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 | 1,713,456,901 | I_kwDOHa8MBc5mIUcF | 3,461 | https://github.com/huggingface/diffusers/issues/3461 | https://api.github.com/repos/huggingface/diffusers/issues/3461 | Why Dreambooth script (diffusers edition) has 35% less performance and using from 4Gig vRAM more than Dreambooth script (ShivamShrirao edition )??! | ### Describe the bug
Hi!
I did some train my face with 15 images using from dreambooth script from these repos:
**https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth.py (diffusers edition)
https://github.com/ShivamShrirao/diffusers/blob/main/examples/dreambooth/train_dreamboot... | closed | completed | false | 10 | [
"bug"
] | [] | 2023-05-17T09:04:15Z | 2023-11-15T16:22:32Z | 2023-06-01T12:26:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hosein-moayedi | 38,086,978 | MDQ6VXNlcjM4MDg2OTc4 | User | false |
huggingface/diffusers | 1,713,579,207 | I_kwDOHa8MBc5mIyTH | 3,464 | https://github.com/huggingface/diffusers/issues/3464 | https://api.github.com/repos/huggingface/diffusers/issues/3464 | [Attention processor] Create SDPA versions of attention processors | diffusers currently supports the following PT 2.0 variant of attention processors
* AttnProcessor => AttnProcessor2_0
* AttnAddedKVProcessor => AttnAddedKVProcessor2_0
The following are not supported:
* SlicedAttnProcessor
* SlicedAttnAddedKVProcessor
* ~LoRAAttnProcessor~
* CustomDiffusionAttnProcessor
... | closed | completed | false | 1 | [
"stale"
] | [
"sayakpaul"
] | 2023-05-17T10:17:45Z | 2023-07-08T15:03:07Z | 2023-07-08T15:03:07Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,713,878,779 | I_kwDOHa8MBc5mJ7b7 | 3,467 | https://github.com/huggingface/diffusers/issues/3467 | https://api.github.com/repos/huggingface/diffusers/issues/3467 | ControlNet resume from wrong step with gradient_accumulation_steps | ### Describe the bug
The `initial_global_step` is calculated by following code
```python
# diffusers/examples/controlnet/train_controlnet.py line 980-982
global_step = int(path.split("-")[1])
initial_global_step = global_step * args.gradient_accumulation_steps
```
with `args.gradient_accumulation_steps > 1`, `... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-05-17T13:08:35Z | 2023-05-22T22:20:14Z | 2023-05-22T22:20:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | exhyy | 105,833,611 | U_kgDOBk7kiw | User | false |
huggingface/diffusers | 1,713,891,498 | I_kwDOHa8MBc5mJ-iq | 3,468 | https://github.com/huggingface/diffusers/issues/3468 | https://api.github.com/repos/huggingface/diffusers/issues/3468 | Dreambooth not generating model_index.json and thus is not able to make inference | ### Describe the bug
The problem is described in: https://discuss.huggingface.co/t/dreambooth-not-generating-model-index-json-and-thus-is-not-able-to-make-inference/38640
Re-iterate here:
I'm using `runwayml/stable-diffusion-v1-5` model to train a `dreambooth` with the training script in `examples/dreambooth/tra... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-05-17T13:12:46Z | 2025-08-10T06:43:49Z | 2023-05-22T13:37:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xarthurx | 1,921,878 | MDQ6VXNlcjE5MjE4Nzg= | User | false |
huggingface/diffusers | 1,361,015,460 | I_kwDOHa8MBc5RH3Kk | 347 | https://github.com/huggingface/diffusers/issues/347 | https://api.github.com/repos/huggingface/diffusers/issues/347 | Unable to find a version that satisfies the requirement huggingface-hub to install diffusers | ### Describe the bug
while trying to setup development environment by using command(pip install -e ".[dev]") within diffusers cloned repro folder
I am getting ERROR: Could not find a version that satisfies the requirement huggingface-hub<1.0,>=0.8.1
I have tried downgrading default huggingface_hub = 0.4.1 to 0... | closed | completed | false | 6 | [
"bug"
] | [] | 2022-09-04T02:42:54Z | 2022-09-05T14:39:33Z | 2022-09-05T14:39:04Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rashmimarganiatgithub | 24,475,962 | MDQ6VXNlcjI0NDc1OTYy | User | false |
huggingface/diffusers | 1,714,734,845 | I_kwDOHa8MBc5mNMb9 | 3,475 | https://github.com/huggingface/diffusers/issues/3475 | https://api.github.com/repos/huggingface/diffusers/issues/3475 | "Common Diffusion Noise Schedules and Sample Steps are Flawed" integration | I was reading the paper [Common Diffusion Noise Schedules and Sample Steps are Flawed](https://arxiv.org/abs/2305.08891) and found it pretty interesting. It proposes a few simple changes that could be useful when integrated into existing schedulers in the `diffusers` library. Namely:
1. *"rescale the noise schedule to... | closed | completed | false | 23 | [
"New scheduler",
"Good second issue"
] | [] | 2023-05-17T22:31:03Z | 2024-11-21T15:36:27Z | 2024-11-21T15:36:26Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vvvm23 | 44,398,246 | MDQ6VXNlcjQ0Mzk4MjQ2 | User | false |
huggingface/diffusers | 1,715,056,650 | I_kwDOHa8MBc5mObAK | 3,480 | https://github.com/huggingface/diffusers/issues/3480 | https://api.github.com/repos/huggingface/diffusers/issues/3480 | Having a "scale" parameter for text encoder lora layers | **Is your feature request related to a problem? Please describe.**
No, not related to a problem
**Describe the solution you'd like**
A similar parameter to "scale" of unet lora layers, in case of text encoder lora layers
**Additional context**
This is an inference parameter that will determine the influence of... | closed | completed | false | 9 | [] | [] | 2023-05-18T06:13:10Z | 2023-06-07T13:36:05Z | 2023-06-07T13:35:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | abhijitpal1247 | 69,110,711 | MDQ6VXNlcjY5MTEwNzEx | User | false |
huggingface/diffusers | 1,715,214,947 | I_kwDOHa8MBc5mPBpj | 3,481 | https://github.com/huggingface/diffusers/issues/3481 | https://api.github.com/repos/huggingface/diffusers/issues/3481 | Stable Diffusion for text2image task with DDIMscheduler, fixed weight, prompts and random seed but different inference steps would generate dissimilar images | ### Describe the bug
I fix the random seeds and prompts, only change the number of inferences, and use DDIMscheduler with eta=0. But I find that the generated images are very different with different inference steps.
### Reproduction
```
import torch
from diffusers import StableDiffusionPipeline, EulerAnce... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-05-18T08:12:45Z | 2023-05-21T06:38:53Z | 2023-05-21T06:38:53Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhuyu-cs | 46,120,052 | MDQ6VXNlcjQ2MTIwMDUy | User | false |
huggingface/diffusers | 1,715,365,825 | I_kwDOHa8MBc5mPmfB | 3,482 | https://github.com/huggingface/diffusers/issues/3482 | https://api.github.com/repos/huggingface/diffusers/issues/3482 | finetuning Stable Diffusion Super-Resolution model | Thanks for your works! Diffusers truly help me a lot to learn about diffusion models.
I've been trying to finetune the Stable Diffusion Super-Resolution model on my custom datasets. And I take the train_text_to_image_lora.py as example, but I am stucked on the problem that
`torch.Size([2, 7, 128, 128]) torch.Size([2,... | closed | completed | false | 5 | [] | [] | 2023-05-18T10:06:05Z | 2023-06-01T12:24:21Z | 2023-06-01T12:24:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | R2Bb1T | 65,695,970 | MDQ6VXNlcjY1Njk1OTcw | User | false |
huggingface/diffusers | 1,715,671,224 | I_kwDOHa8MBc5mQxC4 | 3,483 | https://github.com/huggingface/diffusers/issues/3483 | https://api.github.com/repos/huggingface/diffusers/issues/3483 | Why does AttentionBlock perform softmax in float32? | Hi,
// *This question also applies to the newly refactored replacement for AttentionBlock, which sets [`upcast_softmax=True`](https://github.com/huggingface/diffusers/blob/88d269461ca9b5acfae3dedd732438266f526109/src/diffusers/models/unet_2d_blocks.py#L1173) for the replacements of AttentionBlock.*
I'm curious to... | closed | completed | false | 14 | [] | [] | 2023-05-18T13:47:58Z | 2023-07-08T16:59:08Z | 2023-07-08T16:59:08Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | cmdr2 | 844,287 | MDQ6VXNlcjg0NDI4Nw== | User | false |
huggingface/diffusers | 1,715,856,989 | I_kwDOHa8MBc5mReZd | 3,484 | https://github.com/huggingface/diffusers/issues/3484 | https://api.github.com/repos/huggingface/diffusers/issues/3484 | Training text_to_image on multi-GPU with Megatron-LM or DeepSpeed does not work | I have an EC2 instance with 4 GPUs and each GPU has 16GB memory.
By default, hugging face uses Data Parallelism and tries to fit one copy of the model in each GPU. Of course, 16GB is not enough and I ran into CUDA Out of Memory error.
