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 | 2,048,293,463 | I_kwDOHa8MBc56FnpX | 6,232 | https://github.com/huggingface/diffusers/issues/6232 | https://api.github.com/repos/huggingface/diffusers/issues/6232 | Bug in DPMSolverSinglestepScheduler: scheduler.add_noise does not work for some settings | ### Describe the bug
Hello!
When using DPMSolverSinglestepScheduler and the function ```set_timesteps```, the resulting timesteps might have two times the same value. A bug occurs when we use the function ```add_noise```. An error is thrown if the value of t appears multiple times in self.timesteps
Setting ```sch... | closed | completed | false | 6 | [
"bug",
"stale",
"scheduler"
] | [
"yiyixuxu"
] | 2023-12-19T09:47:24Z | 2024-02-18T09:31:43Z | 2024-02-18T09:31:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AlexandreBinninger | 33,605,890 | MDQ6VXNlcjMzNjA1ODkw | User | false |
huggingface/diffusers | 2,048,595,646 | I_kwDOHa8MBc56Gxa- | 6,233 | https://github.com/huggingface/diffusers/issues/6233 | https://api.github.com/repos/huggingface/diffusers/issues/6233 | generator/seed logic with multiple images | When you set a seed on a single image, you can reproduce easily with the same generator. However, if you produced it from a batch of images, if it isn't the first one, you won't be able to reproduce it without making a batch. What's the seed logic in batches of images? | closed | completed | false | 7 | [] | [] | 2023-12-19T12:49:09Z | 2024-01-02T12:57:09Z | 2024-01-02T12:57:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 2,048,606,213 | I_kwDOHa8MBc56G0AF | 6,234 | https://github.com/huggingface/diffusers/issues/6234 | https://api.github.com/repos/huggingface/diffusers/issues/6234 | issue with examples/text_to_image/train_text_to_image_sdxl.py | ### Describe the bug
At line 854, the calculation of crop_top_left x1 is wrong, when doing flipping, or it can put the flipping operation before resizing.
### Reproduction
the calculation of crop_top_left x1 is wrong, when doing flipping
### Logs
_No response_
### System Info
the calculation of crop_top_left x1 ... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-12-19T12:55:34Z | 2024-01-26T15:05:23Z | 2024-01-26T15:05:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mboboGO | 37,262,007 | MDQ6VXNlcjM3MjYyMDA3 | User | false |
huggingface/diffusers | 2,049,180,441 | I_kwDOHa8MBc56JAMZ | 6,235 | https://github.com/huggingface/diffusers/issues/6235 | https://api.github.com/repos/huggingface/diffusers/issues/6235 | Dreambooth + local background image? | Hi all,
I have trained a Dreambooth model on a specific object (product). Now for inference, I want to use my own background photos (not the ones that are generated by the model). So basically I want to load my local background photo, and ask the trained Dreambooth model to place the object (that it has been trained... | closed | completed | false | 1 | [] | [] | 2023-12-19T17:59:20Z | 2023-12-20T03:52:54Z | 2023-12-20T03:52:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alirezaomneky | 152,901,224 | U_kgDOCR0WaA | User | false |
huggingface/diffusers | 2,049,615,284 | I_kwDOHa8MBc56KqW0 | 6,238 | https://github.com/huggingface/diffusers/issues/6238 | https://api.github.com/repos/huggingface/diffusers/issues/6238 | Issue with SDXL VAE | ### Describe the bug
When I try to read an image and encode it with VAE encoder and decode I get a significantly different image. I followed the encoding from SDXL training script and decoding from SDXL pipeline.
### Reproduction
```
from diffusers import StableDiffusionXLPipeline
from PIL import Image
im... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-12-20T00:13:25Z | 2023-12-20T07:33:25Z | 2023-12-20T05:10:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vidit98 | 22,493,755 | MDQ6VXNlcjIyNDkzNzU1 | User | false |
huggingface/diffusers | 2,049,754,720 | I_kwDOHa8MBc56LMZg | 6,239 | https://github.com/huggingface/diffusers/issues/6239 | https://api.github.com/repos/huggingface/diffusers/issues/6239 | Allow VAE encode(inp).latent_dist.sample() function to be better supported by torch FX | ### Describe the bug
Torch randn, which is used in the `vae.encode(inp).latent_dist.sample()` function, cannot take symbol shapes as shape argument. However, using randn_like works since it can get the symbolic shape of the tensor. If the sample function could use torch.randn_like instead of torch.rand, it would be ab... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2023-12-20T03:08:34Z | 2024-01-28T15:04:33Z | 2024-01-28T15:04:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aviator19941 | 17,955,474 | MDQ6VXNlcjE3OTU1NDc0 | User | false |
huggingface/diffusers | 1,383,861,397 | I_kwDOHa8MBc5SfAyV | 624 | https://github.com/huggingface/diffusers/issues/624 | https://api.github.com/repos/huggingface/diffusers/issues/624 | NameError: name 'ort' is not defined | I just got this error. The very recent update to onnx_utils.py appears to be buggy. Does not function at all for me. | closed | completed | false | 2 | [] | [] | 2022-09-23T14:05:32Z | 2022-09-23T17:40:02Z | 2022-09-23T17:40:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jalexs82 | 69,216,807 | MDQ6VXNlcjY5MjE2ODA3 | User | false |
huggingface/diffusers | 2,049,758,927 | I_kwDOHa8MBc56LNbP | 6,240 | https://github.com/huggingface/diffusers/issues/6240 | https://api.github.com/repos/huggingface/diffusers/issues/6240 | device_map="auto" in AutoPipelineForText2Image raises error | ### Describe the bug
Thanks for ur contributions! I'm using `AutoPipelineForText2Image` and set `device_map="auto"` when initializing, but the program raises `Expected all tensors to be on the same device, but found at least two devices, cuda:1 and cuda:0!` when inferring. Even I manually put the essential components ... | closed | completed | false | 17 | [
"bug",
"stale"
] | [
"sayakpaul"
] | 2023-12-20T03:14:42Z | 2024-11-22T01:35:08Z | 2024-11-22T01:35:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Nidhogg-lyz | 70,124,523 | MDQ6VXNlcjcwMTI0NTIz | User | false |
huggingface/diffusers | 2,049,907,199 | I_kwDOHa8MBc56Lxn_ | 6,242 | https://github.com/huggingface/diffusers/issues/6242 | https://api.github.com/repos/huggingface/diffusers/issues/6242 | Can you provide the log with fill50k dataset? | ### Describe the bug
I did as the readme, but the result with fill50k model is not so good as the result in the git
### Reproduction
I did as the readme, but the result with fill50k model is not so good as the result in the git
### Logs
_No response_
### System Info
------
### Who can help?
@ | closed | completed | false | 1 | [
"bug"
] | [] | 2023-12-20T06:14:03Z | 2023-12-21T04:14:04Z | 2023-12-21T04:14:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dulibubai | 36,948,738 | MDQ6VXNlcjM2OTQ4NzM4 | User | false |
huggingface/diffusers | 2,049,967,479 | I_kwDOHa8MBc56MAV3 | 6,243 | https://github.com/huggingface/diffusers/issues/6243 | https://api.github.com/repos/huggingface/diffusers/issues/6243 | IP-Adapter Face Id | ### Model/Pipeline/Scheduler description
A new IP-Adapter Face model has been released. It uses image embeddings from a face recognition model instead of CLIP image embedding. In addition to this, it uses LoRa to improve ID consistency.
Do you plan to support this model? I am happy to contribute
### Open sourc... | closed | completed | false | 1 | [] | [
"yiyixuxu"
] | 2023-12-20T07:11:03Z | 2024-01-15T14:43:56Z | 2024-01-15T14:43:56Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fabiorigano | 57,982,783 | MDQ6VXNlcjU3OTgyNzgz | User | false |
huggingface/diffusers | 2,050,375,878 | I_kwDOHa8MBc56NkDG | 6,248 | https://github.com/huggingface/diffusers/issues/6248 | https://api.github.com/repos/huggingface/diffusers/issues/6248 | MotionAdapter.from_single_file | ### Model/Pipeline/Scheduler description
`Animatediff` is a great technique with exploding interest and advancements; additional custom models start popping around in `ckpt` format, which cannot be loaded to `diffusers` in the meantime.
is it possible to implement `from_single_file` function for motion adapter, simi... | closed | completed | false | 22 | [
"single_file"
] | [
"DN6",
"yiyixuxu"
] | 2023-12-20T11:34:57Z | 2024-10-15T09:29:00Z | 2024-10-15T09:29:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eps696 | 4,808,408 | MDQ6VXNlcjQ4MDg0MDg= | User | false |
huggingface/diffusers | 2,050,596,756 | I_kwDOHa8MBc56OZ-U | 6,250 | https://github.com/huggingface/diffusers/issues/6250 | https://api.github.com/repos/huggingface/diffusers/issues/6250 | KeyError loading SDXL vae from single file | ### Describe the bug
Following https://huggingface.co/docs/diffusers/v0.24.0/en/api/models/autoencoderkl#loading-from-the-original-format fails for the vae of [stabilityai/stable-diffusion-xl-base-1.0](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/tree/main).
