repo stringclasses 1
value | github_id int64 1.27B 4.61B | github_node_id stringlengths 18 24 | number int64 8 13.9k | 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 2026-06-08 13:44:32 | updated_at stringdate 2022-06-12 22:18:01 2026-06-08 13:47:08 | closed_at stringdate 2022-06-12 22:18:01 2026-06-08 10:50:20 ⌀ | author_association stringclasses 3
values | milestone_title stringclasses 0
values | snapshot_id stringclasses 64
values | extracted_at stringdate 2026-04-07 13:34:13 2026-06-08 14:35:15 | author_login stringlengths 3 28 | author_id int64 1.54k 290M | author_node_id stringlengths 12 20 | author_type stringclasses 3
values | author_site_admin bool 1
class |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
huggingface/diffusers | 2,341,044,403 | I_kwDOHa8MBc6LiYCz | 8,432 | https://github.com/huggingface/diffusers/issues/8432 | https://api.github.com/repos/huggingface/diffusers/issues/8432 | from diffusers.loaders.autoencoder import FromOriginalVAEMixin fails with ImportError | On 0.28.2, this
from diffusers.loaders.autoencoder import FromOriginalVAEMixin
will fail with this error
ImportError: cannot import name 'create_diffusers_vae_model_from_ldm' from 'diffusers.loaders.single_file_utils' (/usr/local/lib/python3.10/dist-packages/diffusers/loaders/single_file_utils.py)
I... | closed | completed | false | 6 | [] | [] | 2024-06-07T19:28:32Z | 2024-10-20T23:04:56Z | 2024-10-20T23:04:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | juancamilog | 2,013,726 | MDQ6VXNlcjIwMTM3MjY= | User | false |
huggingface/diffusers | 2,341,240,340 | I_kwDOHa8MBc6LjH4U | 8,434 | https://github.com/huggingface/diffusers/issues/8434 | https://api.github.com/repos/huggingface/diffusers/issues/8434 | Support multiple control nets in the `StableDiffusionControlNetXSPipeline`/`StableDiffusionXLControlNetXSPipeline` | **Is your feature request related to a problem? Please describe.**
The `StableDiffusionControlNetPipeline`/`StableDiffusionXLControlNetPipeline` allow you to specify a list of control nets (i.e. depth and canny)
However, the new `ControlNetXS` pipelines only allow you provide a single control net
**Describe the ... | open | null | false | 9 | [
"good first issue",
"community-examples",
"contributions-welcome"
] | [] | 2024-06-07T22:23:30Z | 2025-08-08T02:14:51Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chrismaltais | 17,098,959 | MDQ6VXNlcjE3MDk4OTU5 | User | false |
huggingface/diffusers | 2,341,390,683 | I_kwDOHa8MBc6Ljslb | 8,435 | https://github.com/huggingface/diffusers/issues/8435 | https://api.github.com/repos/huggingface/diffusers/issues/8435 | Runtime error with from_single_file | ### Describe the bug
Versioning diffusers to 0.29.0.dev0 did not solve the problem.
### Reproduction
pip install diffusers
from diffusers import StableDiffusionPipeline
pipe = StableDiffusionPipeline.from_single_file("https://huggingface.co/gsdf/Counterfeit-V2.5/blob/main/Counterfeit-V2.5_fp16.safetensors")
### L... | closed | completed | false | 9 | [
"bug"
] | [] | 2024-06-08T02:44:20Z | 2024-07-12T10:21:07Z | 2024-07-12T10:21:07Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | suzukimain | 131,413,573 | U_kgDOB9U2RQ | User | false |
huggingface/diffusers | 2,341,739,297 | I_kwDOHa8MBc6LlBsh | 8,437 | https://github.com/huggingface/diffusers/issues/8437 | https://api.github.com/repos/huggingface/diffusers/issues/8437 | in diffuser audio | does audio pipline have input audio ? like how can we make our own audio | closed | completed | false | 1 | [] | [] | 2024-06-08T16:57:17Z | 2024-06-09T09:50:45Z | 2024-06-09T09:38:41Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xalteropsx | 103,671,642 | U_kgDOBi3nWg | User | false |
huggingface/diffusers | 2,342,002,243 | I_kwDOHa8MBc6LmB5D | 8,439 | https://github.com/huggingface/diffusers/issues/8439 | https://api.github.com/repos/huggingface/diffusers/issues/8439 | How to use EDM2 model with diffusers? | model safetensors: https://huggingface.co/RedRocket/Fluffyrock-Unbound/blob/main/Fluffyrock-Unbound-v1-1.safetensors
