sd_v1_5 = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16, revision="fp16")
```
```
FutureWarning: You are loading the variant fp16 from runwayml/st... | closed | completed | false | 0 | [] | [] | 2023-03-08T06:24:58Z | 2023-03-08T08:47:37Z | 2023-03-08T08:47:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | haotian-liu | 6,631,389 | MDQ6VXNlcjY2MzEzODk= | User | false |
huggingface/diffusers | 1,615,071,756 | I_kwDOHa8MBc5gRAoM | 2,603 | https://github.com/huggingface/diffusers/issues/2603 | https://api.github.com/repos/huggingface/diffusers/issues/2603 | Attend-And-Excitte Pipeline can not generate non squared images. | ### Describe the bug
While using `StableDiffusionAttendAndExcitePipeline` class.
I was unable to generate non-square images, changing to the older weights from `2-1` did not help to resolve the problem.
Or the solution proposed in [pr](https://github.com/huggingface/diffusers/pull/2476#issuecomment-1457979392), I... | closed | completed | false | 3 | [
"bug"
] | [
"yiyixuxu"
] | 2023-03-08T10:55:49Z | 2023-03-14T10:19:26Z | 2023-03-14T10:19:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JanSoltysik | 38,464,373 | MDQ6VXNlcjM4NDY0Mzcz | User | false |
huggingface/diffusers | 1,615,344,035 | I_kwDOHa8MBc5gSDGj | 2,606 | https://github.com/huggingface/diffusers/issues/2606 | https://api.github.com/repos/huggingface/diffusers/issues/2606 | get stuck when save_state using DeepSpeed backend under training train_text_to_image_lora | ### Describe the bug
When using DeepSpeed backend, training is ok but get stuck in `accelerator.save_state(save_path)`. If use MULTI_GPU, the process is OK.
The training script is
```
accelerate launch train_text_to_image_lora.py \
--pretrained_model_name_or_path="pretrain_models/stable-diffusion-v1-4/" ... | closed | completed | false | 19 | [
"bug",
"good first issue"
] | [] | 2023-03-08T14:11:21Z | 2024-11-25T13:04:42Z | 2024-11-25T13:04:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | better629 | 7,869,691 | MDQ6VXNlcjc4Njk2OTE= | User | false |
huggingface/diffusers | 1,615,375,158 | I_kwDOHa8MBc5gSKs2 | 2,607 | https://github.com/huggingface/diffusers/issues/2607 | https://api.github.com/repos/huggingface/diffusers/issues/2607 | MLP in the diffusers library | Hello,
I'm currently dealing with very small size images, (1,8,8) for example, and would like to train a latent denoising model using an MLP or a ResMLP instead of a UNet2DModel, are there any available implementations for such a task? Do you suggest another architecture?
Thanks
| closed | completed | false | 2 | [
"stale"
] | [] | 2023-03-08T14:31:35Z | 2023-04-15T15:03:25Z | 2023-04-15T15:03:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yass43 | 62,470,686 | MDQ6VXNlcjYyNDcwNjg2 | User | false |
huggingface/diffusers | 1,615,658,689 | I_kwDOHa8MBc5gTP7B | 2,609 | https://github.com/huggingface/diffusers/issues/2609 | https://api.github.com/repos/huggingface/diffusers/issues/2609 | Canny ControlNet checkpoint for Stable Diffusion 2.1 | I have converted great checkpoint from @thibaudart in ckpt format to diffusers format and saved only ControlNet part in fp16 so it only takes 700mb of space.
Both 2.1 and 2.1-base work, but 2.1-base seems to work better
In order to convert I manually took part starting from [this](https://github.com/huggingface/dif... | closed | completed | false | 8 | [] | [] | 2023-03-08T17:39:00Z | 2023-03-16T10:51:37Z | 2023-03-16T10:51:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | thepowerfuldeez | 11,796,343 | MDQ6VXNlcjExNzk2MzQz | User | false |
huggingface/diffusers | 1,352,761,078 | I_kwDOHa8MBc5QoX72 | 261 | https://github.com/huggingface/diffusers/issues/261 | https://api.github.com/repos/huggingface/diffusers/issues/261 | Inpainting example is using data from the masked parts with strength 1 | ### Describe the bug
It is my understanding that if we provide `strength=1` to the inpainting pipeline, that the model would try to fill the missing parts without taking into account what was in the image before (where the mask is).