### Reproduction
```
from diff... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-12-20T13:54:04Z | 2024-01-16T14:37:51Z | 2023-12-21T11:54:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | thethomasboyer | 46,932,488 | MDQ6VXNlcjQ2OTMyNDg4 | User | false |
huggingface/diffusers | 2,050,785,383 | I_kwDOHa8MBc56PIBn | 6,252 | https://github.com/huggingface/diffusers/issues/6252 | https://api.github.com/repos/huggingface/diffusers/issues/6252 | Lora Weights not Loaded from Checkpoint | ### Describe the bug
After training with diffusers/examples/text_to_image/train_text_to_image_lora.py script
the lora checkpoint couldnt loaded to pipeline
### Reproduction
```python
!accelerate launch /content/diffusers/examples/text_to_image/train_text_to_image_lora.py \
--pretrained_model_name_or_path="sta... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-12-20T15:26:23Z | 2024-02-05T15:05:25Z | 2024-02-05T15:05:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | maliozer | 1,297,458 | MDQ6VXNlcjEyOTc0NTg= | User | false |
huggingface/diffusers | 2,050,945,213 | I_kwDOHa8MBc56PvC9 | 6,253 | https://github.com/huggingface/diffusers/issues/6253 | https://api.github.com/repos/huggingface/diffusers/issues/6253 | Do from_config and from_pretrained Generate the Same DDIMScheduler? | Hi,
I'm a bit confused about how from_config and from_pretrained work in the Scheduler, especially for the DDIM Scheduler. When I load a scheduler from a pre-trained model like Stable Diffusion, are these two methods basically doing the same thing / generating the same instance? I mean, if the generated DDIM schedul... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-12-20T16:59:28Z | 2024-02-22T02:43:13Z | 2024-02-22T02:43:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jundanl | 13,307,387 | MDQ6VXNlcjEzMzA3Mzg3 | User | false |
huggingface/diffusers | 2,051,618,196 | I_kwDOHa8MBc56STWU | 6,256 | https://github.com/huggingface/diffusers/issues/6256 | https://api.github.com/repos/huggingface/diffusers/issues/6256 | Feature Request: Add `callback_on_step_begin` in pipelines | **Is your feature request related to a problem? Please describe.**
Currently, diffusers support the `callback_on_step_end` functionality, which enables specific actions or functions at the end of each step within pipelines. However, there's a need to introduce `callback_on_step_begin` to align with the end callback, e... | closed | not_planned | false | 14 | [
"stale"
] | [] | 2023-12-21T03:14:12Z | 2025-01-27T02:00:46Z | 2025-01-27T02:00:46Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | liang-hou | 34,029,775 | MDQ6VXNlcjM0MDI5Nzc1 | User | false |
huggingface/diffusers | 2,051,719,308 | I_kwDOHa8MBc56SsCM | 6,257 | https://github.com/huggingface/diffusers/issues/6257 | https://api.github.com/repos/huggingface/diffusers/issues/6257 | generating colorful sample | unconditional image generation model generating colorful sample
,Is there any way to fix it?? | closed | completed | false | 1 | [] | [] | 2023-12-21T05:23:48Z | 2023-12-24T08:52:58Z | 2023-12-24T08:52:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | swayampragnya-malla | 150,222,718 | U_kgDOCPQ3fg | User | false |
huggingface/diffusers | 2,051,728,090 | I_kwDOHa8MBc56SuLa | 6,258 | https://github.com/huggingface/diffusers/issues/6258 | https://api.github.com/repos/huggingface/diffusers/issues/6258 | Stable Video Diffusion Pipeline: forward chunking degrades image quality | ### Describe the bug
When using the Stable Video Diffusion Pipeline with forward chunking enabled the generated video is very noisy. Below is the example code that I'm using (from the official example at https://huggingface.co/docs/diffusers/main/en/using-diffusers/svd). When the code to enable forward chunking is com... | closed | completed | false | 8 | [
"bug"
] | [
"DN6"
] | 2023-12-21T05:34:47Z | 2024-01-02T23:45:09Z | 2024-01-02T23:45:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | BlakeOne | 30,273,164 | MDQ6VXNlcjMwMjczMTY0 | User | false |
huggingface/diffusers | 2,052,492,837 | I_kwDOHa8MBc56Vo4l | 6,274 | https://github.com/huggingface/diffusers/issues/6274 | https://api.github.com/repos/huggingface/diffusers/issues/6274 | T2I-Adapter with multiple adapters is broken | ### Describe the bug
pipelines `StableDiffusionAdapterPipeline` and `StableDiffusionXLAdapterPipeline` produce error when running with more than one `T2IAdapter`.
this can be a list of adapters or via using `MultiAdapter`
this impacts for SD15 and SDXL variants and its not specific to exact adapter model.
###... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-12-21T14:08:20Z | 2023-12-25T14:16:45Z | 2023-12-25T14:16:44Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 2,052,762,040 | I_kwDOHa8MBc56Wqm4 | 6,275 | https://github.com/huggingface/diffusers/issues/6275 | https://api.github.com/repos/huggingface/diffusers/issues/6275 | T2I-Adapters implementation does not support all official adapter models | ### Describe the bug
While testing adapters for SD15, `canny` and `sketch` models are found to be incompatible with current `T2IAdapter` implementation in diffusers
looking at models itself, seems that those models actually have different configuration:
all working models have same shape:
```json
T2IAdapter(
... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-12-21T16:49:52Z | 2023-12-25T16:08:37Z | 2023-12-25T16:08:37Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 2,052,886,128 | I_kwDOHa8MBc56XI5w | 6,277 | https://github.com/huggingface/diffusers/issues/6277 | https://api.github.com/repos/huggingface/diffusers/issues/6277 | Diffusers example train_text_to_image_lora.py broken gradients? | ### Describe the bug
Gradient won't backprop when running the example lora training.
### Reproduction
running the command as specified:
export MODEL_NAME="CompVis/stable-diffusion-v1-4"
export DATASET_NAME="lambdalabs/pokemon-blip-captions"
accelerate launch --mixed_precision="fp16" train_text_to_image_lora... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-12-21T18:25:32Z | 2023-12-22T05:20:34Z | 2023-12-22T05:20:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jaymefosa | 92,468 | MDQ6VXNlcjkyNDY4 | User | false |
huggingface/diffusers | 2,053,033,898 | I_kwDOHa8MBc56Xs-q | 6,278 | https://github.com/huggingface/diffusers/issues/6278 | https://api.github.com/repos/huggingface/diffusers/issues/6278 | Could not run train_lcm_distill_sd_wds with multi-GPU setting | ### Describe the bug
Got the following error while running the train_lcm_distill_sd_wds.py script:
`AttributeError: 'DistributedDataParallel' object has no attribute 'config'`
Occurs only when multi-gpu setting is used.
Changing the line at 1178 from:
`w_embedding = guidance_scale_embedding(w, embedding_dim=unet... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-12-21T20:40:05Z | 2023-12-27T08:55:23Z | 2023-12-27T08:55:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | akameswa | 124,394,284 | U_kgDOB2obLA | User | false |
huggingface/diffusers | 2,053,200,434 | I_kwDOHa8MBc56YVoy | 6,280 | https://github.com/huggingface/diffusers/issues/6280 | https://api.github.com/repos/huggingface/diffusers/issues/6280 | StableDiffusionXLControlNetImg2ImgPipeline has no "from_single_file" | ### Describe the bug
As stated in the title, `StableDiffusionXLControlNetImg2ImgPipeline` currently does not support `from_single_file`. It will fail with:
```
AttributeError: type object 'StableDiffusionXLControlNetImg2ImgPipeline' has no attribute 'from_single_file'
```
### Reproduction
Try creating a StableDif... | closed | completed | false | 10 | [
"bug",
"single_file"
] | [
"DN6"
] | 2023-12-22T00:16:40Z | 2024-03-19T19:30:00Z | 2024-03-19T19:30:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andypotato | 8,491,849 | MDQ6VXNlcjg0OTE4NDk= | User | false |
huggingface/diffusers | 2,053,204,498 | I_kwDOHa8MBc56YWoS | 6,281 | https://github.com/huggingface/diffusers/issues/6281 | https://api.github.com/repos/huggingface/diffusers/issues/6281 | ImportError: cannot import name 'MultiControlNetModel' from 'diffusers.pipelines.controlnet' | ### Describe the bug
When I am trying to run ip_adapter_faceid i am getting error
I am trying to make a gradio app for this repo : https://huggingface.co/h94/IP-Adapter-FaceID
The error is below
```
Traceback (most recent call last):
File "G:\IP-Adapter-FaceID\web-ui.py", line 8, in <module>
from ip_... | closed | completed | false | 8 | [
"bug"
] | [] | 2023-12-22T00:22:24Z | 2025-03-17T08:59:30Z | 2023-12-22T13:24:28Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FurkanGozukara | 19,240,467 | MDQ6VXNlcjE5MjQwNDY3 | User | false |
huggingface/diffusers | 2,053,229,448 | I_kwDOHa8MBc56YcuI | 6,283 | https://github.com/huggingface/diffusers/issues/6283 | https://api.github.com/repos/huggingface/diffusers/issues/6283 | StableDiffusionXLControlNetImg2ImgPipeline can not use (strength = 0) | ### Describe the bug
When I create a `StableDiffusionXLControlNetImg2ImgPipeline` and set the strength parameter to zero (0.0) I will receive the following error message:
```
cannot reshape tensor of 0 elements into shape [0, -1, 1, 512] because the unspecified dimension size -1 can be any value and is ambiguous
WA... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-12-22T01:05:10Z | 2024-11-27T01:36:19Z | 2024-02-07T15:05:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | andypotato | 8,491,849 | MDQ6VXNlcjg0OTE4NDk= | User | false |
huggingface/diffusers | 2,053,264,498 | I_kwDOHa8MBc56YlRy | 6,284 | https://github.com/huggingface/diffusers/issues/6284 | https://api.github.com/repos/huggingface/diffusers/issues/6284 | SDXL Lora training: testing results are different from validation results | I am trying to train an SDXL Lora to generate ’Kuromi‘ by using the [train_text_to_image_lora_sdxl.py](https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image_lora_sdxl.py). The training is done and the validation results are as I expected.
However, the testing results are weird ... | closed | completed | false | 4 | [] | [] | 2023-12-22T02:02:13Z | 2023-12-25T01:35:19Z | 2023-12-25T01:35:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yanjk3 | 48,338,375 | MDQ6VXNlcjQ4MzM4Mzc1 | User | false |
huggingface/diffusers | 2,053,497,519 | I_kwDOHa8MBc56ZeKv | 6,285 | https://github.com/huggingface/diffusers/issues/6285 | https://api.github.com/repos/huggingface/diffusers/issues/6285 | Uploading Fine-tuned SDXL Model to HuggingFace Repository: Push Rejected Due to Missing or Corrupt Local Objects | Hi,
I am attempting to upload a fine-tuned SDXL model to the HuggingFace Repository but am encountering issues. Specifically, git LFS cannot locate the local object.