yaml: https://huggingface.co/RedRocket/Fluffyrock-Unbound/raw/main/Fluffyrock-Unbound-v1-1.yaml
colab demo:
https://colab.research.google.com/drive/1LSGvjWXNVjs6Tthcpf0F5VwuTFJ_d-oB
results:
... | open | null | false | 1 | [
"stale"
] | [] | 2024-06-09T03:39:05Z | 2024-09-14T15:10:19Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | s9anus98a | 151,509,142 | U_kgDOCQfYlg | User | false |
huggingface/diffusers | 2,342,026,740 | I_kwDOHa8MBc6LmH30 | 8,440 | https://github.com/huggingface/diffusers/issues/8440 | https://api.github.com/repos/huggingface/diffusers/issues/8440 | ValueError when using `FromOriginalModelMixin` in subclasses | ### Describe the bug
ValueError when using `FromOriginalModelMixin` in subclasses due to the following strict class check:
https://github.com/huggingface/diffusers/blob/83bc6c94eaeb6f7704a2a428931cf2d9ad973ae9/src/diffusers/loaders/single_file_model.py#L147-L151
Could it be possible to extend `from_single_file` ... | closed | completed | false | 3 | [
"bug"
] | [] | 2024-06-09T05:10:25Z | 2024-07-05T05:07:11Z | 2024-07-05T05:07:11Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | luocfprime | 79,003,314 | MDQ6VXNlcjc5MDAzMzE0 | User | false |
huggingface/diffusers | 2,342,034,351 | I_kwDOHa8MBc6LmJuv | 8,441 | https://github.com/huggingface/diffusers/issues/8441 | https://api.github.com/repos/huggingface/diffusers/issues/8441 | create_dynamic_module infinite recursion when setting HF_HOME using relative path | ### Describe the bug
create_dynamic_module infinite recursion when setting HF_HOME using relative path
### Reproduction
```py
import os
os.environ["HF_HOME"] = ".cache/huggingface"
from diffusers.utils.dynamic_modules_utils import DIFFUSERS_DYNAMIC_MODULE_NAME, create_dynamic_module
from diffusers.utils.... | open | null | false | 6 | [
"bug",
"stale"
] | [] | 2024-06-09T05:39:17Z | 2024-11-09T15:05:42Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | luocfprime | 79,003,314 | MDQ6VXNlcjc5MDAzMzE0 | User | false |
huggingface/diffusers | 2,342,102,328 | I_kwDOHa8MBc6LmaU4 | 8,443 | https://github.com/huggingface/diffusers/issues/8443 | https://api.github.com/repos/huggingface/diffusers/issues/8443 | backwards compatibility breaking change from sharded checkpoints | ### Describe the bug
Since #7830 we have SDXL checkpoints sharded automatically, which did not happen before. These checkpoints are not loadable on earlier versions.
### Reproduction
I think all one has to do is load `from_pretrained()` and then `save_pretrained()`
### Logs
```shell
$ ls /home/user/trai... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-06-09T08:31:45Z | 2024-06-12T12:49:15Z | 2024-06-12T12:49:14Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bghira | 59,658,056 | MDQ6VXNlcjU5NjU4MDU2 | User | false |
huggingface/diffusers | 1,409,558,356 | I_kwDOHa8MBc5UBCdU | 845 | https://github.com/huggingface/diffusers/issues/845 | https://api.github.com/repos/huggingface/diffusers/issues/845 | Allow safety checker pipeline configuration that returns boolean array but does not black out images | **Is your feature request related to a problem? Please describe.**
See this PR comment: https://github.com/huggingface/diffusers/pull/815#discussion_r994418216
TLDR: with recent changes in #815, developers have the ability to disable the safety checker. Currently, the only options available to devs is to either hav... | closed | completed | false | 43 | [
"stale"
] | [] | 2022-10-14T16:01:29Z | 2023-10-25T21:33:27Z | 2023-03-09T15:04:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | theahura | 6,380,637 | MDQ6VXNlcjYzODA2Mzc= | User | false |