I tried to use Tensor Parallelism and Pipeline Parallelism through **_Megatron-... | closed | completed | false | 6 | [] | [] | 2023-05-18T15:44:51Z | 2023-07-01T00:51:35Z | 2023-06-20T05:31:47Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zipliucc | 128,197,214 | U_kgDOB6QiXg | User | false |
huggingface/diffusers | 1,716,848,340 | I_kwDOHa8MBc5mVQbU | 3,487 | https://github.com/huggingface/diffusers/issues/3487 | https://api.github.com/repos/huggingface/diffusers/issues/3487 | New inpaint pipeline with parameter 'strength' generates poor results when using EulerAncestralDiscreteScheduler | ### Describe the bug
Hi, I recently found the update for inpainting with strength, it's really cool. [code](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py#L660)
But when using EulerAncestralDiscreteScheduler, just simply set strength ... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-05-19T08:23:24Z | 2023-05-25T00:22:47Z | 2023-05-20T03:16:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Magicalppj | 72,586,584 | MDQ6VXNlcjcyNTg2NTg0 | User | false |
huggingface/diffusers | 1,717,161,642 | I_kwDOHa8MBc5mWc6q | 3,489 | https://github.com/huggingface/diffusers/issues/3489 | https://api.github.com/repos/huggingface/diffusers/issues/3489 | lpw_stable_diffusion_onnx AttributeError: 'OnnxRuntimeModel' object has no attribute 'config' OnnxStableDiffusionImg2ImgPipeline | code:
device = "cuda"
model_id = "./sd_v15_onnx"
pipe = OnnxStableDiffusionImg2ImgPipeline.from_pretrained(
model_id,
custom_pipeline="lpw_stable_diffusion_onnx",
torch_dtype=torch.float16,
provider="CUDAExecutionProvider"
).to(device)
logs:
╭───────────────────────... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-05-19T12:21:15Z | 2023-07-04T15:03:09Z | 2023-07-04T15:03:09Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dushwe | 19,389,452 | MDQ6VXNlcjE5Mzg5NDUy | User | false |
huggingface/diffusers | 1,717,332,196 | I_kwDOHa8MBc5mXGjk | 3,491 | https://github.com/huggingface/diffusers/issues/3491 | https://api.github.com/repos/huggingface/diffusers/issues/3491 | OSError: fatal: unable to access 'https://huggingface.co/nageshmashette/ddpm-butterflies-128/': server certificate verification failed. CAfile: none CRLfile: none | after running 10 to 20 epochs i am getting error like:-
OSError: fatal: unable to access 'https://huggingface.co/nageshmashette/ddpm-butterflies-128/': server certificate verification failed. CAfile: none CRLfile: none
but while running 1 or 2 epochs its running fine and getting results. please help on this issue.
... | closed | completed | false | 4 | [] | [] | 2023-05-19T14:18:18Z | 2023-05-24T10:20:17Z | 2023-05-24T10:20:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | NageshMashette | 47,878,506 | MDQ6VXNlcjQ3ODc4NTA2 | User | false |
huggingface/diffusers | 1,718,288,810 | I_kwDOHa8MBc5mawGq | 3,493 | https://github.com/huggingface/diffusers/issues/3493 | https://api.github.com/repos/huggingface/diffusers/issues/3493 | AttnProcessor2_0 Significantly Slower | ### Describe the bug
The new AttnProcessor2_0 makes my finetuning code run at 3.8 it/s, whereas if I force AttnProcessor the speed is 5.8 it/s. This is without `torch.compile`. With `torch.compile` AttnProcessor2_0 is even worse at 3.5 it/s, but AttnProcessor remains at about 5.8 it/s.
Since training code can be... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-05-20T20:56:29Z | 2023-07-07T09:46:11Z | 2023-07-04T15:03:07Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fpgaminer | 1,585,817 | MDQ6VXNlcjE1ODU4MTc= | User | false |
huggingface/diffusers | 1,718,352,063 | I_kwDOHa8MBc5ma_i_ | 3,494 | https://github.com/huggingface/diffusers/issues/3494 | https://api.github.com/repos/huggingface/diffusers/issues/3494 | Multidiffusion creates corrupted images with different schedulers | Thanks for the great implement of `StableDiffusionPanoramaPipeline`.
However, I found that `StableDiffusionPanoramaPipeline` will create corrupted images when using some specific schedulers.
For example, if we use `DPMSolverMultistepScheduler`, a corrupted image will be generated.
```python
import torch
from dif... | closed | completed | false | 5 | [] | [] | 2023-05-21T02:49:55Z | 2023-05-23T03:26:36Z | 2023-05-23T03:26:36Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Isotr0py | 41,363,108 | MDQ6VXNlcjQxMzYzMTA4 | User | false |
huggingface/diffusers | 1,718,385,638 | I_kwDOHa8MBc5mbHvm | 3,495 | https://github.com/huggingface/diffusers/issues/3495 | https://api.github.com/repos/huggingface/diffusers/issues/3495 | stable_diffusion_tensorrt_txt2img not so fast | ### Describe the bug
I'm now benchmarking https://github.com/huggingface/diffusers/blob/main/examples/community/stable_diffusion_tensorrt_txt2img.py and got very bad results in comparison with https://github.com/NVIDIA/TensorRT/tree/main/demo/Diffusion
1.9 sec vs. 0.9 sec on A100 GPU
### Reproduction
Dockerfile
``... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-05-21T06:16:09Z | 2023-07-10T09:19:52Z | 2023-05-26T15:47:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexeigor | 2,498,713 | MDQ6VXNlcjI0OTg3MTM= | User | false |
huggingface/diffusers | 1,718,498,314 | I_kwDOHa8MBc5mbjQK | 3,497 | https://github.com/huggingface/diffusers/issues/3497 | https://api.github.com/repos/huggingface/diffusers/issues/3497 | 'StableDiffusionInpaintPipeline' has no attribute 'from_ckpt' | ### Describe the bug
Title is self-explanatory. There's no way to load inpaint model from safetensors using from_ckpt
I'm using the same approach as written in 'Load Safetensors' docs page:
```
pipeline = StableDiffusionPipeline.from_ckpt(
"https://huggingface.co/WarriorMama777/OrangeMixs/blob/main/Models/... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-05-21T13:08:37Z | 2023-06-01T12:21:18Z | 2023-06-01T12:21:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Pirog17000 | 3,022,852 | MDQ6VXNlcjMwMjI4NTI= | User | false |
huggingface/diffusers | 1,718,509,263 | I_kwDOHa8MBc5mbl7P | 3,498 | https://github.com/huggingface/diffusers/issues/3498 | https://api.github.com/repos/huggingface/diffusers/issues/3498 | I would like to use the foolhardi remacri upscaler | Hi, how do I do this? it seems kind of going overboard to use an entire pipeline for this when I have one made for other specific uses | closed | completed | false | 3 | [
"stale"
] | [] | 2023-05-21T13:43:39Z | 2023-06-28T15:03:14Z | 2023-06-28T15:03:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,361,095,249 | I_kwDOHa8MBc5RIKpR | 350 | https://github.com/huggingface/diffusers/issues/350 | https://api.github.com/repos/huggingface/diffusers/issues/350 | Typical CUDA Memory usage for `UNet2DConditionModel` | Hi,
Thanks for this wonderful and easy-to-use repository. I would like to know the typical VRAM usage of `UNet2DConditionModel` if I use the default parameters for one forward and backward pass. I am using a batch size of 1 with mixed precision but I am still getting
```
*** RuntimeError: CUDA out of memory. Tried... | closed | completed | false | 5 | [] | [] | 2022-09-04T09:35:26Z | 2022-09-08T06:15:26Z | 2022-09-08T06:15:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ovshake | 22,028,532 | MDQ6VXNlcjIyMDI4NTMy | User | false |
huggingface/diffusers | 1,718,614,481 | I_kwDOHa8MBc5mb_nR | 3,501 | https://github.com/huggingface/diffusers/issues/3501 | https://api.github.com/repos/huggingface/diffusers/issues/3501 | Textual Inversion Embedding & Compel Library | I'm trying to use Textual Inversion Embeddings in my current implementation of diffusers. After multiple tests I came to realize it did not work because I was using Compel to convert my prompt into embeddings. When removing Compel and passing the prompt directly to diffusers inference method it worked.
What is not cle... | closed | completed | false | 7 | [] | [] | 2023-05-21T19:28:06Z | 2023-06-05T11:56:54Z | 2023-06-05T11:56:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexisrolland | 13,064,696 | MDQ6VXNlcjEzMDY0Njk2 | User | false |
huggingface/diffusers | 1,719,047,351 | I_kwDOHa8MBc5mdpS3 | 3,502 | https://github.com/huggingface/diffusers/issues/3502 | https://api.github.com/repos/huggingface/diffusers/issues/3502 | I made the most sophisticated pipeline thus far but I broke some rules | My pipeline can divide an area into multiple areas and can use multicontrolnet at the same time. The only 2 things my pipeline can't do are upscaling and img2img (can't use loras when loading from ckpt either). here's the pipeline:
[Pipeline.txt](https://github.com/huggingface/diffusers/files/11528093/Pipeline.txt)
... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-05-22T07:11:33Z | 2023-06-30T15:02:58Z | 2023-06-30T15:02:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,361,115,497 | I_kwDOHa8MBc5RIPlp | 351 | https://github.com/huggingface/diffusers/issues/351 | https://api.github.com/repos/huggingface/diffusers/issues/351 | The model is not loaded directly into `CUDA` | ### Describe the bug
In the [Getting started with Diffusers] https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/diffusers_intro.ipynb) notebook available in the README. The model is not automatically loaded into `CUDA` with :
```python
image_pipe = DDPMPipeline.from_pretrained("google... | closed | completed | false | 2 | [
"bug"
] | [] | 2022-09-04T11:09:11Z | 2022-09-04T15:09:38Z | 2022-09-04T15:09:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Mustapha-AJEGHRIR | 66,799,406 | MDQ6VXNlcjY2Nzk5NDA2 | User | false |
huggingface/diffusers | 1,721,624,939 | I_kwDOHa8MBc5mnelr | 3,514 | https://github.com/huggingface/diffusers/issues/3514 | https://api.github.com/repos/huggingface/diffusers/issues/3514 | inpainting implement different from SD-webui | hi, recently I tried the diffuser inpaintpipeline and the SD-webui img2img inpaint, I found the result is different. As for the diffuser have change the part I have masked for unchanging. I wonder what the differences from these two method? | closed | completed | false | 5 | [
"stale"
] | [] | 2023-05-23T09:26:59Z | 2023-06-30T15:02:56Z | 2023-06-30T15:02:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AmberCheng | 18,375,990 | MDQ6VXNlcjE4Mzc1OTkw | User | false |
huggingface/diffusers | 1,721,800,375 | I_kwDOHa8MBc5moJa3 | 3,521 | https://github.com/huggingface/diffusers/issues/3521 | https://api.github.com/repos/huggingface/diffusers/issues/3521 | apply image on part of an image as noise? | Hi, I am not sure if it is possible, but is it possible to take an image X and apply to it on the bottom right image Y as noise? the idea is to add multiple elements to an image and then rerender the whole of it to get one image made from those multiple elements | closed | completed | false | 5 | [] | [] | 2023-05-23T10:53:31Z | 2023-05-25T06:52:29Z | 2023-05-25T06:52:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,723,005,286 | I_kwDOHa8MBc5msvlm | 3,537 | https://github.com/huggingface/diffusers/issues/3537 | https://api.github.com/repos/huggingface/diffusers/issues/3537 | Script for converting diffusers controlnet to original controlnet | **Is your feature request related to a problem? Please describe.**
Hi, I trained a controlnet with diffusers example and I want it in the original controlnet format.