### Repository Structure (before upload model, immediately after creation)
```Bash
/REPO
└── .gitattributes
```
#### Contents of .gitattribut... | closed | completed | false | 5 | [
"stale"
] | [] | 2023-12-22T07:20:28Z | 2024-01-27T06:34:42Z | 2024-01-27T06:34:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ofzlo | 90,624,848 | MDQ6VXNlcjkwNjI0ODQ4 | User | false |
huggingface/diffusers | 2,053,575,051 | I_kwDOHa8MBc56ZxGL | 6,287 | https://github.com/huggingface/diffusers/issues/6287 | https://api.github.com/repos/huggingface/diffusers/issues/6287 | setup.py lists many dependencies that aren't actually used | ### Describe the bug
At the least torchvision, tensorflow, tensorboard dependencies aren't used.
### Reproduction
n/a
### Logs
_No response_
### System Info
n/a
### Who can help?
_No response_ | closed | completed | false | 3 | [
"bug"
] | [] | 2023-12-22T08:35:47Z | 2023-12-26T20:46:08Z | 2023-12-26T20:46:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yurivict | 271,906 | MDQ6VXNlcjI3MTkwNg== | User | false |
huggingface/diffusers | 2,053,625,439 | I_kwDOHa8MBc56Z9Zf | 6,288 | https://github.com/huggingface/diffusers/issues/6288 | https://api.github.com/repos/huggingface/diffusers/issues/6288 | The default learning rate of LCM LoRA training is too small | ### Describe the bug
I am not sure if this script been tested, but it seems that the default lr (1e-6) is too small in my test. The model won't learn anything util I have increase it to 1e-4.
```
export MODEL_NAME="runwayml/stable-diffusion-v1-5"
export OUTPUT_DIR="path/to/saved/model"
accelerate launch train_... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-12-22T09:18:05Z | 2024-02-10T04:57:55Z | 2023-12-25T15:59:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Zeqiang-Lai | 26,198,430 | MDQ6VXNlcjI2MTk4NDMw | User | false |
huggingface/diffusers | 2,053,839,777 | I_kwDOHa8MBc56axuh | 6,292 | https://github.com/huggingface/diffusers/issues/6292 | https://api.github.com/repos/huggingface/diffusers/issues/6292 | Color channel order for watermark embedding | ### Describe the bug
The encoder from the invisible watermark library expects input images with the channel order BGR, which is the default in OpenCV. This can be seen [here](https://github.com/ShieldMnt/invisible-watermark/blob/68d0376d94a4701ed240af0841ec12e00676e325/imwatermark/maxDct.py#L21).
As far as I can se... | closed | completed | false | 10 | [
"bug",
"help wanted",
"contributions-welcome"
] | [] | 2023-12-22T12:14:14Z | 2024-04-26T00:44:54Z | 2024-04-26T00:44:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | btlorch | 11,959,715 | MDQ6VXNlcjExOTU5NzE1 | User | false |
huggingface/diffusers | 2,054,011,781 | I_kwDOHa8MBc56bbuF | 6,295 | https://github.com/huggingface/diffusers/issues/6295 | https://api.github.com/repos/huggingface/diffusers/issues/6295 | Artifacts with DPM++ 2M SDE Karras, even when using `use_lu_lambdas` | ### Describe the bug
The results obtained using DPM++ 2M SDE Karras contain artifacts that suggest there is some bug. This happens across different XL models, although is not so visible in the base SDXL. Using Automatic1111 with the same models and the same kind of scheduler results in high-quality results, and it see... | closed | completed | false | 15 | [
"bug"
] | [
"yiyixuxu"
] | 2023-12-22T14:36:05Z | 2024-01-19T17:08:24Z | 2024-01-19T17:08:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ivanprado | 895,720 | MDQ6VXNlcjg5NTcyMA== | User | false |
huggingface/diffusers | 2,054,724,574 | I_kwDOHa8MBc56eJve | 6,297 | https://github.com/huggingface/diffusers/issues/6297 | https://api.github.com/repos/huggingface/diffusers/issues/6297 | Actual version of train_dreambooth_lora_sdxl can't train on my photos. | ### Describe the bug
I have attempted to train models using several recent versions of the `train_dreambooth_lora_sdxl.py` script found at https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth_lora_sdxl.py, using my own photos as well as animal photos. Unfortunately, none of the LoRa ... | closed | completed | false | 11 | [
"bug",
"stale"
] | [] | 2023-12-23T08:57:48Z | 2024-01-29T15:11:29Z | 2024-01-29T15:11:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | stilletto | 9,992,342 | MDQ6VXNlcjk5OTIzNDI= | User | false |
huggingface/diffusers | 2,054,832,927 | I_kwDOHa8MBc56ekMf | 6,298 | https://github.com/huggingface/diffusers/issues/6298 | https://api.github.com/repos/huggingface/diffusers/issues/6298 | No attribute error of CLIPImageProjection when using GLIGEN pipeline | ### Describe the bug
I'm facing an exactly same issue as #4993, where there is an `AttributeError: module diffusers has no attribute CLIPImageProjection` when using `StableDiffusionGLIGENTextImagePipeline`. It was fixed back then according to #4993, but the error popped out again now (which was also observed by anothe... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-12-23T16:17:52Z | 2023-12-24T15:27:23Z | 2023-12-24T15:27:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zjysteven | 19,763,611 | MDQ6VXNlcjE5NzYzNjEx | User | false |
huggingface/diffusers | 2,055,125,197 | I_kwDOHa8MBc56frjN | 6,309 | https://github.com/huggingface/diffusers/issues/6309 | https://api.github.com/repos/huggingface/diffusers/issues/6309 | [i8n-ja] Translating docs to Japanese | <!--
Note: Please search to see if an issue already exists for the language you are trying to translate.
-->
Hi! I'm Yasuna🤗
Let's bring the documentation to all the Japanese-speaking community 🌐.
Who would want to translate? Please follow the 🤗 [TRANSLATING guide](https://github.com/huggingface/diffusers... | open | reopened | false | 8 | [
"documentation",
"stale",
"contributions-welcome"
] | [] | 2023-12-24T14:52:25Z | 2024-04-12T15:05:04Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | YasunaCoffee | 74,343,879 | MDQ6VXNlcjc0MzQzODc5 | User | false |
huggingface/diffusers | 2,055,352,157 | I_kwDOHa8MBc56gi9d | 6,312 | https://github.com/huggingface/diffusers/issues/6312 | https://api.github.com/repos/huggingface/diffusers/issues/6312 | Prompt not equal, tends to cut in half, i have no negative prompts though. | ### Describe the bug
Basically did a prompt, with 77 words. Uh it doesn't seem to work, since the torch size is cut in half and not equal to positive or negative prompt. Not sure how to fix it, I've read people on github with similar issues but do not know how they solved them or if the compel solution works. I do no... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-12-25T04:05:55Z | 2024-01-05T05:38:52Z | 2023-12-26T20:46:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Truebob | 154,024,136 | U_kgDOCS44yA | User | false |
huggingface/diffusers | 2,055,355,013 | I_kwDOHa8MBc56gjqF | 6,313 | https://github.com/huggingface/diffusers/issues/6313 | https://api.github.com/repos/huggingface/diffusers/issues/6313 | The implementation of a Null-Text Inversion Pipeline | ### Model/Pipeline/Scheduler description
**NullTextInversionPipeline**
A pipeline implementation of _Null-text Inversion for Editing Real Images using Guided Diffusion Models_.
As I did not find it in Diffusers, I'd like to add this feature to Diffusers pipeline.
The NullTextInversionPipeline allows users to c... | closed | completed | false | 6 | [] | [] | 2023-12-25T04:11:53Z | 2024-01-27T03:59:40Z | 2024-01-27T03:59:39Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Junsheng121 | 103,119,560 | U_kgDOBiV6yA | User | false |
huggingface/diffusers | 2,055,454,831 | I_kwDOHa8MBc56g8Bv | 6,318 | https://github.com/huggingface/diffusers/issues/6318 | https://api.github.com/repos/huggingface/diffusers/issues/6318 | Does the IP Adapter support mounting multiple IP Adapter models simultaneously and using multiple reference images at the same time? | null | closed | completed | false | 38 | [
"stale",
"IPAdapter"
] | [
"yiyixuxu"
] | 2023-12-25T07:10:05Z | 2024-03-14T15:19:29Z | 2024-03-14T15:19:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | cjt222 | 17,508,662 | MDQ6VXNlcjE3NTA4NjYy | User | false |
huggingface/diffusers | 2,055,543,025 | I_kwDOHa8MBc56hRjx | 6,320 | https://github.com/huggingface/diffusers/issues/6320 | https://api.github.com/repos/huggingface/diffusers/issues/6320 | Emu2 seems to fail to find remote model | ### System Info
- `transformers` version: 4.37.0.dev0
- Platform: Linux-6.6.8-arch1-1-x86_64-with-glibc2.38
- Python version: 3.10.12
- Huggingface_hub version: 0.19.4
- Safetensors version: 0.4.1
- Accelerate version: 0.25.0
- Accelerate config: not found
- PyTorch version (GPU?): 2.1.2 (True)
- Tensorflow... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-12-24T18:26:45Z | 2024-11-02T17:28:32Z | 2024-02-04T15:05:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CoffeeVampir3 | 48,565,901 | MDQ6VXNlcjQ4NTY1OTAx | User | false |
huggingface/diffusers | 2,055,596,242 | I_kwDOHa8MBc56hejS | 6,321 | https://github.com/huggingface/diffusers/issues/6321 | https://api.github.com/repos/huggingface/diffusers/issues/6321 | openai/clip-vit-large-patch14 not detected | ### Describe the bug
In my cog environment, for some reason, it refuses to detect the already downloaded clip.
### Reproduction
```
clip_model=CLIPModel.from_pretrained("openai/clip-vit-large-patch14",local_files_only=True,cache_dir="model_cache")
tokenizer = CLIPTokenizer.from_pretrained("openai/clip-vit... | closed | completed | false | 13 | [
"bug",
"stale"
] | [] | 2023-12-25T10:34:55Z | 2024-02-05T09:32:02Z | 2024-01-28T09:44:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 2,055,697,614 | I_kwDOHa8MBc56h3TO | 6,323 | https://github.com/huggingface/diffusers/issues/6323 | https://api.github.com/repos/huggingface/diffusers/issues/6323 | TensorRTStableDiffusionImg2ImgPipeline: can't set attribute 'guidance_scale' | ### Describe the bug
`guidance_scale` became a not setable prop. So this example will report an error.