It does not appear to be the case and is really apparent when trying to zoomout of ... | closed | completed | false | 5 | [
"stale"
] | [
"anton-l"
] | 2022-08-26T21:42:27Z | 2022-10-31T15:03:34Z | 2022-10-31T15:03:34Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | leszekhanusz | 4,641,499 | MDQ6VXNlcjQ2NDE0OTk= | User | false |
huggingface/diffusers | 1,615,831,420 | I_kwDOHa8MBc5gT6F8 | 2,610 | https://github.com/huggingface/diffusers/issues/2610 | https://api.github.com/repos/huggingface/diffusers/issues/2610 | clang Error while setting up locally on Mac | ### Describe the bug
Getting this error while setting up diffusers locally on Mac
`clang: error: unsupported option '-fopenmp'`
tried these steps from https://stackoverflow.com/questions/60005176/how-to-deal-with-clang-error-unsupported-option-fopenmp-on-travis
```
brew install llvm libomp
export CPP=/u... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-03-08T19:53:00Z | 2023-03-08T20:40:42Z | 2023-03-08T20:40:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nilooy | 32,486,682 | MDQ6VXNlcjMyNDg2Njgy | User | false |
huggingface/diffusers | 1,615,857,469 | I_kwDOHa8MBc5gUAc9 | 2,611 | https://github.com/huggingface/diffusers/issues/2611 | https://api.github.com/repos/huggingface/diffusers/issues/2611 | stabilityai/stable-diffusion-2-depth is missing diffusion_py torch_model.bin | ```
/home/ubuntu/.local/lib/python3.8/site-packages/diffusers/models/modeling_utils.py:851: FutureWarning: You are loading the variant fp16 from stabilityai/stable-diffusion-2-depth via `revision='fp16'`. This behavio
r is deprecated and will be removed in diffusers v1. One should use `variant='fp16'` instead. Howeve... | closed | completed | false | 3 | [] | [] | 2023-03-08T20:07:52Z | 2023-03-10T11:49:09Z | 2023-03-10T11:49:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jacobzlogar | 10,762,368 | MDQ6VXNlcjEwNzYyMzY4 | User | false |
huggingface/diffusers | 1,615,998,719 | I_kwDOHa8MBc5gUi7_ | 2,613 | https://github.com/huggingface/diffusers/issues/2613 | https://api.github.com/repos/huggingface/diffusers/issues/2613 | Support for adding multiple LoRA layers to Diffusers | Given the popularity of LoRA training for Stable Diffusion, it would be great if the support for adding LoRAs to Diffusers could be expanded to support adding multiple LoRAs. This would be an expansion of this feature - https://github.com/huggingface/diffusers/pull/2129#issuecomment-1460742806. | closed | completed | false | 31 | [
"wip"
] | [
"patrickvonplaten"
] | 2023-03-08T21:52:34Z | 2024-07-25T03:32:34Z | 2023-10-09T06:53:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kirit93 | 8,190,572 | MDQ6VXNlcjgxOTA1NzI= | User | false |
huggingface/diffusers | 1,616,202,850 | I_kwDOHa8MBc5gVUxi | 2,614 | https://github.com/huggingface/diffusers/issues/2614 | https://api.github.com/repos/huggingface/diffusers/issues/2614 | Cannot change pre-trained-model download directory | ### Describe the bug
I tried the very first example **Text-to-Image generation with Stable Diffusion**.
I have set every single variable I can find on Internet to the current path, but it keeps downloading model to `~/.cache`. I have very limited home space, so it keeps telling me no space on the disk.
### Reprod... | closed | completed | false | 12 | [
"bug"
] | [] | 2023-03-09T00:42:36Z | 2023-12-21T16:53:24Z | 2023-03-13T23:08:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xingzhehe | 28,001,407 | MDQ6VXNlcjI4MDAxNDA3 | User | false |
huggingface/diffusers | 1,616,404,824 | I_kwDOHa8MBc5gWGFY | 2,615 | https://github.com/huggingface/diffusers/issues/2615 | https://api.github.com/repos/huggingface/diffusers/issues/2615 | WOw, soo mcuh repetition of code | A lot of pipelines seem to be a copy of each other with changes?
This repo has 11k stars but follows very bad software eng. practices.
Can you write common functionaly of pipelines in some other class and then inherit from those??? | closed | completed | false | 1 | [] | [] | 2023-03-09T04:44:11Z | 2023-03-09T08:58:00Z | 2023-03-09T08:58:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghost | 10,137 | MDQ6VXNlcjEwMTM3 | User | false |
huggingface/diffusers | 1,616,490,723 | I_kwDOHa8MBc5gWbDj | 2,616 | https://github.com/huggingface/diffusers/issues/2616 | https://api.github.com/repos/huggingface/diffusers/issues/2616 | NoneType object has no attribute 'replace' when text_to_image/train_text_to_image_lora save save_attn_procs | ### Describe the bug
inside `examples/text_to_image/train_text_to_image_lora.py`
```
unet = unet.to(torch.float32)
unet.save_attn_procs(args.output_dir)
```
use `save_attn_procs` but the `weights_name` is None which cause the problem in
https://github.com/huggingface/diffusers/blob/main/src/... | closed | completed | false | 9 | [
"bug",
"stale"
] | [
"sayakpaul"
] | 2023-03-09T06:11:59Z | 2023-04-16T15:03:16Z | 2023-04-16T15:03:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | better629 | 7,869,691 | MDQ6VXNlcjc4Njk2OTE= | User | false |
huggingface/diffusers | 1,616,737,781 | I_kwDOHa8MBc5gXXX1 | 2,618 | https://github.com/huggingface/diffusers/issues/2618 | https://api.github.com/repos/huggingface/diffusers/issues/2618 | torch.FloatTensor as image input for StableDiffusionControlNetPipeline | ### Describe the bug
I get "tuple index out of range" when using torch.FloatTensor as input
### Reproduction
```
output = pipe(
[p + ', best quality, extremely detailed'],
image=torch.FloatTensor(depth_cut),
negative_prompt=['longbody, lowres, bad anatomy, ba... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-03-09T09:00:05Z | 2023-04-16T15:03:15Z | 2023-04-16T15:03:15Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jingyangcarl | 24,415,835 | MDQ6VXNlcjI0NDE1ODM1 | User | false |
huggingface/diffusers | 1,353,133,472 | I_kwDOHa8MBc5Qpy2g | 262 | https://github.com/huggingface/diffusers/issues/262 | https://api.github.com/repos/huggingface/diffusers/issues/262 | Add support for external latent specification | https://colab.research.google.com/github/pcuenca/diffusers-examples/blob/main/notebooks/stable-diffusion-seeds.ipynb <- this notebook (linked in the README) makes it seem as if we can specify the latent externally, but looking at the code that doesn't seem to work.