**Describe the solution you'd like**
The same script as this one, but reversed, going from diffusers controlnet to original controlnet.
https://gith... | closed | completed | false | 7 | [] | [] | 2023-05-24T01:07:32Z | 2023-06-06T14:18:53Z | 2023-05-26T05:42:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | iszotic | 22,536,473 | MDQ6VXNlcjIyNTM2NDcz | User | false |
huggingface/diffusers | 1,723,053,536 | I_kwDOHa8MBc5ms7Xg | 3,538 | https://github.com/huggingface/diffusers/issues/3538 | https://api.github.com/repos/huggingface/diffusers/issues/3538 | Do we need to finetune controlnet when we merge sd with different lora | When we have trained a lora with sd and merge lora with sd to get a new base model, do we need to finetune controlnet with new base model to get better result? | closed | completed | false | 1 | [] | [] | 2023-05-24T02:01:38Z | 2023-06-06T04:54:18Z | 2023-06-06T04:54:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yftongbupt | 74,176,966 | MDQ6VXNlcjc0MTc2OTY2 | User | false |
huggingface/diffusers | 1,723,090,326 | I_kwDOHa8MBc5mtEWW | 3,539 | https://github.com/huggingface/diffusers/issues/3539 | https://api.github.com/repos/huggingface/diffusers/issues/3539 | Question about main_process_first | Hi, @sayakpaul @patrickvonplaten
sayakpaul
I am new to diffuser lib and accelerator , in the example/train_text_to_image.py https://github.com/huggingface/diffusers/blob/c13dbd5c3a53017d27de35ad77b8d57f04c8ec7c/examples/text_to_image/train_text_to_image.py#L700
I see the main_process_first is called, I am... | closed | completed | false | 2 | [] | [] | 2023-05-24T02:50:20Z | 2023-05-24T04:21:50Z | 2023-05-24T04:11:16Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | cnnAndBn | 69,999,981 | MDQ6VXNlcjY5OTk5OTgx | User | false |
huggingface/diffusers | 1,723,312,741 | I_kwDOHa8MBc5mt6pl | 3,540 | https://github.com/huggingface/diffusers/issues/3540 | https://api.github.com/repos/huggingface/diffusers/issues/3540 | 'prompt_embeds' and 'negative_prompt_embeds' do not work with TextToVideoZeroPipeline | ### Describe the bug
When passing `prompt_embeds` and `negative_prompt_embeds` to `TextToVideoZeroPipeline`, error is occured:
```
TypeError: TextToVideoZeroPipeline.__call__() got an unexpected keyword argument 'prompt_embeds'
```
### Reproduction
```
import torch
import imageio
from diffusers import TextTo... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-05-24T07:02:50Z | 2023-07-01T15:03:04Z | 2023-07-01T15:03:04Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | antonioglass | 71,437,038 | MDQ6VXNlcjcxNDM3MDM4 | User | false |
huggingface/diffusers | 1,723,543,012 | I_kwDOHa8MBc5muy3k | 3,544 | https://github.com/huggingface/diffusers/issues/3544 | https://api.github.com/repos/huggingface/diffusers/issues/3544 | Do latents need to be scaled by the scheduler first? | https://github.com/huggingface/diffusers/blob/c13dbd5c3a53017d27de35ad77b8d57f04c8ec7c/src/diffusers/pipelines/controlnet/pipeline_controlnet.py#L952-L963
I'm not sure if it's a bug but `latents` that is mentioned inside the if-block isn't scaled using `scheduler.scale_model_input()` unlike `latent_model_input`. | closed | completed | false | 2 | [] | [] | 2023-05-24T09:21:46Z | 2023-05-26T13:02:33Z | 2023-05-26T13:02:33Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | offchan42 | 15,215,732 | MDQ6VXNlcjE1MjE1NzMy | User | false |
huggingface/diffusers | 1,723,571,104 | I_kwDOHa8MBc5mu5ug | 3,545 | https://github.com/huggingface/diffusers/issues/3545 | https://api.github.com/repos/huggingface/diffusers/issues/3545 | Can't control cache directory | **What API design would you like to have changed or added to the library? Why?**
I would like to be able to control the cache directory for pipelines.
The HF Datasets library already has a good solution for thie: https://huggingface.co/docs/datasets/cache:
- The cache directory can be set through an environment ... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-05-24T09:34:53Z | 2023-07-01T15:03:03Z | 2023-07-01T15:03:03Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | spencerkohan | 221,514 | MDQ6VXNlcjIyMTUxNA== | User | false |
huggingface/diffusers | 1,723,637,590 | I_kwDOHa8MBc5mvJ9W | 3,548 | https://github.com/huggingface/diffusers/issues/3548 | https://api.github.com/repos/huggingface/diffusers/issues/3548 | Open Parti Prompts: The better way to evaluate diffusion models. | Quantitative measures like FID or CLIP-score say very little about image quality in practice, better measures are needed!
Introducing Open Parti Prompts, a community-driven approach to compare models with human preference ❤️
The leaderboard is live and can be accessed here:
👉 https://huggingface.co/spaces/Ope... | closed | completed | false | 4 | [
"stale"
] | [] | 2023-05-24T10:09:12Z | 2023-07-22T15:03:11Z | 2023-07-22T15:03:11Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | patrickvonplaten | 23,423,619 | MDQ6VXNlcjIzNDIzNjE5 | User | false |
huggingface/diffusers | 1,724,465,538 | I_kwDOHa8MBc5myUGC | 3,551 | https://github.com/huggingface/diffusers/issues/3551 | https://api.github.com/repos/huggingface/diffusers/issues/3551 | Using LoRA will deactivate xFormers, even if it is enabled | ### Describe the bug
Discussed in https://github.com/huggingface/diffusers/pull/3437#issuecomment-1559822652 . It appears that memory usage increases significantly when using LoRA, even in environments using xFormers. I investigated the cause using [this script](https://gist.github.com/takuma104/e2139bda7f74cd977350e1... | closed | completed | false | 9 | [
"bug"
] | [] | 2023-05-24T17:34:56Z | 2023-05-26T12:03:50Z | 2023-05-26T12:03:49Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | takuma104 | 10,776 | MDQ6VXNlcjEwNzc2 | User | false |
huggingface/diffusers | 1,725,059,170 | I_kwDOHa8MBc5m0lBi | 3,553 | https://github.com/huggingface/diffusers/issues/3553 | https://api.github.com/repos/huggingface/diffusers/issues/3553 | StableDiffusionPipeline callback problem | ### Describe the bug
I want to display the results of the image generation process in diffusers, so I called the callback function. However, when I convert the received tensor variable to an image in the callback, I found that the image has a size of only 64x64 pixels, and it looks very strange. Is there a problem wit... | closed | completed | false | 3 | [
"bug",
"stale"
] | [] | 2023-05-25T04:18:55Z | 2023-07-02T15:03:05Z | 2023-07-02T15:03:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hckj588ku | 132,969,506 | U_kgDOB-z0Ig | User | false |
huggingface/diffusers | 1,726,821,086 | I_kwDOHa8MBc5m7TLe | 3,564 | https://github.com/huggingface/diffusers/issues/3564 | https://api.github.com/repos/huggingface/diffusers/issues/3564 | [Pipelines] Support for Uni-ControlNet | ### Model/Pipeline/Scheduler description
From the official repository, [Uni-ControlNet](https://github.com/ShihaoZhaoZSH/Uni-ControlNet) (All-in-One Control to Text-to-Image Diffusion Models) categorizes various conditions into two distinct groups: local conditions and global conditions. Accordingly, two additional ... | closed | completed | false | 10 | [
"New pipeline/model",
"community-examples"
] | [] | 2023-05-26T03:00:41Z | 2024-11-22T01:38:16Z | 2024-11-22T01:29:57Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wfng92 | 43,742,196 | MDQ6VXNlcjQzNzQyMTk2 | User | false |
huggingface/diffusers | 1,726,839,326 | I_kwDOHa8MBc5m7Xoe | 3,565 | https://github.com/huggingface/diffusers/issues/3565 | https://api.github.com/repos/huggingface/diffusers/issues/3565 | [examples/text_to_image] cuda out of memory, though I followed the instructions of train_text_to_image.py | ### Describe the bug
when I ran the script: examples/text_to_image/text_to_image.py, using the follwing command:
export MODEL_NAME="CompVis/stable-diffusion-v1-4"
export dataset_name="lambdalabs/pokemon-blip-captions"
python train_text_to_image.py \
--pretrained_model_name_or_path=$MODEL_NAME \
--dataset_... | closed | completed | false | 16 | [
"bug",
"stale"
] | [] | 2023-05-26T03:22:02Z | 2023-07-13T14:06:01Z | 2023-07-04T15:03:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | N1cekiko | 38,829,641 | MDQ6VXNlcjM4ODI5NjQx | User | false |
huggingface/diffusers | 1,726,881,203 | I_kwDOHa8MBc5m7h2z | 3,566 | https://github.com/huggingface/diffusers/issues/3566 | https://api.github.com/repos/huggingface/diffusers/issues/3566 | Unexpected results with controlnet_reference community pipeline | ### Describe the bug
I can't seem to get expected results from the new controlnet_reference pipeline.
I get subpar results with style_fidelity=0.0, but with anything greater than 0.0 I get weird/collapsed images.