### Reproduction
I'm using the example code:
https://github.com/huggingface/diffusers/blob/main/examples/community/README.md#tensorrt-image2image-stable-diffusion-pipeline
### Logs
```shell
Traceback (most re... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-12-25T13:48:18Z | 2023-12-25T15:46:58Z | 2023-12-25T15:46:58Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jannchie | 29,743,310 | MDQ6VXNlcjI5NzQzMzEw | User | false |
huggingface/diffusers | 2,055,698,978 | I_kwDOHa8MBc56h3oi | 6,324 | https://github.com/huggingface/diffusers/issues/6324 | https://api.github.com/repos/huggingface/diffusers/issues/6324 | AttributeError: 'bool' object has no attribute '__module__'. Did you mean: '__mod__'? | ### Describe the bug
`not_compiled_module` could return `bool`.
This causes this example to report an error:
https://github.com/huggingface/diffusers/blob/main/examples/community/README.md?rgh-link-date=2023-12-25T13%3A48%3A18Z#tensorrt-image2image-stable-diffusion-pipeline
### Reproduction
https://github.com/hu... | closed | completed | false | 15 | [
"bug",
"stale"
] | [] | 2023-12-25T13:50:58Z | 2025-08-05T02:39:23Z | 2024-02-03T15:04:33Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jannchie | 29,743,310 | MDQ6VXNlcjI5NzQzMzEw | User | false |
huggingface/diffusers | 2,055,699,836 | I_kwDOHa8MBc56h318 | 6,325 | https://github.com/huggingface/diffusers/issues/6325 | https://api.github.com/repos/huggingface/diffusers/issues/6325 | RAVE | ### Model/Pipeline/Scheduler description
RAVE (Randomized Noise Shuffling for Fast and Consistent Video Editing with Diffusion Models) is a fast text-guided video editing framework. It supports editing videos of any length while preserving motion and semantic structure. From some very basic testing, I really like th... | closed | completed | false | 9 | [] | [] | 2023-12-25T13:52:45Z | 2024-02-04T15:05:07Z | 2024-02-04T15:05:07Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | a-r-r-o-w | 72,266,394 | MDQ6VXNlcjcyMjY2Mzk0 | User | false |
huggingface/diffusers | 2,056,023,184 | I_kwDOHa8MBc56jGyQ | 6,331 | https://github.com/huggingface/diffusers/issues/6331 | https://api.github.com/repos/huggingface/diffusers/issues/6331 | SAG for inpainting | Hello, SAG is great and produces great results, but there are no pipelines that use it for inpainting. Can it be done?
this is the pipeline btw:
https://github.com/huggingface/diffusers/blob/8e5921cac14c84a670b4e9d89d717b5356e5b9e9/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_sag.py | closed | completed | false | 4 | [
"stale"
] | [] | 2023-12-26T04:44:59Z | 2024-01-27T03:28:33Z | 2024-01-27T03:28:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xinli2008 | 71,311,087 | MDQ6VXNlcjcxMzExMDg3 | User | false |
huggingface/diffusers | 2,056,124,626 | I_kwDOHa8MBc56jfjS | 6,333 | https://github.com/huggingface/diffusers/issues/6333 | https://api.github.com/repos/huggingface/diffusers/issues/6333 | FINETUNE PRE-TRAINED LORA BY PEFT | ### Describe the bug
Hi there, I am going to fine-tune LCM LoRA on my own pre-trained models by the PR https://github.com/huggingface/diffusers/blob/87f87a70631d01e765d2e8cdbc1078a6e2908312/examples/consistency_distillation/train_lcm_distill_lora_sdxl.py#L811. However, I don't know how to load pre-trained LCM LoRA ... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-12-26T07:29:48Z | 2023-12-27T14:20:18Z | 2023-12-27T14:20:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | George0726 | 38,740,075 | MDQ6VXNlcjM4NzQwMDc1 | User | false |
huggingface/diffusers | 2,056,183,631 | I_kwDOHa8MBc56jt9P | 6,335 | https://github.com/huggingface/diffusers/issues/6335 | https://api.github.com/repos/huggingface/diffusers/issues/6335 | StableDiffusionControlNetImg2ImgPipeline can't follow pipeline parameters to control the size of output | ### Describe the bug
I wish to turn an image of resolution 512x631 into a new image of the same resolution. It is vital that they should have the same resolution, otherwise I can't use them in downstream tasks. And the StableDiffusionControlNetImg2ImgPipeline keeps returning images of resolution 512x624, regardless of... | closed | completed | false | 7 | [
"bug"
] | [
"yiyixuxu"
] | 2023-12-26T08:43:39Z | 2023-12-29T02:57:55Z | 2023-12-29T02:57:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sunhaozhepy | 73,462,159 | MDQ6VXNlcjczNDYyMTU5 | User | false |
huggingface/diffusers | 2,056,283,330 | I_kwDOHa8MBc56kGTC | 6,336 | https://github.com/huggingface/diffusers/issues/6336 | https://api.github.com/repos/huggingface/diffusers/issues/6336 | TensorRT pipeline: Returns an error when compiling VAE when using AutoencoderTiny | ### Describe the bug
TensorRT returns an error when compiling VAE when using AutoencoderTiny.
### Reproduction
```
import random
from PIL import Image
import torch
from diffusers import (
LCMScheduler,
AutoencoderTiny,
DPMSolverMultistepScheduler,
)
from diffusers.pipelines.stable_diffusion impo... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-12-26T10:37:43Z | 2023-12-27T05:14:27Z | 2023-12-27T05:14:27Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Jannchie | 29,743,310 | MDQ6VXNlcjI5NzQzMzEw | User | false |
huggingface/diffusers | 2,056,688,083 | I_kwDOHa8MBc56lpHT | 6,345 | https://github.com/huggingface/diffusers/issues/6345 | https://api.github.com/repos/huggingface/diffusers/issues/6345 | add padding_mask_crop to all inpaint pipelines | We just merged in this PR https://github.com/huggingface/diffusers/pull/6072 that introduces `padding_mask_crop` argument for stable diffusion inpaint pipeline for the first time
It is similar to the feature in auto1111 when you select "inpaint area = only masked" - basically, it will crop out a region around the ma... | closed | completed | false | 4 | [
"good first issue",
"help wanted",
"contributions-welcome"
] | [] | 2023-12-26T20:30:20Z | 2024-01-22T05:42:24Z | 2024-01-22T05:42:24Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 2,056,867,460 | I_kwDOHa8MBc56mU6E | 6,347 | https://github.com/huggingface/diffusers/issues/6347 | https://api.github.com/repos/huggingface/diffusers/issues/6347 | Add GlueGen | ### Model/Pipeline/Scheduler description
GlueGen is a minimal adapter that allow alignment between any encoder (Text Encoder of different language, Multilingual Roberta, AudioClip) and CLIP text encoder used in standard Stable Diffusion model.
Below are the results produced by using community available weights in E... | closed | completed | false | 2 | [
"community-examples",
"contributions-welcome"
] | [] | 2023-12-27T03:24:34Z | 2024-01-05T12:05:42Z | 2024-01-05T12:05:42Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rootonchair | 23,548,268 | MDQ6VXNlcjIzNTQ4MjY4 | User | false |
huggingface/diffusers | 2,056,891,204 | I_kwDOHa8MBc56matE | 6,348 | https://github.com/huggingface/diffusers/issues/6348 | https://api.github.com/repos/huggingface/diffusers/issues/6348 | "Enhancing `diffusers.AutoencoderKL` for 3D Image Support in Medical and Video Applications" | **Is your feature request related to a problem? Please describe.**
Currently, the VAE in the `diffusers` library supports processing two-dimensional images as input. This limitation makes it less convenient for applications that require handling three-dimensional data, such as medical images, point clouds, and video d... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-12-27T04:15:01Z | 2024-02-03T15:04:31Z | 2024-02-03T15:04:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yusijin02 | 101,894,130 | U_kgDOBhLH8g | User | false |
huggingface/diffusers | 2,057,077,938 | I_kwDOHa8MBc56nISy | 6,351 | https://github.com/huggingface/diffusers/issues/6351 | https://api.github.com/repos/huggingface/diffusers/issues/6351 | when i run train_lcm_distill_lora_sdxl.py, it cause error:'DistributedDataParallel' object has no attribute 'disable_adapters' | ### Describe the bug
when i run train_lcm_distill_lora_sdxl.py, it cause error:'DistributedDataParallel' object has no attribute 'disable_adapters'
### Reproduction
export MODEL_NAME="/home/kas/general_model/diffusers_workspace/text_to_image/cvitai/dreamshaper_xl1.0"
export OUTPUT_DIR="/home/kas/kas_workspace/cjt/s... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-12-27T08:31:56Z | 2023-12-29T09:16:49Z | 2023-12-28T02:07:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | cjt222 | 17,508,662 | MDQ6VXNlcjE3NTA4NjYy | User | false |
huggingface/diffusers | 2,057,107,130 | I_kwDOHa8MBc56nPa6 | 6,352 | https://github.com/huggingface/diffusers/issues/6352 | https://api.github.com/repos/huggingface/diffusers/issues/6352 | how to choose save precision for lora file in training | I'm confused about my lora precision(fp16,bf16,float) and whether i can choose precision about my lora weights. I searched for the params about the **StableDiffusionXLPipeline.save_lora_weights** function used to save lora in sdxl text2img training script and didnt find params like 'save_precision' or sth.
anyone ca... | closed | not_planned | false | 1 | [] | [] | 2023-12-27T09:02:47Z | 2023-12-28T08:21:29Z | 2023-12-28T08:21:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DoctorTar | 147,801,017 | U_kgDOCM9DuQ | User | false |
huggingface/diffusers | 2,057,109,388 | I_kwDOHa8MBc56nP-M | 6,353 | https://github.com/huggingface/diffusers/issues/6353 | https://api.github.com/repos/huggingface/diffusers/issues/6353 | get error when following the training sdxl controlnet example | I followed the example here (https://github.com/huggingface/diffusers/blob/main/examples/controlnet/README_sdxl.md) and get error:
```
2023-12-27 17:01:07.430871: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the foll... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-12-27T09:05:05Z | 2024-01-27T03:28:01Z | 2024-01-27T03:28:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | universewill | 6,790,730 | MDQ6VXNlcjY3OTA3MzA= | User | false |
huggingface/diffusers | 2,057,126,614 | I_kwDOHa8MBc56nULW | 6,354 | https://github.com/huggingface/diffusers/issues/6354 | https://api.github.com/repos/huggingface/diffusers/issues/6354 | Add LoRA support for ControlNet | **Is your feature request related to a problem? Please describe.**
Recently I've been trying to port [LooseControl](https://github.com/shariqfarooq123/LooseControl) to A1111 WebUI and ComfyUI. The model weights are just a LoRA of ControlNet-depth, but to load the LoRA into ControlNet, the author had to hack thourgh `C... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-12-27T09:22:03Z | 2024-05-10T15:43:50Z | 2024-01-27T03:27:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ifsheldon | 39,153,080 | MDQ6VXNlcjM5MTUzMDgw | User | false |
huggingface/diffusers | 2,057,397,861 | I_kwDOHa8MBc56oWZl | 6,357 | https://github.com/huggingface/diffusers/issues/6357 | https://api.github.com/repos/huggingface/diffusers/issues/6357 | Difference API results and local diffuser results | ### Describe the bug
Hi, I train a diffuser model and push to the hub. And I find that I cannot reproduce the result of the hub api using codes like this and it seems the results in the API are better.