Here is a [proposed tweak](https://gist.github.com/... | closed | completed | false | 2 | [] | [] | 2022-08-27T20:07:28Z | 2022-08-28T08:20:06Z | 2022-08-28T08:20:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | gordicaleksa | 29,271,842 | MDQ6VXNlcjI5MjcxODQy | User | false |
huggingface/diffusers | 1,617,118,382 | I_kwDOHa8MBc5gY0Su | 2,622 | https://github.com/huggingface/diffusers/issues/2622 | https://api.github.com/repos/huggingface/diffusers/issues/2622 | is it necessary to add a description of each picture in the training set？ | Hey, thanks for this great project, I'm a bit confused.
For painting style learning, is it necessary to add a description of each picture in the training set? For example, 1) use a unique word to represent the style of painting, such as <sks style> used in dreambooth lora, and there is no image description file; 2) us... | closed | completed | false | 8 | [] | [
"sayakpaul"
] | 2023-03-09T12:18:49Z | 2024-08-18T14:36:35Z | 2023-03-14T09:38:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jason-kid | 31,425,629 | MDQ6VXNlcjMxNDI1NjI5 | User | false |
huggingface/diffusers | 1,617,461,629 | I_kwDOHa8MBc5gaIF9 | 2,626 | https://github.com/huggingface/diffusers/issues/2626 | https://api.github.com/repos/huggingface/diffusers/issues/2626 | Provide example for training ControlNet | null | closed | not_planned | false | 1 | [] | [] | 2023-03-09T15:22:35Z | 2023-03-10T13:10:33Z | 2023-03-09T15:30:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ghost | 10,137 | MDQ6VXNlcjEwMTM3 | User | false |
huggingface/diffusers | 1,617,580,252 | I_kwDOHa8MBc5galDc | 2,628 | https://github.com/huggingface/diffusers/issues/2628 | https://api.github.com/repos/huggingface/diffusers/issues/2628 | Adding cosine with decay | **Is your feature request related to a problem? Please describe.**
Hi! We wanted to add cosine with decay optimizer to diffusers since in training muse and other experiments, it's very valuable.
**Describe the solution you'd like**
I'll make a pr.
| closed | completed | false | 5 | [
"stale"
] | [] | 2023-03-09T16:30:14Z | 2023-04-10T00:01:03Z | 2023-04-10T00:01:03Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | isamu-isozaki | 23,430,101 | MDQ6VXNlcjIzNDMwMTAx | User | false |
huggingface/diffusers | 1,353,296,551 | I_kwDOHa8MBc5Qqaqn | 263 | https://github.com/huggingface/diffusers/issues/263 | https://api.github.com/repos/huggingface/diffusers/issues/263 | `DDIMScheduler` does not work unless `set_timesteps` is used | ### Describe the bug
Need to set number of inference steps since "None" is default. Expected to have a default value.
Adding a very simple PR.