I believe there are a few issues and have tried many things, but to keep things simple below is an tes... | closed | completed | false | 14 | [
"bug",
"stale"
] | [] | 2023-05-26T04:20:47Z | 2023-07-04T15:03:04Z | 2023-07-04T15:03:03Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | reimager | 20,918,094 | MDQ6VXNlcjIwOTE4MDk0 | User | false |
huggingface/diffusers | 1,726,940,290 | I_kwDOHa8MBc5m7wSC | 3,567 | https://github.com/huggingface/diffusers/issues/3567 | https://api.github.com/repos/huggingface/diffusers/issues/3567 | Using a custom pipeline with from_ckpt. | Hey, it seems like loading custom pipelines with from_ckpt is currently not implemented. Is there a workaround for this and is this feature planned for the future? Thanks in advance! | closed | completed | false | 9 | [
"stale"
] | [] | 2023-05-26T05:39:38Z | 2024-09-29T00:01:14Z | 2023-07-10T15:03:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sertdfyguhi | 71,869,572 | MDQ6VXNlcjcxODY5NTcy | User | false |
huggingface/diffusers | 1,727,147,796 | I_kwDOHa8MBc5m8i8U | 3,569 | https://github.com/huggingface/diffusers/issues/3569 | https://api.github.com/repos/huggingface/diffusers/issues/3569 | [New Pipeline]: Uni-ControlNet: All-in-One Control to Text-to-Image Diffusion Models | Github: https://github.com/ShihaoZhaoZSH/Uni-ControlNet
Project Page: https://shihaozhaozsh.github.io/unicontrolnet/
Paper: https://arxiv.org/abs/2305.16322 | closed | completed | false | 4 | [] | [] | 2023-05-26T08:24:47Z | 2023-06-01T16:43:48Z | 2023-06-01T16:43:48Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kadirnar | 36,204,372 | MDQ6VXNlcjM2MjA0Mzcy | User | false |
huggingface/diffusers | 1,727,207,585 | I_kwDOHa8MBc5m8xih | 3,570 | https://github.com/huggingface/diffusers/issues/3570 | https://api.github.com/repos/huggingface/diffusers/issues/3570 | cannot disable downloading when loading ckpt | ### Describe the bug
I want to load a ckpt without Internet connection(to github, specifically), but loading the ckpt triggers downloading the config file from stable diffusion repo on github, even if it has been downloaded before, and local_files_only set to True.
### Reproduction
#no Internet connection
from di... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-05-26T09:05:16Z | 2023-07-04T15:03:02Z | 2023-07-04T15:03:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | reveperdu | 109,431,485 | U_kgDOBoXKvQ | User | false |
huggingface/diffusers | 1,727,727,056 | I_kwDOHa8MBc5m-wXQ | 3,573 | https://github.com/huggingface/diffusers/issues/3573 | https://api.github.com/repos/huggingface/diffusers/issues/3573 | "class_labels should be provided" in local upscaler pipeline | Hi, I'm getting this:
│ /usr/local/lib/python3.8/site-packages/torch/utils/_contextlib.py:115 in decorate_context │
│ │
│ 112 │ @functools.wraps(func) ... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-05-26T14:20:17Z | 2023-07-07T15:03:40Z | 2023-07-07T15:03:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,727,935,917 | I_kwDOHa8MBc5m_jWt | 3,574 | https://github.com/huggingface/diffusers/issues/3574 | https://api.github.com/repos/huggingface/diffusers/issues/3574 | from_ckpt not working properly on windows | ### Describe the bug
When attempting to use the `StableDiffusionPipeline.from_ckpt()` with HF checkpoint link on Windows OS, a `HFValidationError` exception is raised. This issue stems from how the Path is processed in the code [here](https://github.com/huggingface/diffusers/blob/67cf0445ef48b1f913b90ce0025ac0c75673e3... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-05-26T16:43:26Z | 2023-06-05T10:55:38Z | 2023-06-05T10:55:38Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | LyubimovVladislav | 43,727,166 | MDQ6VXNlcjQzNzI3MTY2 | User | false |
huggingface/diffusers | 1,728,074,626 | I_kwDOHa8MBc5nAFOC | 3,575 | https://github.com/huggingface/diffusers/issues/3575 | https://api.github.com/repos/huggingface/diffusers/issues/3575 | What sampler does the official Runway ML demo use? | What sampler does the official Runway ML demo use? https://runwayml-stable-diffusion-v1-5.hf.space/
I get much more creative results using this instead of A111 locally and I want to replicate it | closed | completed | false | 2 | [
"stale"
] | [] | 2023-05-26T18:37:55Z | 2023-07-04T15:03:00Z | 2023-07-04T15:03:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | phokphokphok | 134,711,075 | U_kgDOCAeHIw | User | false |
huggingface/diffusers | 1,728,162,699 | I_kwDOHa8MBc5nAauL | 3,576 | https://github.com/huggingface/diffusers/issues/3576 | https://api.github.com/repos/huggingface/diffusers/issues/3576 | TypeError when using XFormers with Textual Inversion | ### Describe the bug
I'm encountering an issue when trying to use xformers with textual inversion. Using textual inversion works as expected with the following setup:
pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16).to("cuda")
pipe.load_textual_inversion("/content/8387", token="cha... | closed | completed | false | 5 | [
"bug"
] | [] | 2023-05-26T19:57:26Z | 2023-05-30T10:26:25Z | 2023-05-30T10:26:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | realimposter | 51,212,151 | MDQ6VXNlcjUxMjEyMTUx | User | false |
huggingface/diffusers | 1,728,499,073 | I_kwDOHa8MBc5nBs2B | 3,578 | https://github.com/huggingface/diffusers/issues/3578 | https://api.github.com/repos/huggingface/diffusers/issues/3578 | Fail to load checkpoint for AutoencoderKL | ### Describe the bug
```
File "/home/xx/.local/lib/python3.8/site-packages/diffusers/models/modeling_utils.py", line 102, in load_state_dict
return safetensors.torch.load_file(checkpoint_file, device="cpu")
File "/home/xx/.local/lib/python3.8/site-packages/safetensors/torch.py", line 261, in load_file
... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-05-27T06:50:42Z | 2023-07-04T15:02:59Z | 2023-07-04T15:02:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TiankaiHang | 52,823,230 | MDQ6VXNlcjUyODIzMjMw | User | false |
huggingface/diffusers | 1,728,576,957 | I_kwDOHa8MBc5nB_29 | 3,579 | https://github.com/huggingface/diffusers/issues/3579 | https://api.github.com/repos/huggingface/diffusers/issues/3579 | callback preview image noise | ```py
import torch
import torchvision
from PIL import Image
from diffusers import EulerDiscreteScheduler, StableDiffusionPipeline
device = "cuda" if torch.cuda.is_available() else "cpu"
pipe = StableDiffusionPipeline.from_pretrained(
"prompthero/openjourney-v4", torch_dtype=torch.float16, safety_checker=... | closed | completed | false | 10 | [
"stale"
] | [] | 2023-05-27T09:10:23Z | 2024-11-11T02:13:10Z | 2023-07-04T15:02:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | enescakircali | 31,443,145 | MDQ6VXNlcjMxNDQzMTQ1 | User | false |
huggingface/diffusers | 1,361,673,427 | I_kwDOHa8MBc5RKXzT | 358 | https://github.com/huggingface/diffusers/issues/358 | https://api.github.com/repos/huggingface/diffusers/issues/358 | Apple MPS error in unet_2d_condition.py | ### Describe the bug
When you use an LMSDiscreteScheduler on an Apple Silicon machine, you'll get the following error:
Cannot convert a MPS Tensor to float64 dtype as the MPS framework doesn't support float64. Please use float32 instead
The offending line is 134 in unet_2d_condition.py.
The current code is:
ti... | closed | completed | false | 14 | [
"bug"
] | [
"pcuenca"
] | 2022-09-05T09:42:51Z | 2022-09-13T09:23:22Z | 2022-09-13T09:23:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FahimF | 181,110 | MDQ6VXNlcjE4MTExMA== | User | false |
huggingface/diffusers | 1,728,654,797 | I_kwDOHa8MBc5nCS3N | 3,580 | https://github.com/huggingface/diffusers/issues/3580 | https://api.github.com/repos/huggingface/diffusers/issues/3580 | TypeError when using XFormers (diffusers version: 0.17.0.dev0) | ### Describe the bug
I was generating images using Hugging face Diffusers Stable Diffusion pipeline on Google colab, I got the following bugs.
TypeError: XFormersAttnProcessor.call() got an unexpected keyword argument 'temb'
![image](https://github.com/huggingface/diffusers/assets/91832589/574853a6-b9cd-46a2-bec... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-05-27T11:10:41Z | 2023-06-01T12:40:55Z | 2023-06-01T12:40:55Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | deepgoyal19 | 91,832,589 | U_kgDOBXlBDQ | User | false |
huggingface/diffusers | 1,728,750,653 | I_kwDOHa8MBc5nCqQ9 | 3,582 | https://github.com/huggingface/diffusers/issues/3582 | https://api.github.com/repos/huggingface/diffusers/issues/3582 | StableDiffusionControlNetInpaintPipeline `strength=1` doesn't work | ### Describe the bug
cc @williamberman
Lets say you have an image with a subject and a background, you put a mask over the background to repaint it and set the `strength=1`, so in theory the background should be completely ignored. But this isn't what happens, the background carries some weight in the regeneration.... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-05-27T14:13:00Z | 2023-07-17T22:19:52Z | 2023-07-17T22:19:52Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | undernf | 134,790,842 | U_kgDOCAi-ug | User | false |
huggingface/diffusers | 1,728,775,919 | I_kwDOHa8MBc5nCwbv | 3,583 | https://github.com/huggingface/diffusers/issues/3583 | https://api.github.com/repos/huggingface/diffusers/issues/3583 | [Pipelines] Support for Mixture of Diffusers | Github: https://github.com/albarji/mixture-of-diffusers
Paper: https://arxiv.org/abs/2302.02412
HF Demo: https://huggingface.co/spaces/albarji/mixture-of-diffusers | closed | completed | false | 3 | [] | [] | 2023-05-27T15:03:35Z | 2023-06-13T12:15:01Z | 2023-06-13T12:15:01Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kadirnar | 36,204,372 | MDQ6VXNlcjM2MjA0Mzcy | User | false |
huggingface/diffusers | 1,728,782,774 | I_kwDOHa8MBc5nCyG2 | 3,584 | https://github.com/huggingface/diffusers/issues/3584 | https://api.github.com/repos/huggingface/diffusers/issues/3584 | DreamBooth Diffusers tutorial failed | ### Describe the bug
Issue with the DreamBooth Diffusers tutorial. After executing the accelerate launch command, the script breaks with the following runtimeError: `RuntimeError: Given groups=1, weight of size [320, 4, 3, 3], expected input[1, 3, 512, 512] to have 4 channels, but got 3 channels instead`.