```
from diffusers import StableDiffusionPipeline, EulerDiscreteScheduler
import torch
model_id = "my_repo"
sc... | closed | completed | false | 13 | [
"bug",
"stale"
] | [] | 2023-12-27T13:51:46Z | 2024-03-03T04:24:30Z | 2024-03-03T04:24:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CaffreyR | 84,232,793 | MDQ6VXNlcjg0MjMyNzkz | User | false |
huggingface/diffusers | 2,057,473,074 | I_kwDOHa8MBc56oowy | 6,359 | https://github.com/huggingface/diffusers/issues/6359 | https://api.github.com/repos/huggingface/diffusers/issues/6359 | EulerDiscreteScheduler not have get_velocity function and add_noise get error for prediction_type == "v_prediction"? | ### Describe the bug
Finetune stable video diffusion but found two problems:
1. EulerDiscreteScheduler not have get_velocity function ?
2. EulerDiscreteScheduler add_noise get error for prediction_type == "v_prediction"?
### Reproduction
1. EulerDiscreteScheduler not have get_velocity function ?
https://... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-12-27T14:59:45Z | 2024-01-15T14:13:24Z | 2023-12-31T04:28:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiarenyf | 12,405,578 | MDQ6VXNlcjEyNDA1NTc4 | User | false |
huggingface/diffusers | 2,057,670,996 | I_kwDOHa8MBc56pZFU | 6,362 | https://github.com/huggingface/diffusers/issues/6362 | https://api.github.com/repos/huggingface/diffusers/issues/6362 | When fine-tuning sdxl I get Out of Memory when mapping the dataset | ### Describe the bug
When using `train_text_to_image_sdxl.py` I'm getting out of memory when mapping the dataset:
```
Resolving data files: 100% 1512/1512 [00:00<00:00, 44098.99it/s]
Map: 0% 0/1511 [00:44<?, ? examples/s]
Traceback (most recent call last):
File "/content/drive/MyDrive/Colab/SD_LoRA/train_... | closed | completed | false | 11 | [
"bug"
] | [] | 2023-12-27T18:35:59Z | 2024-01-02T15:28:52Z | 2024-01-02T15:28:15Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AeroDEmi | 44,657,733 | MDQ6VXNlcjQ0NjU3NzMz | User | false |
huggingface/diffusers | 2,057,692,396 | I_kwDOHa8MBc56peTs | 6,363 | https://github.com/huggingface/diffusers/issues/6363 | https://api.github.com/repos/huggingface/diffusers/issues/6363 | train_text_to_image_lora.py raise ValueError("Attempting to unscale FP16 gradients.") | ### Describe the bug
When looking at the **examples/text_to_image** documentation, I experimented with the **train_text_to_image_lora.py** following the examples in the documentation. But I found that the run with ```raise ValueError("Attempting to unscale FP16 gradients.")``` error.
I found that the cause of t... | closed | completed | false | 16 | [
"bug",
"stale"
] | [] | 2023-12-27T19:04:30Z | 2025-04-11T14:57:55Z | 2024-10-09T15:25:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | billvsme | 3,324,744 | MDQ6VXNlcjMzMjQ3NDQ= | User | false |
huggingface/diffusers | 2,057,805,887 | I_kwDOHa8MBc56p6A_ | 6,366 | https://github.com/huggingface/diffusers/issues/6366 | https://api.github.com/repos/huggingface/diffusers/issues/6366 | train_dreambooth_lora_sdxl_advanced.py and train_dreambooth_lora_sdxl.py do not load previously saved checkpoints correctly | ### Describe the bug
both examples/dreambooth/train_dreambooth_lora_sdxl.py and examples/advanced_diffusion_training/train_dreambooth_lora_sdxl_advanced.py seem to have an issue when resuming training from a previously saved checkpoint.
Training and saving checkpoints seems to work correctly, however, when resuming... | closed | completed | false | 15 | [
"bug",
"stale"
] | [] | 2023-12-27T22:05:03Z | 2024-01-29T15:18:09Z | 2024-01-29T15:18:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | prushik | 1,657,917 | MDQ6VXNlcjE2NTc5MTc= | User | false |
huggingface/diffusers | 2,057,896,707 | I_kwDOHa8MBc56qQMD | 6,368 | https://github.com/huggingface/diffusers/issues/6368 | https://api.github.com/repos/huggingface/diffusers/issues/6368 | ValueError: Checkpoint not supported because layer lora_unet_down_blocks_0_downsamplers_0_conv.alpha not supported. | ### Describe the bug
Hello, I'm loading the lora model from civtai, which includes convolutional layers, cause error. Are diffusers not supported now?
ValueError: Checkpoint not supported because layer lora_unet_down_blocks_0_downsamplers_0_conv.alpha not supported.
### Reproduction
from diffusers import StableDi... | closed | completed | false | 10 | [
"bug",
"stale",
"lora"
] | [] | 2023-12-28T01:43:15Z | 2024-07-10T23:23:36Z | 2024-03-23T03:10:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | cjt222 | 17,508,662 | MDQ6VXNlcjE3NTA4NjYy | User | false |
huggingface/diffusers | 2,058,038,392 | I_kwDOHa8MBc56qyx4 | 6,370 | https://github.com/huggingface/diffusers/issues/6370 | https://api.github.com/repos/huggingface/diffusers/issues/6370 | How to use diffusers lora in the AUTOMATIC1111 | Thanks for your great work, I use the train_text_to_image_lora_sdxl.py to train my custom dataset and get these output, And I get the good result. But I want to use the AUTOMATIC1111 to use the lora weight, I move the pytorch_lora_weights to the AUTOMATIC1111 lora folder But get the error report:`AssertionError: conver... | closed | completed | false | 3 | [] | [] | 2023-12-28T06:17:19Z | 2024-01-02T13:38:26Z | 2024-01-02T13:38:26Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chongxian | 18,145,013 | MDQ6VXNlcjE4MTQ1MDEz | User | false |
huggingface/diffusers | 2,058,115,182 | I_kwDOHa8MBc56rFhu | 6,371 | https://github.com/huggingface/diffusers/issues/6371 | https://api.github.com/repos/huggingface/diffusers/issues/6371 | get assert error when following controlnet_xs example | I followed the controlnet xs example (https://github.com/UmerHA/diffusers/tree/cnxs-training/examples/controlnet_xs) and get assert error.
```
Detected kernel version 5.4.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum ve... | closed | completed | false | 2 | [] | [] | 2023-12-28T07:53:57Z | 2023-12-28T08:15:24Z | 2023-12-28T08:15:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | universewill | 6,790,730 | MDQ6VXNlcjY3OTA3MzA= | User | false |
huggingface/diffusers | 2,058,216,870 | I_kwDOHa8MBc56reWm | 6,372 | https://github.com/huggingface/diffusers/issues/6372 | https://api.github.com/repos/huggingface/diffusers/issues/6372 | [Pipeline] SVD + ControlNet | ### Model/Pipeline/Scheduler description
Recently, AnimateDiff+ControlNet support was added to diffusers. I was hoping to try and implement SVD+ControlNet too but didn't find the time, however @CiaraStrawberry has a great implementation for it. Thank you for this very cool pipeline!