### Reproduction
```
scheduler = DDIMScheduler(
beta_start=0.00085, beta_end=0.012, beta_schedule="scaled_linear"
)
# skipping this step will... | closed | completed | false | 6 | [
"bug"
] | [] | 2022-08-28T09:23:09Z | 2022-09-13T16:04:31Z | 2022-09-13T16:04:31Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | samedii | 1,872,050 | MDQ6VXNlcjE4NzIwNTA= | User | false |
huggingface/diffusers | 1,617,789,537 | I_kwDOHa8MBc5gbYJh | 2,632 | https://github.com/huggingface/diffusers/issues/2632 | https://api.github.com/repos/huggingface/diffusers/issues/2632 | Training script lr schedulers configuration not taking into account gradient accumulation correctly | ### Describe the bug
Re internal discussion: it sounds like accelerate handles prepared lr schedulers step function only being called when gradient updates actually occur. This means that we shouldn't have to change our lr scheduler configuration to artificially stretch out the number of steps based off of the number ... | closed | completed | false | 1 | [
"bug"
] | [
"williamberman"
] | 2023-03-09T18:32:49Z | 2023-04-19T15:04:17Z | 2023-04-19T15:04:17Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,617,849,488 | I_kwDOHa8MBc5gbmyQ | 2,633 | https://github.com/huggingface/diffusers/issues/2633 | https://api.github.com/repos/huggingface/diffusers/issues/2633 | Asymmetric tiling | Hello. I'm trying to achieve tiling asymmetrically using Diffusers, in a similar fashion to the asymmetric tiling in Automatic1111's extension https://github.com/tjm35/asymmetric-tiling-sd-webui.
My understanding is that I must traverse all layers to alter the padding, in my case circular in X and constant in Y, bu... | closed | completed | false | 12 | [
"good first issue",
"question"
] | [] | 2023-03-09T19:09:34Z | 2025-07-29T08:48:27Z | 2023-03-11T03:27:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alejobrainz | 6,945,663 | MDQ6VXNlcjY5NDU2NjM= | User | false |
huggingface/diffusers | 1,618,265,335 | I_kwDOHa8MBc5gdMT3 | 2,635 | https://github.com/huggingface/diffusers/issues/2635 | https://api.github.com/repos/huggingface/diffusers/issues/2635 | which scheduler is "DPM++ 2s a Karras" and "DPM++ 2M Karras" ? | ### Model/Pipeline/Scheduler description
I used scheduler 【DPM ++ 2s a Karras】 and 【DPM++ 2M Karras】before in sd-webui, and I know these two scheduler has better performance.
And I found that diffusers has single/multi-DPM solver , but I don't know which is DPM ++ 2s a Karras or DPM++ 2M Karras.
Could you please t... | closed | completed | false | 8 | [] | [] | 2023-03-10T02:34:48Z | 2023-12-19T11:03:31Z | 2023-03-10T03:25:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ym-ctw | 124,122,174 | U_kgDOB2X0Pg | User | false |
huggingface/diffusers | 1,618,450,981 | I_kwDOHa8MBc5gd5ol | 2,636 | https://github.com/huggingface/diffusers/issues/2636 | https://api.github.com/repos/huggingface/diffusers/issues/2636 | Resume from checkpoint loops through skipped steps and sets progress bar too short | ### Describe the bug
We don't have to loop through each previous step to get to the resume point. This is on the order of minutes for moderate training runs.
the length of the tqdm progress bar is the length of the entire iterable, not the start to the end point. There's a separate argument that determines the end ... | closed | completed | false | 1 | [
"bug"
] | [
"williamberman"
] | 2023-03-10T06:47:28Z | 2023-04-19T15:04:14Z | 2023-04-19T15:04:14Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,618,719,122 | I_kwDOHa8MBc5ge7GS | 2,638 | https://github.com/huggingface/diffusers/issues/2638 | https://api.github.com/repos/huggingface/diffusers/issues/2638 | runwayml/stable-diffusion-inpainting not working as expected | runwayml/stable-diffusion-inpainting not working as expected. It produced good results untill yesterday. what can be the issue for it to not work same as yesterday | closed | completed | false | 2 | [
"stale"
] | [] | 2023-03-10T10:20:50Z | 2023-04-17T15:03:34Z | 2023-04-17T15:03:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Pragya251990 | 127,494,813 | U_kgDOB5lqnQ | User | false |
huggingface/diffusers | 1,620,116,802 | I_kwDOHa8MBc5gkQVC | 2,644 | https://github.com/huggingface/diffusers/issues/2644 | https://api.github.com/repos/huggingface/diffusers/issues/2644 | Does dreambooth support be trained on two 12GB GPUs? | Hi, I am relatively new to diffusion. I am training model stable-diffusion-v1-5. I set use 2 GPUs through `accelerate config`.
```
In which compute environment are you running? ([0] This machine, [1] AWS (Amazon SageMaker)): 0
Which type of machine are you using? ([0] No distributed training, [1] multi-CPU, [2] mu... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-03-11T19:19:19Z | 2023-04-19T15:04:12Z | 2023-04-19T15:04:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wangtsiao | 70,847,926 | MDQ6VXNlcjcwODQ3OTI2 | User | false |
huggingface/diffusers | 1,620,123,762 | I_kwDOHa8MBc5gkSBy | 2,645 | https://github.com/huggingface/diffusers/issues/2645 | https://api.github.com/repos/huggingface/diffusers/issues/2645 | ControlNet pipeline failed on mac M1 with "Assertion error: torch not compiled with cuda enabled" | ### Describe the bug
**ControlNet pipeline failed on mac M1 with "Assertion error: torch not compiled with cuda enabled"**
I've managed to follow the M1/M2 instructions to run baseline SD diffusers as described here: https://huggingface.co/docs/diffusers/optimization/mps
However, other pipelines failed with `As... | closed | completed | false | 14 | [
"bug",
"stale"
] | [] | 2023-03-11T19:44:53Z | 2023-07-05T03:59:35Z | 2023-04-19T15:04:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | RELNO | 14,296,774 | MDQ6VXNlcjE0Mjk2Nzc0 | User | false |
huggingface/diffusers | 1,620,215,590 | I_kwDOHa8MBc5gkocm | 2,646 | https://github.com/huggingface/diffusers/issues/2646 | https://api.github.com/repos/huggingface/diffusers/issues/2646 | tiled VAE fails with IndexError | ### Describe the bug
decoding with `enable_vae_tiling` fails with the Stable Diffusion 2.1 model on certain sizes, throwing an IndexError.