I've not ... | closed | completed | false | 19 | [
"bug"
] | [
"sayakpaul"
] | 2023-05-27T15:19:24Z | 2023-06-11T11:35:11Z | 2023-06-06T16:42:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kiriamcf | 42,093,327 | MDQ6VXNlcjQyMDkzMzI3 | User | false |
huggingface/diffusers | 1,728,842,722 | I_kwDOHa8MBc5nDAvi | 3,585 | https://github.com/huggingface/diffusers/issues/3585 | https://api.github.com/repos/huggingface/diffusers/issues/3585 | Use a model that synthesizes lora into stable-diffusion-v1-5 failed | ### Describe the bug
I trained a lora model with train_text_to_image_lora.py. And I try to merge it to stable-diffusion-v1.5 with following code.
When I try to use my merged model, the following error occurred.
**KeyError: 'down_blocks.0.attentions.0.transformer_blocks.0.attn1.processor.to_q_lora.down.weight'**
##... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-05-27T17:12:37Z | 2023-06-01T12:06:24Z | 2023-06-01T12:06:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | fangshuoz | 75,928,120 | MDQ6VXNlcjc1OTI4MTIw | User | false |
huggingface/diffusers | 1,729,888,110 | I_kwDOHa8MBc5nG_9u | 3,591 | https://github.com/huggingface/diffusers/issues/3591 | https://api.github.com/repos/huggingface/diffusers/issues/3591 | ControlNet's loss curve is confusing. | ### Describe the bug
I trained the ControlNet as instructed by the README documentation, but got a confusing loss curve trend. I want to make sure that this curve trend is normal.
### Reproduction
**The code I use for training is as below:**
export MODEL_DIR="runwayml/stable-diffusion-v1-5"
export OUTPUT_DIR=".... | closed | completed | false | 3 | [
"bug"
] | [] | 2023-05-29T02:23:43Z | 2023-06-03T01:43:03Z | 2023-06-01T12:02:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | WeiChunyu-star | 54,472,778 | MDQ6VXNlcjU0NDcyNzc4 | User | false |
huggingface/diffusers | 1,729,982,174 | I_kwDOHa8MBc5nHW7e | 3,593 | https://github.com/huggingface/diffusers/issues/3593 | https://api.github.com/repos/huggingface/diffusers/issues/3593 | [LoRA] enabling xformers with LoRA fails during the second pipeline call (PT 2.0) | I am on the following:
```bash
- `diffusers` version: 0.17.0.dev0
- Platform: Linux-5.15.107+-x86_64-with-glibc2.31
- Python version: 3.10.11
- PyTorch version (GPU?): 2.0.1+cu118 (True)
- Huggingface_hub version: 0.14.1
- Transformers version: 4.29.2
- Accelerate version: 0.19.0
- xFormers version: 0.0.20
... | closed | completed | false | 5 | [] | [] | 2023-05-29T03:58:45Z | 2023-06-01T12:00:18Z | 2023-06-01T12:00:18Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,730,606,236 | I_kwDOHa8MBc5nJvSc | 3,595 | https://github.com/huggingface/diffusers/issues/3595 | https://api.github.com/repos/huggingface/diffusers/issues/3595 | from_ckpt not working | ### Describe the bug
I tried to use from_ckpt with a huggingface link from #3466 but i got a log that says "global_step key not found in model" and it downloaded the diffusers version of the model.
### Reproduction
import torch
from torch import autocast
from diffusers import StableDiffusionPipeline, EulerDis... | closed | completed | false | 6 | [
"bug",
"stale"
] | [] | 2023-05-29T11:34:26Z | 2023-07-04T17:49:54Z | 2023-07-04T17:49:54Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | aycaecemgul | 50,622,243 | MDQ6VXNlcjUwNjIyMjQz | User | false |
huggingface/diffusers | 1,730,679,734 | I_kwDOHa8MBc5nKBO2 | 3,596 | https://github.com/huggingface/diffusers/issues/3596 | https://api.github.com/repos/huggingface/diffusers/issues/3596 | LMS Comparison | ### Describe the bug
The Invoke community has noted that the LMS sampler performs extremely poorly following our full migration to Diffusers.
Per @patrickvonplaten - Sharing original LMS code
```"""SAMPLING ONLY."""
import torch
import numpy as np
from tqdm import tqdm
from functools import partial
from l... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-05-29T12:27:56Z | 2023-06-01T12:12:19Z | 2023-06-01T12:12:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hipsterusername | 31,807,370 | MDQ6VXNlcjMxODA3Mzcw | User | false |
huggingface/diffusers | 1,730,690,471 | I_kwDOHa8MBc5nKD2n | 3,597 | https://github.com/huggingface/diffusers/issues/3597 | https://api.github.com/repos/huggingface/diffusers/issues/3597 | lpw_stable_diffusion_onnx doesn't work | ### Describe the bug
lpw_stable_diffusion_onnx doesn't work
prompt ="((ghibli style)), san (mononoke hime), 1girl, armlet, bangs, black hair, black undershirt, circlet, earrings, facepaint, floating hair, forest, fur cape, green eyes, jewelry, looking at viewer, nature, outdoors, parted bangs, short hair, sleevel... | closed | completed | false | 2 | [
"bug",
"stale"
] | [] | 2023-05-29T12:36:48Z | 2023-07-06T15:03:25Z | 2023-07-06T15:03:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dushwe | 19,389,452 | MDQ6VXNlcjE5Mzg5NDUy | User | false |
huggingface/diffusers | 1,730,700,289 | I_kwDOHa8MBc5nKGQB | 3,598 | https://github.com/huggingface/diffusers/issues/3598 | https://api.github.com/repos/huggingface/diffusers/issues/3598 | Artefacts with `LMSDiscreteScheduler` at certain step counts | ### Describe the bug
`LMSDiscreteScheduler` often produces images with artefacts.
This has been an issue for some time (several weeks at least), and I'm surprised there aren't existing reports. Perhaps the reason is that `LMSDiscrete` is not a particularly popular scheduler.
Here are my notes:
- as step count i... | closed | completed | false | 8 | [
"bug",
"stale"
] | [] | 2023-05-29T12:45:00Z | 2023-07-06T15:03:23Z | 2023-07-06T15:03:23Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | psychedelicious | 4,822,129 | MDQ6VXNlcjQ4MjIxMjk= | User | false |
huggingface/diffusers | 1,730,829,740 | I_kwDOHa8MBc5nKl2s | 3,599 | https://github.com/huggingface/diffusers/issues/3599 | https://api.github.com/repos/huggingface/diffusers/issues/3599 | Adding Daam | **Is your feature request related to a problem? Please describe.**
This is a feature request. I think it'll be nice to add in daam from [here](https://github.com/castorini/daam) to diffusers. This is a way to get the attention maps from each token to see their contributions for generation like this
.
Would you have the plan to implement for TensorRT? | closed | completed | false | 7 | [
"stale"
] | [] | 2022-06-28T06:33:43Z | 2022-10-31T15:03:43Z | 2022-10-31T15:03:43Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | UdonDa | 25,411,643 | MDQ6VXNlcjI1NDExNjQz | User | false |
huggingface/diffusers | 1,730,834,651 | I_kwDOHa8MBc5nKnDb | 3,600 | https://github.com/huggingface/diffusers/issues/3600 | https://api.github.com/repos/huggingface/diffusers/issues/3600 | how to merge inpainting diffusion model using checkpoint_merger.py | 当我使用下面方法进行merger时 报错了：
merged_pipe = pipe.merge([ "runwayml/stable-diffusion-inpainting", "andite/anything-v4.0", "runwayml/stable-diffusion-v1-5"], local_files_only=False, safety_checker=None, force = True, interp = "add_diff", alpha = 1).to("cuda", torch_dtype=torch.float16)
error：
│ 285 │ def add_differenc... | closed | completed | false | 3 | [
"stale"
] | [] | 2023-05-29T14:11:08Z | 2023-07-11T15:03:50Z | 2023-07-11T15:03:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yanchaoguo | 11,827,802 | MDQ6VXNlcjExODI3ODAy | User | false |
huggingface/diffusers | 1,730,946,193 | I_kwDOHa8MBc5nLCSR | 3,601 | https://github.com/huggingface/diffusers/issues/3601 | https://api.github.com/repos/huggingface/diffusers/issues/3601 | how do I get dtype from inside the pipeline (custom pipelines) | Hi, pretty much the title. Let's say I wanted to print the dtype inside the init function of a custom pipeline. How do I do that? | closed | completed | false | 2 | [] | [] | 2023-05-29T15:26:05Z | 2023-05-30T10:55:30Z | 2023-05-30T10:55:29Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alexblattner | 15,870,094 | MDQ6VXNlcjE1ODcwMDk0 | User | false |
huggingface/diffusers | 1,730,961,333 | I_kwDOHa8MBc5nLF-1 | 3,602 | https://github.com/huggingface/diffusers/issues/3602 | https://api.github.com/repos/huggingface/diffusers/issues/3602 | What is the default for VAE option? | If "VAE" is not specified for "Stable Diffusion," what is the default applied? | closed | completed | false | 4 | [] | [] | 2023-05-29T15:42:19Z | 2023-06-08T10:30:27Z | 2023-06-08T10:30:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Michi-123 | 26,673,104 | MDQ6VXNlcjI2NjczMTA0 | User | false |
huggingface/diffusers | 1,731,115,128 | I_kwDOHa8MBc5nLrh4 | 3,603 | https://github.com/huggingface/diffusers/issues/3603 | https://api.github.com/repos/huggingface/diffusers/issues/3603 | Torch.compile dynamic input support | `torch.compile` support in diffusers is interesting in theory, but in reality users tend to change settings, use extensions and swap stuff out, forcing lengthy recompiles whenever that happens. Dynamic input support would help mitigate this.