It would be awesome if diffuse... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-12-28T09:41:36Z | 2024-02-04T15:05:02Z | 2024-02-04T15:05:02Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | a-r-r-o-w | 72,266,394 | MDQ6VXNlcjcyMjY2Mzk0 | User | false |
huggingface/diffusers | 2,058,306,208 | I_kwDOHa8MBc56r0Kg | 6,374 | https://github.com/huggingface/diffusers/issues/6374 | https://api.github.com/repos/huggingface/diffusers/issues/6374 | 加载 qwen微调后的lora 权重 | ### Describe the bug
加载qwen 微调后的lora ，命令：
export USE_MODELSCOPE_HUB=1
CUDA_VISIBLE_DEVICES=0 python src/cli_demo.py \
--model_name_or_path qwen/Qwen-7B \
--adapter_name_or_path ./output/sft/checkpoint-4000/ \
--template default \
--finetuning_type lora
### Reproduction
1
### Logs
```shell
12... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-12-28T11:08:02Z | 2024-05-23T05:00:16Z | 2024-02-04T15:05:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fangcao1314 | 15,755,678 | MDQ6VXNlcjE1NzU1Njc4 | User | false |
huggingface/diffusers | 2,058,439,973 | I_kwDOHa8MBc56sU0l | 6,377 | https://github.com/huggingface/diffusers/issues/6377 | https://api.github.com/repos/huggingface/diffusers/issues/6377 | Unloading LoRA raises errors on refiner SDXL models when employing only ONE text encoder | **Unloading LoRA raises errors on refiner SDXL models when employing only ONE text encoder:**
https://github.com/huggingface/diffusers/blob/7f551e29ff4ad05615cb38530a8940811f9e5936/src/diffusers/loaders/lora.py#L967C9-L967C28
on self._remove_text_encoder_monkey_patch()
https://github.com/huggingface/diffusers/bl... | closed | not_planned | false | 16 | [
"stale"
] | [] | 2023-12-28T13:24:17Z | 2024-11-21T15:44:19Z | 2024-11-21T15:44:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | arnonrodman | 4,532,476 | MDQ6VXNlcjQ1MzI0NzY= | User | false |
huggingface/diffusers | 2,058,694,363 | I_kwDOHa8MBc56tS7b | 6,379 | https://github.com/huggingface/diffusers/issues/6379 | https://api.github.com/repos/huggingface/diffusers/issues/6379 | unet.save_pretrained() only save a small file (2kb or 2Mb) | ### Describe the bug
I try to finetune the text-to-video model using the `UNetSpatioTemporalConditionModel` as the U-Net backbone on my private dataset, and the training process was normal, but when I need to use `unet.save_pretrained(os.path.join(args.output_dir, "unet"))`, it only save a config file and a very sma... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-12-28T18:01:47Z | 2024-11-19T09:19:02Z | 2024-11-19T09:19:02Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Wang-Xiaodong1899 | 44,627,574 | MDQ6VXNlcjQ0NjI3NTc0 | User | false |
huggingface/diffusers | 2,059,011,083 | I_kwDOHa8MBc56ugQL | 6,382 | https://github.com/huggingface/diffusers/issues/6382 | https://api.github.com/repos/huggingface/diffusers/issues/6382 | I am encountering a key error with the IP adapter when using the train_lcm_distill_lora_sdxl.py script to train the LCM Lora model for SDXL. After training, when I use it together with SD XL and IP adapter, it throws an IP adapter key error. | ### Describe the bug
I am encountering a key error with the IP adapter when using the train_lcm_distill_lora_sdxl.py script to train the LCM Lora model for SDXL. After training, when I use it together with SD XL and IP adapter, it throws an IP adapter key error.
Loading adapter weights from state_dict led to unexpe... | closed | completed | false | 11 | [
"bug"
] | [] | 2023-12-29T02:41:54Z | 2024-01-02T03:12:16Z | 2024-01-02T03:12:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | cjt222 | 17,508,662 | MDQ6VXNlcjE3NTA4NjYy | User | false |
huggingface/diffusers | 2,059,015,137 | I_kwDOHa8MBc56uhPh | 6,383 | https://github.com/huggingface/diffusers/issues/6383 | https://api.github.com/repos/huggingface/diffusers/issues/6383 | Does diffusers get_scheduler() support multi-stage learning rates? | ### Describe the bug
I want to train my model with mutli-stage lr, see to accelerate training. I want to assign a larger lr the first stage, then a smaller one in the second stage. I don't see a example or method in the diffusers `get_scheduler() ` methods. So does diffusers support this implementation ?
### Repr... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-12-29T02:52:52Z | 2023-12-29T04:05:02Z | 2023-12-29T04:05:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | stardusts-hj | 54,565,340 | MDQ6VXNlcjU0NTY1MzQw | User | false |
huggingface/diffusers | 2,059,185,898 | I_kwDOHa8MBc56vK7q | 6,384 | https://github.com/huggingface/diffusers/issues/6384 | https://api.github.com/repos/huggingface/diffusers/issues/6384 | How to map A1111 reference_only parameters into diffusers? | Thanks for the community to implement the reference_only functionality in A1111, but how can the parameters correspond to each other? I have tried to reproduce the effect of webui in the diffusers library, but I can't seem to do it. I'm using the StableDiffusionReferencePipeline community pipeline.
My questions are... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-12-29T08:16:15Z | 2024-01-28T15:29:43Z | 2024-01-28T15:29:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Logos23333 | 26,246,545 | MDQ6VXNlcjI2MjQ2NTQ1 | User | false |
huggingface/diffusers | 2,059,199,791 | I_kwDOHa8MBc56vOUv | 6,385 | https://github.com/huggingface/diffusers/issues/6385 | https://api.github.com/repos/huggingface/diffusers/issues/6385 | Abnormally high VRAM required when using StableVideoDiffusionPipeline | ### Describe the bug
Accordingto the doc https://huggingface.co/docs/diffusers/v0.24.0/en/using-diffusers/svd#low-memory it says applying all the low-memory tricks shall make the VRAM needed no more than 8-GB. Yet I still met CUDA OOM problem with an abnormal VRAM requirement of about 40 GB. I have the code almost tot... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-12-29T08:36:37Z | 2024-01-03T02:19:22Z | 2024-01-03T02:19:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | feiyangsuo | 47,778,268 | MDQ6VXNlcjQ3Nzc4MjY4 | User | false |
huggingface/diffusers | 2,059,214,483 | I_kwDOHa8MBc56vR6T | 6,386 | https://github.com/huggingface/diffusers/issues/6386 | https://api.github.com/repos/huggingface/diffusers/issues/6386 | Improve sample process‘ performance | HI！
I want to obtain the x_0_prediction when the sampling process is halfway through. Is it possible to achieve this? | closed | completed | false | 8 | [] | [] | 2023-12-29T08:57:25Z | 2024-01-05T00:58:49Z | 2024-01-05T00:58:49Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ShaochengShen | 94,623,045 | U_kgDOBaPVRQ | User | false |
huggingface/diffusers | 2,060,253,686 | I_kwDOHa8MBc56zPn2 | 6,388 | https://github.com/huggingface/diffusers/issues/6388 | https://api.github.com/repos/huggingface/diffusers/issues/6388 | gradient-checkpointing ignored in SDXL+LoRA training | There is `--gradient-checkpointing` option, but never handled. Is this intended?
https://github.com/huggingface/diffusers/blob/9f283b01d20aa8d9e5251dbaad19432bfb2ffe11/examples/text_to_image/train_text_to_image_lora_sdxl.py#L302-306 | closed | completed | false | 1 | [] | [] | 2023-12-29T16:49:02Z | 2024-01-01T03:21:05Z | 2024-01-01T03:21:05Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | 2510 | 8,462,436 | MDQ6VXNlcjg0NjI0MzY= | User | false |
huggingface/diffusers | 1,385,333,901 | I_kwDOHa8MBc5SkoSN | 639 | https://github.com/huggingface/diffusers/issues/639 | https://api.github.com/repos/huggingface/diffusers/issues/639 | Parameter Initialization doesn't match with Latent Diffusion Model | ### Describe the bug
Thanks for releasing this great work, it really makes using diffusion a easy thing!
But later when I tried to train a `UNet2DConditionModel` from scratch, I found the parameter initialization method is a little different from the official implementation.
For example,
https://github.com/h... | closed | completed | false | 4 | [
"stale"
] | [
"patil-suraj"
] | 2022-09-26T02:56:50Z | 2022-11-28T15:03:47Z | 2022-11-28T15:03:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Karbo123 | 42,935,526 | MDQ6VXNlcjQyOTM1NTI2 | User | false |
huggingface/diffusers | 2,060,335,273 | I_kwDOHa8MBc56zjip | 6,391 | https://github.com/huggingface/diffusers/issues/6391 | https://api.github.com/repos/huggingface/diffusers/issues/6391 | Distributed training on 2 GPUs but with Batch Size 1? | Hello,
I am trying to see if I can train a Stable Diffusion model on two GPUs, but there should only be one model that is loaded onto both GPUs, instead of having each GPU load its own model in what would be basically training on BS 2 but with two GPUs.
Fine-tuning SDXL with the text encoder and a high-quality op... | closed | completed | false | 1 | [] | [] | 2023-12-29T19:03:35Z | 2023-12-30T03:50:37Z | 2023-12-30T03:50:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | johnman3032 | 134,149,718 | U_kgDOB_72Vg | User | false |
huggingface/diffusers | 2,060,405,473 | I_kwDOHa8MBc56z0rh | 6,392 | https://github.com/huggingface/diffusers/issues/6392 | https://api.github.com/repos/huggingface/diffusers/issues/6392 | load_lora_weights function fails to load custom lora | ### Describe the bug
**Base model**: segmind/tiny-sd
**LoRA**: akameswa/lcm-lora-tiny-sd
**Error:**
Got the following error using load_lora_weights.
`ValueError: Target modules {'base_model.model.up_blocks.0.attentions.0.transformer_blocks.0.attn1.to_q', ...} not found in the base model. Please check the target ... | closed | completed | false | 11 | [
"bug",
"stale",
"training",
"peft"
] | [] | 2023-12-29T21:15:37Z | 2024-03-17T15:07:06Z | 2024-03-17T15:07:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | akameswa | 124,394,284 | U_kgDOB2obLA | User | false |
huggingface/diffusers | 2,060,417,152 | I_kwDOHa8MBc56z3iA | 6,393 | https://github.com/huggingface/diffusers/issues/6393 | https://api.github.com/repos/huggingface/diffusers/issues/6393 | Private HF link download failing through "from_single_file" | ### Describe the bug
Getting a 401 error while running ```from_single_file()``` on private HF model link.
Same access token and url working fine while downloading directly using ```huggingface_hub.hf_hub_download``` and on ```diffusers==0.24.0```
### Reproduction
On ```diffusers > 0.24.0 ```
```
pipe = Stable... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-12-29T21:39:26Z | 2023-12-30T10:44:25Z | 2023-12-30T10:44:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nivedwho | 62,166,124 | MDQ6VXNlcjYyMTY2MTI0 | User | false |
huggingface/diffusers | 2,060,641,324 | I_kwDOHa8MBc560uQs | 6,397 | https://github.com/huggingface/diffusers/issues/6397 | https://api.github.com/repos/huggingface/diffusers/issues/6397 | Add a section about how training data resizing might affect the quality of the end models | Have been chatting with @bghira on our Discord forum about the negative effects of not carefully resizing the training data.
@bghira noticed that the way we resize the images in our training examples can introduce unwanted artifacts in the samples generated by the end fine-tuned model. In short, it ruins the fideli... | closed | completed | false | 18 | [] | [
"asomoza"
] | 2023-12-30T05:33:50Z | 2025-04-08T07:20:22Z | 2025-04-08T06:56:08Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 2,060,679,003 | I_kwDOHa8MBc5603db | 6,398 | https://github.com/huggingface/diffusers/issues/6398 | https://api.github.com/repos/huggingface/diffusers/issues/6398 | Memory leak? | ### Describe the bug
When I load a model and then have a loop to continuously generate images, the processing time gets longer and longer. Is there a memory leak, or is there something else I need to do to clean up after each loop?