### Reproduction
```py
import diffusers, torch
pipeline = diffusers.StableDiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-2-1", torch_dtype=torch.float16, revis... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-03-12T03:01:59Z | 2023-04-19T15:04:09Z | 2023-04-19T15:04:09Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | keturn | 83,819 | MDQ6VXNlcjgzODE5 | User | false |
huggingface/diffusers | 1,620,943,913 | I_kwDOHa8MBc5gnaQp | 2,650 | https://github.com/huggingface/diffusers/issues/2650 | https://api.github.com/repos/huggingface/diffusers/issues/2650 | [Docs] Put together a doc for "diffusion models in prod" | Internal discussion thread: https://huggingface.slack.com/archives/C03HBN1C8CW/p1678308485758429 | closed | completed | false | 3 | [
"wip"
] | [
"pcuenca",
"sayakpaul"
] | 2023-03-13T08:02:06Z | 2024-11-21T15:39:30Z | 2024-11-21T15:39:30Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,621,047,389 | I_kwDOHa8MBc5gnzhd | 2,651 | https://github.com/huggingface/diffusers/issues/2651 | https://api.github.com/repos/huggingface/diffusers/issues/2651 | Provide a way to find out which model branches map to which pipelines | Right now a model repo may have multiple branches (e.g., main, fp16, bf16, etc).
There seems to be a de facto convention that a given branch is designed for a specific purpose. E.g., The "main" branch contains the model weights with fp32 precision for a pytorch model, whereas the fp16 contains the same weights in fp... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-03-13T09:11:33Z | 2023-04-20T15:03:28Z | 2023-04-20T15:03:28Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | antoche | 1,627,384 | MDQ6VXNlcjE2MjczODQ= | User | false |
huggingface/diffusers | 1,622,174,189 | I_kwDOHa8MBc5gsGnt | 2,659 | https://github.com/huggingface/diffusers/issues/2659 | https://api.github.com/repos/huggingface/diffusers/issues/2659 | [SD] generated images are slightly varying with batch size | ### Describe the bug
I'm not sure whether this is an expected behavior, but there are slight (but noticeable!) differences in generated images depending on the batch size. I've tried different schedulers, models, and even pipelines (txt2img and img2img). I haven't used xformers but I used fp16. Are such differences ex... | closed | not_planned | false | 14 | [
"bug",
"stale"
] | [] | 2023-03-13T20:01:47Z | 2023-05-22T15:33:27Z | 2023-05-22T15:33:26Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vvsotnikov | 21,967,926 | MDQ6VXNlcjIxOTY3OTI2 | User | false |
huggingface/diffusers | 1,622,555,569 | I_kwDOHa8MBc5gtjux | 2,661 | https://github.com/huggingface/diffusers/issues/2661 | https://api.github.com/repos/huggingface/diffusers/issues/2661 | Checkpoint saving at 3000/15000 yields to write error. | ### System info:
- `diffusers` version: 0.15.0.dev0 - Platform: Linux-4.18.0-372.19.1.el8_6.x86_64-x86_64-with-glibc2... | closed | completed | false | 0 | [] | [] | 2023-03-14T01:24:28Z | 2023-03-14T01:25:15Z | 2023-03-14T01:25:15Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Chasexj | 47,127,634 | MDQ6VXNlcjQ3MTI3NjM0 | User | false |
huggingface/diffusers | 1,622,557,096 | I_kwDOHa8MBc5gtkGo | 2,662 | https://github.com/huggingface/diffusers/issues/2662 | https://api.github.com/repos/huggingface/diffusers/issues/2662 | Checkpoint saving leads to write error. | ### System info:
- `diffusers` version: 0.15.0.dev0 - Platform: Linux-4.18.0-372.19.1.el8_6.x86_64-x86_64-with-glibc2... | closed | completed | false | 6 | [] | [] | 2023-03-14T01:25:55Z | 2023-04-19T19:36:24Z | 2023-04-19T19:36:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Chasexj | 47,127,634 | MDQ6VXNlcjQ3MTI3NjM0 | User | false |
huggingface/diffusers | 1,622,585,456 | I_kwDOHa8MBc5gtrBw | 2,663 | https://github.com/huggingface/diffusers/issues/2663 | https://api.github.com/repos/huggingface/diffusers/issues/2663 | Model fails to load if StableDiffusionSafetyChecker safetensors are present | ### Describe the bug
This used to work fine a day ago:
```python
from diffusers import StableDiffusionPipeline
p = StableDiffusionPipeline.from_pretrained("prompthero/openjourney")
```
Now it fails with the following exception:
```python
NotImplementedError: Make sure `_init_weights` is implemented for <c... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-03-14T01:56:50Z | 2023-10-30T16:06:40Z | 2023-05-14T15:03:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | antoche | 1,627,384 | MDQ6VXNlcjE2MjczODQ= | User | false |