I am testing stable diffusion with `pipe.unet = torch.compile(pipe.unet, dy... | closed | completed | false | 7 | [
"stale"
] | [] | 2023-05-29T18:30:14Z | 2023-10-24T15:16:14Z | 2023-07-07T15:03:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AlphaAtlas | 46,462,706 | MDQ6VXNlcjQ2NDYyNzA2 | User | false |
huggingface/diffusers | 1,731,385,462 | I_kwDOHa8MBc5nMth2 | 3,604 | https://github.com/huggingface/diffusers/issues/3604 | https://api.github.com/repos/huggingface/diffusers/issues/3604 | train loss doesn't decrease while finetuning using training script | Hello, first of all thank you so much for all your contribution.
I am using the newly uploaded [train_text_to_image.py](https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image.py) to fine-tune the text-to-image stable diffusion model on my own dataset using 1 A100 GPU.
Howeve... | closed | completed | false | 6 | [
"stale"
] | [] | 2023-05-30T01:38:40Z | 2023-08-30T13:53:00Z | 2023-07-08T15:03:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | emily-swatchon | 100,738,598 | U_kgDOBgEmJg | User | false |
huggingface/diffusers | 1,732,088,670 | I_kwDOHa8MBc5nPZNe | 3,610 | https://github.com/huggingface/diffusers/issues/3610 | https://api.github.com/repos/huggingface/diffusers/issues/3610 | Streaming version | Looking for a way to optimize training on a large dataset. I would like to be able to use load_dataset(streaming=True...)
From what I see, current version doesn't support it. How hard is to adapt the current code? Any other suggestions to deal with the problem?
| closed | completed | false | 4 | [
"stale"
] | [] | 2023-05-30T11:41:42Z | 2023-07-07T15:03:36Z | 2023-07-07T15:03:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xenia-kra | 81,739,495 | MDQ6VXNlcjgxNzM5NDk1 | User | false |
huggingface/diffusers | 1,732,527,167 | I_kwDOHa8MBc5nREQ_ | 3,611 | https://github.com/huggingface/diffusers/issues/3611 | https://api.github.com/repos/huggingface/diffusers/issues/3611 | Readme image is broken | Looks like the readme image is somehow broken.
<img width="422" alt="image" src="https://github.com/huggingface/diffusers/assets/44067394/e88be2ea-e4c2-4379-8a35-45ce1fe89441">
| closed | completed | false | 1 | [] | [] | 2023-05-30T15:42:36Z | 2023-05-30T22:18:56Z | 2023-05-30T22:18:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | omerXfaruq | 44,067,394 | MDQ6VXNlcjQ0MDY3Mzk0 | User | false |
huggingface/diffusers | 1,732,530,504 | I_kwDOHa8MBc5nRFFI | 3,612 | https://github.com/huggingface/diffusers/issues/3612 | https://api.github.com/repos/huggingface/diffusers/issues/3612 | How should I create a ControlNet without a model using diffusers and specify its preprocessor as reference_adain+attn | **Is your feature request related to a problem? Please describe.**
I need to implement the same function in diffusers as ControlNet in SD WebUI with reference_adain+attn preprocessor specified, but I can't find the available method.
**Describe the solution you'd like**
I can create a ControlNet without a model and... | closed | completed | false | 28 | [] | [] | 2023-05-30T15:44:48Z | 2024-07-14T16:23:44Z | 2023-06-13T15:46:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vvlookman | 61,679,792 | MDQ6VXNlcjYxNjc5Nzky | User | false |
huggingface/diffusers | 1,732,545,053 | I_kwDOHa8MBc5nRIod | 3,613 | https://github.com/huggingface/diffusers/issues/3613 | https://api.github.com/repos/huggingface/diffusers/issues/3613 | train dreambooth out of gpu memory with torch2.0 | ### Describe the bug
if torch version is 2.0 and not use --use_8bit_adam, will out of memory.
but use --use_8bit_adam, result is wrong.
when i change torch version to 1.13, everything is ok.
### Reproduction
below is shell command
`
export MODEL_NAME="CompVis/stable-diffusion-v1-4"
export INSTANCE_DIR="./do... | closed | completed | false | 16 | [
"bug",
"stale"
] | [] | 2023-05-30T15:54:18Z | 2023-08-19T15:03:34Z | 2023-08-19T15:03:34Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | conansherry | 2,148,224 | MDQ6VXNlcjIxNDgyMjQ= | User | false |
huggingface/diffusers | 1,733,079,888 | I_kwDOHa8MBc5nTLNQ | 3,618 | https://github.com/huggingface/diffusers/issues/3618 | https://api.github.com/repos/huggingface/diffusers/issues/3618 | Stable Diffusion Pipeline duplicates its process when run from embedded interpreter on Mac | ### Describe the bug
Hi there,
I am having a super bizarre bug. When using the basic four-line Stable Diffusion Pipeline example on Mac from an embedded python interpreter inside a C++ program (code below), whenever I get to the inference call (pipe()), it seems to duplicate the C++ process multiple times. Each o... | closed | not_planned | false | 1 | [
"bug"
] | [] | 2023-05-30T22:04:09Z | 2023-05-31T04:16:54Z | 2023-05-31T04:16:54Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jonahclarsen | 97,376,233 | U_kgDOBc3X6Q | User | false |
huggingface/diffusers | 1,733,160,164 | I_kwDOHa8MBc5nTezk | 3,619 | https://github.com/huggingface/diffusers/issues/3619 | https://api.github.com/repos/huggingface/diffusers/issues/3619 | Text2Video-Zero CrossFrameAttention doesn't work | ### Describe the bug
I try to follow the instructions, but finding the result is the same with or without the following two lines of code.
```
pipe.unet.set_attn_processor(CrossFrameAttnProcessor(batch_size=2))
pipe.controlnet.set_attn_processor(CrossFrameAttnProcessor(batch_size=2))
```
### Reproduction
from ... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-05-30T23:34:40Z | 2023-06-08T17:28:14Z | 2023-06-08T17:28:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | minzhang-1 | 34,903,683 | MDQ6VXNlcjM0OTAzNjgz | User | false |
huggingface/diffusers | 1,733,290,353 | I_kwDOHa8MBc5nT-lx | 3,620 | https://github.com/huggingface/diffusers/issues/3620 | https://api.github.com/repos/huggingface/diffusers/issues/3620 | [LoRA] Discussions on ensuring robust LoRA support in Diffusers | For the last few months, we have been collaborating with our contributors to ensure we support LoRA effectively and efficiently from Diffusers:
**1. Training support**
✅ [DreamBooth](https://github.com/huggingface/diffusers/tree/main/examples/dreambooth) (letting users perform LoRA fine-tuning of both UNet and ... | closed | completed | false | 17 | [] | [] | 2023-05-31T02:36:36Z | 2023-07-12T02:43:45Z | 2023-07-11T06:17:33Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,733,296,922 | I_kwDOHa8MBc5nUAMa | 3,621 | https://github.com/huggingface/diffusers/issues/3621 | https://api.github.com/repos/huggingface/diffusers/issues/3621 | [LoRA] Repurcussions of using monkey-patching for text encoder LoRA | We use monkey-patching to support LoRA fine-tuning of the text encoder and loading of the LoRA params corresponding to the text encoder. While it helps us in preventing additional dependencies like `peft`<sup>*</sup>, it comes with baggage.
As pointed out by @rvorias in https://github.com/huggingface/diffusers/pull... | closed | completed | false | 9 | [] | [] | 2023-05-31T02:46:31Z | 2023-06-20T11:09:47Z | 2023-06-20T11:09:46Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sayakpaul | 22,957,388 | MDQ6VXNlcjIyOTU3Mzg4 | User | false |
huggingface/diffusers | 1,733,324,728 | I_kwDOHa8MBc5nUG-4 | 3,622 | https://github.com/huggingface/diffusers/issues/3622 | https://api.github.com/repos/huggingface/diffusers/issues/3622 | Tutorial 'Effective and efficient diffusion' gets an import error. | ### Describe the bug
Hi,
Thanks for the great job!
There is a minor error in the tutorial [https://huggingface.co/docs/diffusers/stable_diffusion](https://huggingface.co/docs/diffusers/stable_diffusion), Tutorial 'Effective and efficient diffusion', the section of 'Memory'
![image](https://github.com/huggingf... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-05-31T03:27:49Z | 2023-06-01T11:26:50Z | 2023-06-01T11:26:50Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lhaippp | 22,257,640 | MDQ6VXNlcjIyMjU3NjQw | User | false |
huggingface/diffusers | 1,733,367,603 | I_kwDOHa8MBc5nURcz | 3,623 | https://github.com/huggingface/diffusers/issues/3623 | https://api.github.com/repos/huggingface/diffusers/issues/3623 | Unable to load larger models | ### Describe the bug
Running
```
StableDiffusionPipeline.from_ckpt("path_to_safetensorfile")
```
or the
```
./scripts/convert_original_stable_diffusion_to_diffusers.py
```
Causes some weights to be removed from the models. The model does not work as well as it should. The results from same settings (se... | closed | completed | false | 6 | [
"bug"
] | [] | 2023-05-31T04:30:22Z | 2023-06-28T20:31:07Z | 2023-06-01T02:32:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | JaLnYn | 33,004,730 | MDQ6VXNlcjMzMDA0NzMw | User | false |
huggingface/diffusers | 1,733,622,813 | I_kwDOHa8MBc5nVPwd | 3,624 | https://github.com/huggingface/diffusers/issues/3624 | https://api.github.com/repos/huggingface/diffusers/issues/3624 | TypeError: getattr(): attribute name must be string | ### Describe the bug
]
`text_config_dict` is provided which will be used to initialize `CLIPTextConfig`. The value `text_config["id2label"]` will be overriden.