This is a loop generating SD 1.5 images:
![image](https://github.com/huggingface... | closed | completed | false | 12 | [
"bug",
"stale"
] | [] | 2023-12-30T08:28:50Z | 2024-02-06T15:04:57Z | 2024-02-06T15:04:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 2,060,712,120 | I_kwDOHa8MBc560_i4 | 6,399 | https://github.com/huggingface/diffusers/issues/6399 | https://api.github.com/repos/huggingface/diffusers/issues/6399 | pytorch_lora_weights lora is not visible inside Automatic nor ComfyUI | ### Describe the bug
I trained a lora with https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/SDXL_DreamBooth_LoRA_.ipynb#scrollTo=GOiZSXHfx3mJ
this colab. Everything works fine with inference inside Colab. But that model weights file which named pytorch_lora_weights.safetensors is n... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-12-30T10:53:53Z | 2023-12-30T19:06:27Z | 2023-12-30T19:06:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ahgsql | 35,695,543 | MDQ6VXNlcjM1Njk1NTQz | User | false |
huggingface/diffusers | 1,385,364,603 | I_kwDOHa8MBc5Skvx7 | 640 | https://github.com/huggingface/diffusers/issues/640 | https://api.github.com/repos/huggingface/diffusers/issues/640 | Incorrect type annotations on `StableDiffusionPipeline` | ### Describe the bug
`StableDiffusionPipeline::__call__` has some parameters annotated as `Optional` when they really aren't (i.e. they shouldn't be `None`)
A potential fix would be to just remove the `Optional` annotation on these parameters, as in this commit: https://github.com/tasercake/diffusers/commit/7a5c37a... | closed | completed | false | 4 | [
"stale"
] | [] | 2022-09-26T03:42:12Z | 2022-11-01T10:30:11Z | 2022-11-01T10:30:11Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tasercake | 13,855,549 | MDQ6VXNlcjEzODU1NTQ5 | User | false |
huggingface/diffusers | 2,060,943,712 | I_kwDOHa8MBc5614Fg | 6,406 | https://github.com/huggingface/diffusers/issues/6406 | https://api.github.com/repos/huggingface/diffusers/issues/6406 | [Community] The "lpw_stable_diffusion_xl" pipeline doesn't support new run params | ### Describe the bug
Passing args like `clip_skip` or `cfg_scale` to a pipeline instantiated with the "lpw_stable_diffusion_xl" pipeline cause a crash.
### Reproduction
```python
from diffusers import *
from diffusers.utils import make_image_grid
import torch
import numpy as np
universal_args = {
'cac... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-12-31T05:10:54Z | 2024-01-02T15:28:47Z | 2024-01-02T15:28:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | T145 | 1,214,129 | MDQ6VXNlcjEyMTQxMjk= | User | false |
huggingface/diffusers | 2,060,961,306 | I_kwDOHa8MBc5618Ya | 6,407 | https://github.com/huggingface/diffusers/issues/6407 | https://api.github.com/repos/huggingface/diffusers/issues/6407 | AnyText: Multilingual Visual Text Generation And Editing | ### Model/Pipeline/Scheduler description
From the [repository](https://github.com/tyxsspa/AnyText):
> AnyText comprises a diffusion pipeline with two primary elements: an auxiliary latent module and a text embedding module. The former uses inputs like text glyph, position, and masked image to generate latent featu... | closed | completed | false | 15 | [
"community-examples",
"stale",
"contributions-welcome"
] | [] | 2023-12-31T06:44:29Z | 2025-03-10T20:19:42Z | 2025-03-10T20:19:42Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 2,060,994,771 | I_kwDOHa8MBc562EjT | 6,409 | https://github.com/huggingface/diffusers/issues/6409 | https://api.github.com/repos/huggingface/diffusers/issues/6409 | ValueError when `to('cuda')` on Value-guided planning pipeline | ### Describe the bug
When I run the Value-guided planning pipeline colab <https://colab.research.google.com/drive/1rXm8CX4ZdN5qivjJ2lhwhkOmt_m0CvU0> with `to('cuda')` to speed up the diffusion process. It gave me a ValueError.
### Reproduction
<https://colab.research.google.com/drive/1SFl7daLQxd8QyHJP6ndSPznSF... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-12-31T09:09:50Z | 2024-01-02T05:46:23Z | 2024-01-02T05:46:23Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hilookas | 15,031,653 | MDQ6VXNlcjE1MDMxNjUz | User | false |
huggingface/diffusers | 2,061,152,246 | I_kwDOHa8MBc562q_2 | 6,412 | https://github.com/huggingface/diffusers/issues/6412 | https://api.github.com/repos/huggingface/diffusers/issues/6412 | ControlNet Annotators to ONNX | **Is your feature request related to a problem? Please describe.**
At the moment I am able to use [Canny Edge Detection](https://github.com/dakenf/diffusers.js/pull/8), [Pose Estimation and Semantic Segmentation](https://github.com/dakenf/diffusers.js/pull/11) (all in JavaScript) in [diffusers.js](https://github.com/d... | closed | completed | false | 2 | [] | [] | 2023-12-31T19:34:31Z | 2024-01-01T22:45:54Z | 2024-01-01T22:45:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jdp8 | 68,886,123 | MDQ6VXNlcjY4ODg2MTIz | User | false |
huggingface/diffusers | 2,061,382,594 | I_kwDOHa8MBc563jPC | 6,413 | https://github.com/huggingface/diffusers/issues/6413 | https://api.github.com/repos/huggingface/diffusers/issues/6413 | Dreambooth SDXL example inconsistent results | ### Describe the bug
When running the dream booth sdxl code exactly as per the example readme, the results generated are not consistent.
I have had a look at others users push to hub uploads and it appears others have the same issue, i have seen people uploading the same example out of the box and all of the dogs l... | closed | completed | false | 7 | [
"bug"
] | [] | 2024-01-01T07:27:10Z | 2024-01-03T09:19:21Z | 2024-01-03T09:19:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ComandoLando | 22,780,569 | MDQ6VXNlcjIyNzgwNTY5 | User | false |
huggingface/diffusers | 2,061,416,009 | I_kwDOHa8MBc563rZJ | 6,414 | https://github.com/huggingface/diffusers/issues/6414 | https://api.github.com/repos/huggingface/diffusers/issues/6414 | Clip Skip with LPW pipeline? | Hi,
I want to use clip Skip with LPW stable diffusion community pipeline, but it doesn't support it, like the StableDiffusionPipleline does in which I can pass a clip_skip param to be used to that specific inference and it can be changed for other inferences, how can I achieve this functionality similar to StableDiff... | closed | completed | false | 4 | [
"stale"
] | [] | 2024-01-01T08:37:22Z | 2024-02-08T15:05:19Z | 2024-02-08T15:05:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Adeel-Intizar | 56,041,154 | MDQ6VXNlcjU2MDQxMTU0 | User | false |
huggingface/diffusers | 2,061,440,113 | I_kwDOHa8MBc563xRx | 6,415 | https://github.com/huggingface/diffusers/issues/6415 | https://api.github.com/repos/huggingface/diffusers/issues/6415 | IP-Adapter-FaceID and FaceID plus not support | ### Describe the bug
```
diffusers\loaders\unet.py", line 780, in _load_ip_adapter_weights
num_image_text_embeds = state_dict["image_proj"]["latents"].shape[1]
KeyError: 'latents'
```
### Reproduction
```
import cv2
from insightface.app import FaceAnalysis
import numpy as np
from PIL import Image... | closed | completed | false | 28 | [
"bug",
"stale"
] | [] | 2024-01-01T09:26:59Z | 2024-06-12T12:29:34Z | 2024-02-08T15:05:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhaoyun0071 | 35,762,050 | MDQ6VXNlcjM1NzYyMDUw | User | false |
huggingface/diffusers | 2,061,540,693 | I_kwDOHa8MBc564J1V | 6,418 | https://github.com/huggingface/diffusers/issues/6418 | https://api.github.com/repos/huggingface/diffusers/issues/6418 | Finetune stable video diffusion, but the output video frames are all the same as the first frame, why ? | ### Describe the bug
Finetune stable video diffusion, but the output video frames are all the same as the first frame, why ?
### Reproduction
Finetune stable video diffusion, but the output video frames are all the same as the first frame, why ?
### Logs
_No response_
### System Info
Finetune stable video diffus... | closed | completed | false | 3 | [
"bug"
] | [] | 2024-01-01T12:46:13Z | 2024-01-12T03:48:27Z | 2024-01-02T01:37:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jiarenyf | 12,405,578 | MDQ6VXNlcjEyNDA1NTc4 | User | false |
huggingface/diffusers | 2,062,035,121 | I_kwDOHa8MBc566Cix | 6,423 | https://github.com/huggingface/diffusers/issues/6423 | https://api.github.com/repos/huggingface/diffusers/issues/6423 | can't reproduce fill50k's results | **Is your feature request related to a problem? Please describe.**
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...].
**Describe the solution you'd like.**
A clear and concise description of what you want to happen.