huggingface/diffusers | 1,622,667,963 | I_kwDOHa8MBc5gt_K7 | 2,664 | https://github.com/huggingface/diffusers/issues/2664 | https://api.github.com/repos/huggingface/diffusers/issues/2664 | Timestep embedding in UNet1DModel and UNet1DConditionModel | **Is your feature request related to a problem? Please describe.**
It seems the UNet1DModel does not support the timestep embedding as UNet2DModel. For example, [this function](https://github.com/huggingface/diffusers/blob/f20c8f5a1aba27f5972cad50516f18ba516e4d9e/src/diffusers/models/unet_1d_blocks.py#L499) does not c... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-03-14T03:41:15Z | 2023-05-08T15:03:36Z | 2023-05-08T15:03:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yujianll | 46,540,151 | MDQ6VXNlcjQ2NTQwMTUx | User | false |
huggingface/diffusers | 1,623,592,008 | I_kwDOHa8MBc5gxgxI | 2,668 | https://github.com/huggingface/diffusers/issues/2668 | https://api.github.com/repos/huggingface/diffusers/issues/2668 | StableDiffusionControlNetPipeline with lpw_stable_diffusion treats ControlNet image as Img2Img image | ### Describe the bug
If I use StableDiffusionControlNetPipeline with lpw_stable_diffusion and input a ControlNet image, the output image does not follow the ControlNet image. If I input a normal image, the output image will be similar to the input image. It looks that StableDiffusionControlNetPipeline with lpw_stabl... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-03-14T14:23:38Z | 2023-03-16T14:51:39Z | 2023-03-16T11:43:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nodchip | 685,625 | MDQ6VXNlcjY4NTYyNQ== | User | false |
huggingface/diffusers | 1,623,681,240 | I_kwDOHa8MBc5gx2jY | 2,669 | https://github.com/huggingface/diffusers/issues/2669 | https://api.github.com/repos/huggingface/diffusers/issues/2669 | AttributeError: 'NoneType' object has no attribute 'replace' | ### Describe the bug
I follow train dreambooth lora model guid in https://github.com/huggingface/diffusers/tree/main/examples/dreambooth#training-with-low-rank-adaptation-of-large-language-models-lora
when I run accelerate launch train_dreambooth_lora.py, I got error like below
Steps: 100%|██████████████████████... | closed | completed | false | 8 | [
"bug"
] | [] | 2023-03-14T15:07:44Z | 2023-07-25T01:04:14Z | 2023-03-21T03:00:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | killermyth | 1,821,315 | MDQ6VXNlcjE4MjEzMTU= | User | false |
huggingface/diffusers | 1,353,555,144 | I_kwDOHa8MBc5QrZzI | 267 | https://github.com/huggingface/diffusers/issues/267 | https://api.github.com/repos/huggingface/diffusers/issues/267 | Non-squared Image shape | Is it possible to use diffusers on non-squared images?
That would be a very interesting feature! | closed | completed | false | 4 | [
"question"
] | [] | 2022-08-29T01:29:33Z | 2022-09-13T15:57:36Z | 2022-09-13T15:57:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LucasSilvaFerreira | 5,742,873 | MDQ6VXNlcjU3NDI4NzM= | User | false |
huggingface/diffusers | 1,624,471,831 | I_kwDOHa8MBc5g03kX | 2,673 | https://github.com/huggingface/diffusers/issues/2673 | https://api.github.com/repos/huggingface/diffusers/issues/2673 | Can't make StableDiffusionUpscalePipeline work, Please verify the config of `pipeline.unet` or your `image` input. | ### Describe the bug
> expects 8 but
received `num_channels_latents`: 4 + `num_channels_image`: 3 = 7. Please
verify the config of `pipeline.unet` or your `image` input.
So, basically I'm trying to make upscaler pipilne work, but got weird issues, potentially doing something wrong initializing it.
### Rep... | closed | completed | false | 10 | [
"bug",
"stale"
] | [] | 2023-03-14T23:50:18Z | 2024-11-19T14:28:49Z | 2023-04-14T15:05:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Pirog17000 | 3,022,852 | MDQ6VXNlcjMwMjI4NTI= | User | false |
huggingface/diffusers | 1,624,557,321 | I_kwDOHa8MBc5g1McJ | 2,674 | https://github.com/huggingface/diffusers/issues/2674 | https://api.github.com/repos/huggingface/diffusers/issues/2674 | Lora don't work without safetensors but not in requirements | ### Describe the bug
Try last version from github, but it is crashed when I try use lora without installed safetensors, so propose to put it in setup requirements.