The config attributes {'clip_sample': False} were passed to PNDMScheduler, but are not expected and will be ignored. Please verify your scheduler_config.json... | closed | completed | false | 1 | [
"bug"
] | [] | 2023-05-31T08:08:45Z | 2024-03-06T09:21:35Z | 2023-05-31T08:29:01Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | shirwy | 101,863,801 | U_kgDOBhJReQ | User | false |
huggingface/diffusers | 1,734,267,132 | I_kwDOHa8MBc5nXtD8 | 3,627 | https://github.com/huggingface/diffusers/issues/3627 | https://api.github.com/repos/huggingface/diffusers/issues/3627 | Retrieve.py throwing error when following training example on Custom diffusion on hugging face | ### Describe the bug
Retrieve.py script has issues
### Reproduction
follow the instructions on custom diffusion training and run it in a python 3.8 conda env.
https://huggingface.co/docs/diffusers/main/en/training/custom_diffusion
### Logs
```shell
(diff) ubuntu@ip-96-107-149-234:~/project/custom-diffusion_expts/... | closed | completed | false | 12 | [
"bug",
"stale"
] | [] | 2023-05-31T13:44:05Z | 2025-07-17T10:44:42Z | 2023-07-08T15:03:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nighting0le01 | 81,643,693 | MDQ6VXNlcjgxNjQzNjkz | User | false |
huggingface/diffusers | 1,362,193,457 | I_kwDOHa8MBc5RMWwx | 363 | https://github.com/huggingface/diffusers/issues/363 | https://api.github.com/repos/huggingface/diffusers/issues/363 | Possible Apple MPS issue - prompts array fails to generate any images | ### Describe the bug
If you pass an array of prompts (a list of strings) rather than single prompt (a single string) to the pipeline, under Apple `mps` you never get an actual image. You get a lot of NSFW warnings and some images with noise but that's it.
If you run the image generator multiple times in a loop inst... | closed | completed | false | 12 | [
"stale"
] | [] | 2022-09-05T15:37:28Z | 2023-03-21T18:46:11Z | 2022-11-08T15:04:00Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FahimF | 181,110 | MDQ6VXNlcjE4MTExMA== | User | false |
huggingface/diffusers | 1,735,312,436 | I_kwDOHa8MBc5nbsQ0 | 3,630 | https://github.com/huggingface/diffusers/issues/3630 | https://api.github.com/repos/huggingface/diffusers/issues/3630 | Unknown error in UNet2DModel | ### Describe the bug
I try to feed UNet2DModel with (BATCHS, 1, 28, 28) shape sample and (BATCHS,) timestep, but I got
"RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 8 but got size 7 for tensor number 1
in the list." in "diffusers/models/unet_2d_blocks.py:1726 in forward".
Here i... | closed | completed | false | 4 | [
"bug"
] | [] | 2023-06-01T02:13:39Z | 2023-06-01T14:03:48Z | 2023-06-01T14:03:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chuangshizhiqiang | 16,714,437 | MDQ6VXNlcjE2NzE0NDM3 | User | false |
huggingface/diffusers | 1,735,394,976 | I_kwDOHa8MBc5ncAag | 3,631 | https://github.com/huggingface/diffusers/issues/3631 | https://api.github.com/repos/huggingface/diffusers/issues/3631 | Inpainting pipeline throws exception when providing latents | ### Describe the bug
After [this](https://github.com/huggingface/diffusers/commit/d114d80fd2b5a1b7c5f3d2ec67b92a442d8f18f9) commit if I call the inpaint pipeline with latents I get:
```
<cut>
File "/usr/local/lib/python3.8/dist-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return fu... | closed | completed | false | 7 | [
"bug"
] | [] | 2023-06-01T03:55:58Z | 2023-06-02T19:49:07Z | 2023-06-02T16:03:17Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | reimager | 20,918,094 | MDQ6VXNlcjIwOTE4MDk0 | User | false |
huggingface/diffusers | 1,736,714,448 | I_kwDOHa8MBc5nhCjQ | 3,638 | https://github.com/huggingface/diffusers/issues/3638 | https://api.github.com/repos/huggingface/diffusers/issues/3638 | [New Pipeline]: Prompt-Free Diffusion: Taking "Text" out of Text-to-Image Diffusion Models | Repo: https://github.com/SHI-Labs/Prompt-Free-Diffusion
HF Demo: https://huggingface.co/spaces/shi-labs/Prompt-Free-Diffusion
Paper: https://arxiv.org/pdf/2305.16223.pdf | open | null | false | 13 | [
"New pipeline/model",
"Good second issue"
] | [] | 2023-06-01T16:43:23Z | 2023-08-29T08:51:43Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kadirnar | 36,204,372 | MDQ6VXNlcjM2MjA0Mzcy | User | false |
huggingface/diffusers | 1,736,717,993 | I_kwDOHa8MBc5nhDap | 3,639 | https://github.com/huggingface/diffusers/issues/3639 | https://api.github.com/repos/huggingface/diffusers/issues/3639 | [New Pipeline]: GlyphControl: Glyph Conditional Control for Visual Text Generation | Github: https://github.com/AIGText/GlyphControl-release
Demo: https://huggingface.co/spaces/AIGText/GlyphControl
Paper: https://arxiv.org/abs/2305.18259.pdf | closed | completed | false | 1 | [
"New pipeline/model"
] | [] | 2023-06-01T16:46:15Z | 2024-11-22T01:29:37Z | 2024-11-22T01:29:37Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | kadirnar | 36,204,372 | MDQ6VXNlcjM2MjA0Mzcy | User | false |
huggingface/diffusers | 1,737,274,815 | I_kwDOHa8MBc5njLW_ | 3,643 | https://github.com/huggingface/diffusers/issues/3643 | https://api.github.com/repos/huggingface/diffusers/issues/3643 | RuntimeError: upsample_nearest_nhwc only supports output tensors with less than INT_MAX elements when decoding latents | ### Describe the bug
standard stable diffusion pipeline when increasing batch size > 1, crashes when trying to decode latents
### Reproduction
pipe = diffusers.StableDiffusionPipeline.from_pretrained("Lykon/DreamShaper").to("cuda",torch.float16)
pipe.enable_xformers_memory_efficient_attention()
img = pipe("c... | closed | completed | false | 4 | [
"bug",
"stale"
] | [] | 2023-06-02T00:22:30Z | 2023-07-10T15:03:47Z | 2023-07-10T15:03:47Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ethansmith2000 | 98,723,285 | U_kgDOBeJl1Q | User | false |
huggingface/diffusers | 1,737,491,106 | I_kwDOHa8MBc5nkAKi | 3,645 | https://github.com/huggingface/diffusers/issues/3645 | https://api.github.com/repos/huggingface/diffusers/issues/3645 | [Need Help] Training or Finetune on `stable_diffusion_controlnet_inpaint.py` (inpaintinig pipeline with stable diffusion and controlnet) | Hi all,
I am looking for some references to create a training/finetune file for the [stable_diffusion_controlnet_inpaint.py](https://github.com/huggingface/diffusers/blob/main/examples/community/stable_diffusion_controlnet_inpaint.py). I have looked at the file like [train_controlnet.py](https://github.com/huggingfa... | closed | completed | false | 9 | [
"stale"
] | [] | 2023-06-02T05:23:27Z | 2024-01-23T14:07:48Z | 2023-07-10T15:03:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wuyujack | 10,961,996 | MDQ6VXNlcjEwOTYxOTk2 | User | false |
huggingface/diffusers | 1,737,556,479 | I_kwDOHa8MBc5nkQH_ | 3,646 | https://github.com/huggingface/diffusers/issues/3646 | https://api.github.com/repos/huggingface/diffusers/issues/3646 | AttributeError: type object 'ControlNetModel' has no attribute 'from_unet' | Hello everyone, When I train Controlnet, the above error occurs. My environment is as follows:
libgcc_mutex 0.1 main defaults
_openmp_mutex 5.1 1_gnu defaults
absl-py 1.4.0 pypi_0 pypi
accelerate ... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-02T06:36:54Z | 2023-07-10T15:03:44Z | 2023-07-10T15:03:44Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | air-y | 31,497,223 | MDQ6VXNlcjMxNDk3MjIz | User | false |
huggingface/diffusers | 1,737,932,239 | I_kwDOHa8MBc5nlr3P | 3,647 | https://github.com/huggingface/diffusers/issues/3647 | https://api.github.com/repos/huggingface/diffusers/issues/3647 | Dreambooth training with DeepFloyd | ### Describe the bug
I have been trying to get dreambooth working with Deepfloyd IF model. Following the instructions mentioned at https://github.com/huggingface/diffusers/tree/main/examples/dreambooth#full-dreambooth.
Script and instructions aren't working, tried with a few alternatives cli options, everything fai... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2023-06-02T10:32:48Z | 2023-08-14T15:04:18Z | 2023-08-14T15:04:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | mahendra0203 | 1,389,460 | MDQ6VXNlcjEzODk0NjA= | User | false |
huggingface/diffusers | 1,738,017,908 | I_kwDOHa8MBc5nmAx0 | 3,648 | https://github.com/huggingface/diffusers/issues/3648 | https://api.github.com/repos/huggingface/diffusers/issues/3648 | How much VRAM is required to finetune "CompVis/stable-diffusion-v1-4" model ? | I am trying to finetune the above mentioned model on "lambdalabs/pokemon-blip-captions" dataset , My GPU config is as follows, I am using single Tesla T4 with 15.36 GB VRAM , while fine tuning it on pokemon blip captions dataset using the examples provided [here](https://github.com/huggingface/diffusers/tree/main/examp... | closed | completed | false | 8 | [] | [] | 2023-06-02T11:26:45Z | 2023-09-11T05:31:40Z | 2023-06-05T07:10:45Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | atharmzaalo2023 | 134,913,498 | U_kgDOCAqd2g | User | false |
huggingface/diffusers | 1,738,562,088 | I_kwDOHa8MBc5noFoo | 3,651 | https://github.com/huggingface/diffusers/issues/3651 | https://api.github.com/repos/huggingface/diffusers/issues/3651 | Could not run train_dreambooth with multi-GPU setting | ### Describe the bug
I got the following error message while running the `train_dreambooth.py` script:
```
AttributeError: 'DistributedDataParallel' object has no attribute 'config'
```
It only occurs when I use `multi-gpu` setting.