**Describe alternatives you've considered.**
A clear... | closed | completed | false | 3 | [
"stale"
] | [] | 2024-01-02T07:24:49Z | 2024-02-10T15:04:33Z | 2024-02-10T15:04:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dulibubai | 36,948,738 | MDQ6VXNlcjM2OTQ4NzM4 | User | false |
huggingface/diffusers | 2,062,458,628 | I_kwDOHa8MBc567p8E | 6,427 | https://github.com/huggingface/diffusers/issues/6427 | https://api.github.com/repos/huggingface/diffusers/issues/6427 | Conversion of model from diffusers to .ckpt does not work properly | ### Describe the bug
I trained model with [this script](https://github.com/huggingface/diffusers/blob/main/examples/research_projects/dreambooth_inpaint/train_dreambooth_inpaint.py). The result opens and operates normally with StableDiffusionInpaintPipeline.from_pretrained. Then I converted it to .ckpt with [this](h... | closed | completed | false | 5 | [
"bug"
] | [
"DN6"
] | 2024-01-02T13:57:44Z | 2024-02-12T12:00:24Z | 2024-02-12T12:00:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | PMykhailo | 18,684,799 | MDQ6VXNlcjE4Njg0Nzk5 | User | false |
huggingface/diffusers | 2,062,889,985 | I_kwDOHa8MBc569TQB | 6,429 | https://github.com/huggingface/diffusers/issues/6429 | https://api.github.com/repos/huggingface/diffusers/issues/6429 | Lora Usage With Controlnet Pipeline | Can we use both load_lora_weights() and fuse_lora() with controlnet pipelines. @sayakpaul | closed | not_planned | false | 2 | [] | [] | 2024-01-02T20:09:17Z | 2024-01-03T06:27:16Z | 2024-01-03T06:27:15Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | furkantrky | 57,717,026 | MDQ6VXNlcjU3NzE3MDI2 | User | false |
huggingface/diffusers | 2,062,968,873 | I_kwDOHa8MBc569mgp | 6,430 | https://github.com/huggingface/diffusers/issues/6430 | https://api.github.com/repos/huggingface/diffusers/issues/6430 | Custom pipeline docs are not accurate | ### Describe the bug
The description reads as such:
```
custom_pipeline (`str`, *optional*):
<Tip warning={true}>
🧪 This is an experimental feature and may change in the future.
</Tip>
Can be either:
- A string, the... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2024-01-02T21:20:50Z | 2024-02-12T07:34:12Z | 2024-02-11T15:04:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | T145 | 1,214,129 | MDQ6VXNlcjEyMTQxMjk= | User | false |
huggingface/diffusers | 2,063,072,074 | I_kwDOHa8MBc569_tK | 6,431 | https://github.com/huggingface/diffusers/issues/6431 | https://api.github.com/repos/huggingface/diffusers/issues/6431 | None of the pipelines will load anymore | ### Describe the bug
None of the pipelines will load as of today. The call to `from_pretrained` never returns as it is apparently now unable to fetch needed files from huggingface. I tried with and without using a HF secret key and the result is the same. I've been using various pipelines regularly up until today wi... | closed | completed | false | 18 | [
"bug"
] | [] | 2024-01-02T23:39:48Z | 2024-01-12T08:45:26Z | 2024-01-07T15:35:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | BlakeOne | 30,273,164 | MDQ6VXNlcjMwMjczMTY0 | User | false |
huggingface/diffusers | 2,063,599,581 | I_kwDOHa8MBc57AAfd | 6,434 | https://github.com/huggingface/diffusers/issues/6434 | https://api.github.com/repos/huggingface/diffusers/issues/6434 | Is it possible to add vocabulary to a clip model packaged inside of a distributed SD checkpoint? | I tried to make an account over on huggingface to ask on the forum, but it wont redirect me after the captcha so i cant make an account, so il ask here.
Is it possible to add vocabulary to a clip model packaged inside of a distributed SD checkpoint? I have a checkpoint of the .safetensor variety that i downloaded on... | closed | completed | false | 2 | [
"stale"
] | [] | 2024-01-03T09:28:36Z | 2024-02-12T07:36:11Z | 2024-02-11T15:04:30Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TeKett | 64,377,023 | MDQ6VXNlcjY0Mzc3MDIz | User | false |
huggingface/diffusers | 2,063,632,633 | I_kwDOHa8MBc57AIj5 | 6,435 | https://github.com/huggingface/diffusers/issues/6435 | https://api.github.com/repos/huggingface/diffusers/issues/6435 | Default Width, Height in ControlNet does not initialized | ### Describe the bug
I notice that `width`, `height` variables are not initialized as `self.unet.config.sample_size * self.vae_scale_factor` said in the doc
https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/controlnet/pipeline_controlnet.py#L1024
I'm not sure if this is intentional, else... | closed | completed | false | 7 | [
"bug",
"stale"
] | [
"yiyixuxu"
] | 2024-01-03T09:45:13Z | 2024-02-11T15:04:28Z | 2024-02-11T15:04:28Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rootonchair | 23,548,268 | MDQ6VXNlcjIzNTQ4MjY4 | User | false |
huggingface/diffusers | 2,064,115,770 | I_kwDOHa8MBc57B-g6 | 6,439 | https://github.com/huggingface/diffusers/issues/6439 | https://api.github.com/repos/huggingface/diffusers/issues/6439 | Docs: fp16 link from accelerating SDXL guide 404s | ### Describe the bug
Link that 404s:
https://huggingface.co/docs/diffusers/main/en/optimization/fp16.md
Linked from:
https://huggingface.co/docs/diffusers/main/en/tutorials/fast_diffusion
And in code:
https://github.com/huggingface/diffusers/blob/2e4dc3e25d6abaf5f15d7e5c62db75a8bd63f915/docs/source/en/tutor... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-01-03T14:19:23Z | 2024-01-04T17:59:12Z | 2024-01-04T17:59:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fofr | 319,055 | MDQ6VXNlcjMxOTA1NQ== | User | false |
huggingface/diffusers | 2,064,615,511 | I_kwDOHa8MBc57D4hX | 6,441 | https://github.com/huggingface/diffusers/issues/6441 | https://api.github.com/repos/huggingface/diffusers/issues/6441 | from_pretrained can't work in Google Colab | ### Describe the bug
I try to load model using ddpmpipeline.from_pretrained and loading model stop in line "fetching files"
### Reproduction
I try use all diffusers start from 0.24.1 to last
`
image_pipe = DDPMPipeline.from_pretrained("google/ddpm-ema-church-256").to(device)
`
And when collab load code his... | closed | completed | false | 12 | [
"bug"
] | [] | 2024-01-03T20:36:27Z | 2024-01-12T08:45:45Z | 2024-01-12T08:45:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | WiNE-iNEFF | 41,611,046 | MDQ6VXNlcjQxNjExMDQ2 | User | false |
huggingface/diffusers | 2,064,700,998 | I_kwDOHa8MBc57ENZG | 6,442 | https://github.com/huggingface/diffusers/issues/6442 | https://api.github.com/repos/huggingface/diffusers/issues/6442 | ValueError: Attempting to unscale FP16 gradients. | I am trying to resume training a lora in sdxl but when I try to resume it gives an error `ValueError: Attempting to unscale FP16 gradients.`
It works the first time but when I resume training it gives me that error
```
!accelerate launch --mixed_precision="fp16" /content/train_text_to_image_lora_sdxl.py \
--... | closed | completed | false | 11 | [] | [] | 2024-01-03T21:58:55Z | 2025-12-08T19:43:15Z | 2024-01-30T01:25:03Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | loboere | 15,265,895 | MDQ6VXNlcjE1MjY1ODk1 | User | false |
huggingface/diffusers | 2,064,734,361 | I_kwDOHa8MBc57EViZ | 6,443 | https://github.com/huggingface/diffusers/issues/6443 | https://api.github.com/repos/huggingface/diffusers/issues/6443 | Pipeline StableDiffusionSAGPipeline is not compatible with most of schedulers | ### Describe the bug
Like title says, pipeline `StableDiffusionSAGPipeline` is not compatible with most of samplers - basically only `PNDM`, `DDPM` and `DDIM` work, all others produce the error.
Given this is a built-in pipeline, not a community one I'd expect anything but base samplers to work.
### Reproduction
... | closed | completed | false | 4 | [
"bug"
] | [] | 2024-01-03T22:32:08Z | 2024-01-09T03:54:39Z | 2024-01-09T03:54:39Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 2,064,757,039 | I_kwDOHa8MBc57EbEv | 6,444 | https://github.com/huggingface/diffusers/issues/6444 | https://api.github.com/repos/huggingface/diffusers/issues/6444 | AttributeError: 'CLIPTextModel' object has no attribute 'peft_config' when using default train_dreambooth_lora_sdxl_advanced.py but with checkpointing_steps < max_train_steps | ### Describe the bug
When checkpointing at defined steps, such as every 50 steps I get a 'CLIPTextModel' object has no attribute 'peft_config' while trying to save the model. This only happens when checkpointing_steps < max_train_steps, as the code wouldn't run if it was higher than the train steps
### Reproduction
... | closed | completed | false | 7 | [
"bug"
] | [] | 2024-01-03T22:58:34Z | 2024-07-17T17:56:54Z | 2024-01-05T10:05:27Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | brandostrong | 2,181,431 | MDQ6VXNlcjIxODE0MzE= | User | false |
huggingface/diffusers | 2,064,843,105 | I_kwDOHa8MBc57EwFh | 6,445 | https://github.com/huggingface/diffusers/issues/6445 | https://api.github.com/repos/huggingface/diffusers/issues/6445 | ValueError: --caption_column' value 'text' needs to be one of: image, label | I am trying to train lora sdxl in kaggle but it gives me this error ValueError: --caption_column' value 'text' needs to be one of: image, label
```
!python /kaggle/working/train_text_to_image_lora_sdxl.py \
--pretrained_model_name_or_path "stabilityai/stable-diffusion-xl-base-1.0" \
--train_data_dir data/ \
--... | closed | completed | false | 3 | [
"stale"
] | [] | 2024-01-04T01:05:51Z | 2024-11-26T13:08:45Z | 2024-02-13T15:04:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | loboere | 15,265,895 | MDQ6VXNlcjE1MjY1ODk1 | User | false |
huggingface/diffusers | 2,064,847,410 | I_kwDOHa8MBc57ExIy | 6,446 | https://github.com/huggingface/diffusers/issues/6446 | https://api.github.com/repos/huggingface/diffusers/issues/6446 | Stable Diffusion 2-1 Not Loaded on Colab with Minimal Loading | ### Describe the bug
Within this simple loading script:
```python
from diffusers import DiffusionPipeline
pipeline = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-2-1")
```
The progress is stucked there is no error at all,
Fetching 13 files: 0% 0/13 [05:54<?, ?it/s]
### Reproduction
Colab:... | closed | completed | false | 12 | [
"bug"
] | [] | 2024-01-04T01:12:58Z | 2024-01-12T08:46:07Z | 2024-01-12T08:46:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | maliozer | 1,297,458 | MDQ6VXNlcjEyOTc0NTg= | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.