### Reproduction
just try to use diffusers pipeline without installed safetensors
### Logs
_No response_
### System Info
2023-03-15 01:20:20.345859: ... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-03-15T01:21:31Z | 2023-04-22T15:03:37Z | 2023-04-22T15:03:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | stilletto | 9,992,342 | MDQ6VXNlcjk5OTIzNDI= | User | false |
huggingface/diffusers | 1,624,643,805 | I_kwDOHa8MBc5g1hjd | 2,675 | https://github.com/huggingface/diffusers/issues/2675 | https://api.github.com/repos/huggingface/diffusers/issues/2675 | Textual Inversion training colab: ValueError: You have to specify the number of GPUs you would like to use, add `num_processes=...` to your call. | │ in <module>:2 │
│ │
│ 1 import accelerate │
│ ❱ 2 accelera... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-03-15T03:02:07Z | 2023-04-22T15:03:35Z | 2023-04-22T15:03:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | KaiyueSun98 | 94,892,075 | U_kgDOBafwKw | User | false |
huggingface/diffusers | 1,624,699,310 | I_kwDOHa8MBc5g1vGu | 2,676 | https://github.com/huggingface/diffusers/issues/2676 | https://api.github.com/repos/huggingface/diffusers/issues/2676 | "What Schedulers are supported on diffusers? I found some in the webUI that I couldn't find on diffusers, such as "DPM++ 2M Karras" and "DPM++ SDE Karras". Does anyone know their names on diffusers?" | ### Model/Pipeline/Scheduler description
"What Schedulers are supported on diffusers? I found some in the webUI that I couldn't find on diffusers, such as "DPM++ 2M Karras" and "DPM++ SDE Karras". Does anyone know their names on diffusers?"
### Open source status
- [X] The model implementation is available
- [X] The... | closed | completed | false | 8 | [] | [] | 2023-03-15T04:03:14Z | 2023-10-10T00:51:40Z | 2023-03-16T01:20:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | XXXKAY | 119,796,959 | U_kgDOByP03w | User | false |
huggingface/diffusers | 1,624,829,138 | I_kwDOHa8MBc5g2OzS | 2,679 | https://github.com/huggingface/diffusers/issues/2679 | https://api.github.com/repos/huggingface/diffusers/issues/2679 | How to use attend-and-excite pipeline with inpainting or depth pipelines? | If I am not mistaken attend-and-excite pipeline is using text2img model. Is it possible to use it with "runwayml/stable-diffusion-inpainting" or "stabilityai/stable-diffusion-2-depth" pipelines? Thanks. | closed | completed | false | 6 | [
"stale"
] | [
"yiyixuxu"
] | 2023-03-15T06:15:42Z | 2023-04-24T15:03:39Z | 2023-04-24T15:03:39Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | himmetozcan | 44,242,024 | MDQ6VXNlcjQ0MjQyMDI0 | User | false |
huggingface/diffusers | 1,624,840,687 | I_kwDOHa8MBc5g2Rnv | 2,680 | https://github.com/huggingface/diffusers/issues/2680 | https://api.github.com/repos/huggingface/diffusers/issues/2680 | Sovled | null | closed | completed | false | 0 | [] | [] | 2023-03-15T06:25:52Z | 2023-03-15T06:41:31Z | 2023-03-15T06:40:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | haidog-yaqub | 74,460,932 | MDQ6VXNlcjc0NDYwOTMy | User | false |
huggingface/diffusers | 1,624,932,868 | I_kwDOHa8MBc5g2oIE | 2,681 | https://github.com/huggingface/diffusers/issues/2681 | https://api.github.com/repos/huggingface/diffusers/issues/2681 | the FID for the Stable Diffusion | In the NVIDIA paper "eDiff-I: Text-to-Image Diffusion Models with an Ensemble of Expert Denoisers", it said the zero-shot FID for the stable diffusion on the COCO2014 validation set can be 8.59:
And I see ... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-03-15T07:42:42Z | 2024-10-22T09:45:43Z | 2023-04-22T15:03:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiayisunx | 50,946,601 | MDQ6VXNlcjUwOTQ2NjAx | User | false |
huggingface/diffusers | 1,625,084,540 | I_kwDOHa8MBc5g3NJ8 | 2,683 | https://github.com/huggingface/diffusers/issues/2683 | https://api.github.com/repos/huggingface/diffusers/issues/2683 | dreambooth finetune | hello,when using dreambooth to train anime characters, I encountered an embarrassing problem.
1) When using dreambooth lora without train_text_encoder, I get pretty bad results;
2) But when I use dreambooth with train_text_encoder, the results are pretty good.