### Reproduction
Follow the Finetune section in the [tutorial](https://huggin... | closed | completed | false | 0 | [
"bug"
] | [] | 2023-06-02T17:33:52Z | 2023-06-05T10:03:13Z | 2023-06-05T10:03:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dangkhoasdc | 3,388,047 | MDQ6VXNlcjMzODgwNDc= | User | false |
huggingface/diffusers | 1,738,716,949 | I_kwDOHa8MBc5norcV | 3,654 | https://github.com/huggingface/diffusers/issues/3654 | https://api.github.com/repos/huggingface/diffusers/issues/3654 | Stable Diffusion x4 upscaler has a different expected value range than other pipelines | for input/output in `pt` format, `StableDiffusionUpscalePipeline` expect the input to have value range `[-1,1]` and also output `[-1,1]`
see `preprocess` method here, i.e. It will not apply normalization if the input is a pytorch tensor
https://github.com/huggingface/diffusers/blob/94a0c644a8ce5b05a969859e0814ef488... | closed | completed | false | 3 | [] | [] | 2023-06-02T19:39:38Z | 2023-06-07T17:19:11Z | 2023-06-07T17:19:11Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | yiyixuxu | 12,631,849 | MDQ6VXNlcjEyNjMxODQ5 | User | false |
huggingface/diffusers | 1,738,865,435 | I_kwDOHa8MBc5npPsb | 3,655 | https://github.com/huggingface/diffusers/issues/3655 | https://api.github.com/repos/huggingface/diffusers/issues/3655 | Not Able to Import Unet2DModel on MacOS | ### Describe the bug
<img width="1283" alt="Screenshot 2023-06-02 at 2 45 40 PM" src="https://github.com/huggingface/diffusers/assets/3173623/2602e32d-867f-4165-a852-0348543ab11d">
### Reproduction
use this line in a Python notebook: "from diffusers import UNet2DModel, DDIMPipeline, DDPMPipeline, DDIMScheduler, DD... | closed | completed | false | 14 | [
"bug"
] | [] | 2023-06-02T21:55:57Z | 2023-06-07T17:19:17Z | 2023-06-07T17:19:17Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chetan-guevara | 3,173,623 | MDQ6VXNlcjMxNzM2MjM= | User | false |
huggingface/diffusers | 1,739,056,073 | I_kwDOHa8MBc5np-PJ | 3,658 | https://github.com/huggingface/diffusers/issues/3658 | https://api.github.com/repos/huggingface/diffusers/issues/3658 | Fix additional noise latents added to unet input in dreambooth upsampler training scripts | ### Describe the bug
The additional latents are random noise and instead need to be added to the original image as during inference
https://github.com/huggingface/diffusers/blob/7a39691362e5448b4417f37e135158fcda5ae7fb/examples/dreambooth/train_dreambooth.py#L1215-L1222
compared to
https://github.com/hugging... | closed | completed | false | 1 | [
"bug"
] | [
"williamberman"
] | 2023-06-03T01:59:18Z | 2023-06-05T09:47:05Z | 2023-06-05T09:47:05Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | williamberman | 6,505,431 | MDQ6VXNlcjY1MDU0MzE= | User | false |
huggingface/diffusers | 1,739,293,426 | I_kwDOHa8MBc5nq4Ly | 3,660 | https://github.com/huggingface/diffusers/issues/3660 | https://api.github.com/repos/huggingface/diffusers/issues/3660 | `torch.compile` not working with lora or xformers | ### Describe the bug
I'm experiencing some errors when using `torch.compile` with `UNet2DConditionModel`, possibly caused by either lora or xformers, which I can't decide for now.
I did some tests trying to identify the problem but failed. So I'm posting my test results here, hopefully serving as some useful clue... | closed | completed | false | 18 | [
"bug"
] | [] | 2023-06-03T08:45:29Z | 2024-07-12T09:11:54Z | 2023-06-30T07:42:32Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | sunhs | 1,916,614 | MDQ6VXNlcjE5MTY2MTQ= | User | false |
huggingface/diffusers | 1,739,322,191 | I_kwDOHa8MBc5nq_NP | 3,661 | https://github.com/huggingface/diffusers/issues/3661 | https://api.github.com/repos/huggingface/diffusers/issues/3661 | `from_ckpt` not work for Stable Diffusion 2.x weights | ### Describe the bug
`from_ckpt` not work for Stable Diffusion 2.x weights
### Reproduction
```python
from diffusers import StableDiffusionPipeline
pipe = StableDiffusionPipeline.from_ckpt(
'https://huggingface.co/waifu-diffusion/wd-1-5-beta3/blob/main/wd-illusion-fp16.safetensors'
)
```
Just try to load a... | closed | completed | false | 2 | [
"bug"
] | [] | 2023-06-03T09:41:32Z | 2023-06-06T21:39:12Z | 2023-06-06T21:39:12Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | ctrysbita | 8,357,481 | MDQ6VXNlcjgzNTc0ODE= | User | false |
huggingface/diffusers | 1,739,856,096 | I_kwDOHa8MBc5ntBjg | 3,665 | https://github.com/huggingface/diffusers/issues/3665 | https://api.github.com/repos/huggingface/diffusers/issues/3665 | Can I save just differences of weights between trained model by dreambooth and base model? | Hey guys!
I have a question.
When we are training a stable diffusion model by dreambooth script with a 15 images of new inctance (like human face), we will save a result that has about 2 or 4 gig size.
My question is, can we save just difference of weights as a result and when we want load that, we will make new stab... | closed | completed | false | 5 | [] | [] | 2023-06-03T20:00:45Z | 2023-06-05T10:31:33Z | 2023-06-05T10:16:21Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hosein-moayedi | 38,086,978 | MDQ6VXNlcjM4MDg2OTc4 | User | false |
huggingface/diffusers | 1,740,412,899 | I_kwDOHa8MBc5nvJfj | 3,667 | https://github.com/huggingface/diffusers/issues/3667 | https://api.github.com/repos/huggingface/diffusers/issues/3667 | Understanding Dreambooth and text2img | Hi,
i am currently working with the dreambooth and a custom version of mine of the text2img script. I was wondering however if there are any differences between these two if prior preservation is not used (as i do currently).
I get different results with both which kind of confuses me. I couldnt really find mater... | closed | completed | false | 2 | [
"stale"
] | [] | 2023-06-04T14:45:37Z | 2023-07-12T15:03:48Z | 2023-07-12T15:03:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | leabrix | 130,184,959 | U_kgDOB8J2_w | User | false |
huggingface/diffusers | 1,740,741,383 | I_kwDOHa8MBc5nwZsH | 3,669 | https://github.com/huggingface/diffusers/issues/3669 | https://api.github.com/repos/huggingface/diffusers/issues/3669 | General question: what are the steps to debug if the image produced is just wrong? | I have a lora(lycoris) that I have tested with A1111's webui and I'm pretty happy with the result. When I tried to use it with `diffusers` it just give me corrupted image. The lora brings some desired effect (like white background), but the overall image is just not right.
I have included some personal code to use l... | closed | completed | false | 9 | [
"stale"
] | [] | 2023-06-05T01:44:49Z | 2023-07-13T15:03:51Z | 2023-07-13T15:03:51Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wangdong2023 | 8,127,162 | MDQ6VXNlcjgxMjcxNjI= | User | false |
huggingface/diffusers | 1,362,542,535 | I_kwDOHa8MBc5RNr_H | 367 | https://github.com/huggingface/diffusers/issues/367 | https://api.github.com/repos/huggingface/diffusers/issues/367 | Make a Hugging Face "optimized" pipeline example | Currently, using Stable Diffusion with huggingface/diffusers often fails due to VRAM shortage, even when using a 16GB GPU, and even when trying to make fairly humble generations, like an image with a 16:9 aspect ratio (which is important for making videos).
[Basujinal's fork](https://github.com/basujindal/stable-dif... | closed | completed | false | 6 | [] | [] | 2022-09-05T23:31:40Z | 2022-09-07T18:40:00Z | 2022-09-07T18:40:00Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | exo-pla-net | 47,930,742 | MDQ6VXNlcjQ3OTMwNzQy | User | false |
huggingface/diffusers | 1,741,136,513 | I_kwDOHa8MBc5nx6KB | 3,671 | https://github.com/huggingface/diffusers/issues/3671 | https://api.github.com/repos/huggingface/diffusers/issues/3671 | Appropriate Number of Epochs to fine tune train_text_to_image_lora.py | Considering I am having dataset of 16 images , 4 images per subject (There are 4 unique subjects in the images and hence there are 16 images , also subjects are under different conditions like smiling , laughter or anger) , what should be appropriate number of epochs to fine tune lora so as to not overfit , also how to... | closed | completed | false | 2 | [] | [] | 2023-06-05T07:15:44Z | 2023-06-06T21:28:14Z | 2023-06-06T21:28:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | atharmzaalo2023 | 134,913,498 | U_kgDOCAqd2g | User | false |
huggingface/diffusers | 1,741,329,910 | I_kwDOHa8MBc5nypX2 | 3,672 | https://github.com/huggingface/diffusers/issues/3672 | https://api.github.com/repos/huggingface/diffusers/issues/3672 | Concurrent Thread Failure in Image Inference: StableDiffusionPipeline | ### Describe the bug
When I attempt to concurrently initiate two threads for image inference in one pipe, the process fails and the following information is displayed as logs during the inference. I intend to submit a pull request for thread-safety in one inference pipe if anyone else need.
### Reproduction
```p... | closed | completed | false | 23 | [
"bug"
] | [] | 2023-06-05T09:06:07Z | 2025-12-03T19:16:02Z | 2023-06-13T01:52:09Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Fantast616 | 45,848,260 | MDQ6VXNlcjQ1ODQ4MjYw | User | false |
huggingface/diffusers | 1,741,482,008 | I_kwDOHa8MBc5nzOgY | 3,674 | https://github.com/huggingface/diffusers/issues/3674 | https://api.github.com/repos/huggingface/diffusers/issues/3674 | Problem while reading data from metadata.jsonl file while fine tuning train_text_to_image_lora.py | I am trying to fine tune lora model on custom dataset . My Directory structure is as follows
|- Images (folder)
|- train(folder)
| |- metadata.jsonl
| |- ashish_1.jpg
| |- ashish_2.jpg
| |- ...
|- train_text_to_image_lora.py
My metadata.jsonl file looks like this
{"file_name": "train/ashish_1.j... | closed | completed | false | 6 | [] | [] | 2023-06-05T10:34:42Z | 2023-06-06T05:50:28Z | 2023-06-06T05:50:27Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | atharmzaalo2023 | 134,913,498 | U_kgDOCAqd2g | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.