Well, any suggestions? lora+train_text_encoder or anythi... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-03-15T09:16:40Z | 2023-04-22T15:03:32Z | 2023-04-22T15:03:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jason-kid | 31,425,629 | MDQ6VXNlcjMxNDI1NjI5 | User | false |
huggingface/diffusers | 1,625,635,788 | I_kwDOHa8MBc5g5TvM | 2,689 | https://github.com/huggingface/diffusers/issues/2689 | https://api.github.com/repos/huggingface/diffusers/issues/2689 | Image link is broken | ### Describe the bug
On this page, "https://huggingface.co/docs/diffusers/v0.14.0/en/using-diffusers/inpaint", the link to the drawing is broken.
### Reproduction
.
### Logs
_No response_
### System Info
. | closed | completed | false | 1 | [
"bug"
] | [
"yiyixuxu"
] | 2023-03-15T14:26:26Z | 2023-03-17T05:35:29Z | 2023-03-17T05:35:29Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tolgacangoz | 46,008,593 | MDQ6VXNlcjQ2MDA4NTkz | User | false |
huggingface/diffusers | 1,626,016,973 | I_kwDOHa8MBc5g6wzN | 2,692 | https://github.com/huggingface/diffusers/issues/2692 | https://api.github.com/repos/huggingface/diffusers/issues/2692 | MegaEdit | ### Model/Pipeline/Scheduler description
Hi, I wanted to share a project I was working on. MegaEdit. Here's the github link: https://github.com/ethansmith2000/MegaEdit
I'm unsure how this would best be implemented as it is a wrapper for pipelines rather than a pipeline itself, although still very useful, thank you!... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-03-15T18:06:39Z | 2023-04-24T15:03:38Z | 2023-04-24T15:03:38Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ethansmith2000 | 98,723,285 | U_kgDOBeJl1Q | User | false |
huggingface/diffusers | 1,626,150,048 | I_kwDOHa8MBc5g7RSg | 2,695 | https://github.com/huggingface/diffusers/issues/2695 | https://api.github.com/repos/huggingface/diffusers/issues/2695 | ControlNet Training failed on validation using the default Tensorboard report_to option | ### Describe the bug
I tried the training of the ControlNet in the main branch right away. The default option for `--report_to` is set to `tensorboard`, it seems to raise a ValueError and stop the process after generating validation images. As a workaround, using `wandb` did not cause this issue.
### Reproduction... | closed | completed | false | 10 | [
"bug",
"stale"
] | [
"williamberman"
] | 2023-03-15T19:27:49Z | 2024-04-05T19:18:43Z | 2023-06-03T00:16:50Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | takuma104 | 10,776 | MDQ6VXNlcjEwNzc2 | User | false |
huggingface/diffusers | 1,626,221,038 | I_kwDOHa8MBc5g7inu | 2,696 | https://github.com/huggingface/diffusers/issues/2696 | https://api.github.com/repos/huggingface/diffusers/issues/2696 | AssertionError: Torch not compiled with CUDA enabled` after installing the diffuser library | ### Describe the bug
I followed the installation instruction, and install the diffuser in the virtual environment as
`pip install --upgrade diffusers[torch]`
and also installed transformers as `pip install transformers`.
However, when trying the QuickStart, the `pipeline.to("cuda")` gives the following error... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-03-15T20:14:20Z | 2023-03-16T01:11:57Z | 2023-03-16T01:11:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wenouyang | 5,430,158 | MDQ6VXNlcjU0MzAxNTg= | User | false |
huggingface/diffusers | 1,626,555,397 | I_kwDOHa8MBc5g80QF | 2,700 | https://github.com/huggingface/diffusers/issues/2700 | https://api.github.com/repos/huggingface/diffusers/issues/2700 | UniPC solver_p function throws an error on validation steps with train_controlnet.py | I notice when using the UniPC scheduler it works fine for training, but then as soon it hits the validation steps the solver_p is missing arguments which I don't know how to add. This breaks the training, the same when you use it for regular inference. I lost the log before I decided to post this but will try to get ... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-03-16T02:01:45Z | 2023-04-24T15:03:35Z | 2023-04-24T15:03:35Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Leomn1234 | 127,356,389 | U_kgDOB5dN5Q | User | false |
huggingface/diffusers | 1,626,825,253 | I_kwDOHa8MBc5g92Il | 2,703 | https://github.com/huggingface/diffusers/issues/2703 | https://api.github.com/repos/huggingface/diffusers/issues/2703 | Composable Diffusion Pipeline Broken | ### Describe the bug
Composable Diffusion Pipeline appears to be broken when attempting to follow the instructions here:
https://github.com/huggingface/diffusers/blob/268ebcb015d8816257f24e3cd930fb48964314b4/examples/community/README.md#composable-stable-diffusion
### Reproduction
import torch as th
import numpy... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-03-16T06:46:28Z | 2023-04-11T15:50:28Z | 2023-04-11T15:50:28Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nagolinc | 7,775,917 | MDQ6VXNlcjc3NzU5MTc= | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.