huggingface/diffusers | 2,342,531,741 | I_kwDOHa8MBc6LoDKd | 8,450 | https://github.com/huggingface/diffusers/issues/8450 | https://api.github.com/repos/huggingface/diffusers/issues/8450 | Running `diffusers/stable-diffusion-xl-1.0-inpainting-0.1` with `StableDiffusionXLInpaintPipeline` introduces weird noise when strength is set to 1.0 during inference | ### Describe the bug
I am currently comparing the inpainting generated results between the [diffusers/stable-diffusion-xl-1.0-inpainting-0.1](https://huggingface.co/diffusers/stable-diffusion-xl-1.0-inpainting-0.1) model and the [stabilityai/stable-diffusion-2-inpainting](https://huggingface.co/stabilityai/stable-diff... | open | null | false | 15 | [
"bug",
"stale"
] | [] | 2024-06-09T22:58:05Z | 2024-09-14T15:10:07Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jackylu0124 | 28,615,340 | MDQ6VXNlcjI4NjE1MzQw | User | false |
huggingface/diffusers | 2,342,867,388 | I_kwDOHa8MBc6LpVG8 | 8,451 | https://github.com/huggingface/diffusers/issues/8451 | https://api.github.com/repos/huggingface/diffusers/issues/8451 | The training details about Unet parameters frozen (and no script for train sd1.x) | ### Describe the bug
Why are the UNet parameters frozen during training for SD1v5, but not for SDXL? the haggingface training sdxl script sets " Unet.train() "
### Reproduction
diffusers : 0.28
***** huggingface train sdxl *********
vae.requires_grad_(False)
text_encoder_one.requires_grad_(False)
text_encode... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2024-06-10T05:10:28Z | 2024-10-09T15:18:57Z | 2024-10-09T15:18:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hda-xian | 61,269,950 | MDQ6VXNlcjYxMjY5OTUw | User | false |
huggingface/diffusers | 1,409,590,656 | I_kwDOHa8MBc5UBKWA | 846 | https://github.com/huggingface/diffusers/issues/846 | https://api.github.com/repos/huggingface/diffusers/issues/846 | Dreambooth class image generation fails if filename already exists | ### Describe the bug
When it generates class images, the `train_dreambooth.py` script does not check if the filename already exists.
So it might overwrite an existing image with the same name, and in consequence not enough images are present in the folder and the script breaks later.
### Reproduction
This can happe... | closed | completed | false | 3 | [
"bug"
] | [
"patil-suraj"
] | 2022-10-14T16:30:41Z | 2022-10-20T14:12:54Z | 2022-10-20T14:12:53Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | leszekhanusz | 4,641,499 | MDQ6VXNlcjQ2NDE0OTk= | User | false |
huggingface/diffusers | 2,344,758,358 | I_kwDOHa8MBc6LwixW | 8,463 | https://github.com/huggingface/diffusers/issues/8463 | https://api.github.com/repos/huggingface/diffusers/issues/8463 | AttributeError: 'NoneType' object has no attribute 'tokenize' for InstructP2P-SDXL | ### Describe the bug
```bash
Loading pipeline components...: 100%|█████████████████████████████████████████████████████████████████████████████████████████████| 5/5 [00:00<00:00, 7.77it/s]
Traceback (most recent call last):
File "/mnt/bn/sijie-us-nas-8/liming/code/diffusers/test.py", line 12, in <module>
... | closed | completed | false | 6 | [
"bug"
] | [] | 2024-06-10T20:21:27Z | 2024-07-01T10:14:27Z | 2024-06-11T12:57:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | liming-ai | 73,068,772 | MDQ6VXNlcjczMDY4Nzcy | User | false |
huggingface/diffusers | 2,344,852,675 | I_kwDOHa8MBc6Lw5zD | 8,464 | https://github.com/huggingface/diffusers/issues/8464 | https://api.github.com/repos/huggingface/diffusers/issues/8464 | SEGA with cpu offload fails when specifying a seed | Hello :wave:
### Describe the bug
If I use `SemanticStableDiffusionPipeline` + `enable_model_cpu_offload` + a seed (either through `torch.manual_seed(0)` or via `torch.Generator`, I get tensors in different devices
### Reproduction
First install latest `diffusers` and `accelerate` versions
```bash
pip ... | closed | completed | false | 1 | [
"bug"
] | [] | 2024-06-10T21:24:22Z | 2024-06-11T16:37:50Z | 2024-06-11T16:37:50Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | osanseviero | 7,246,357 | MDQ6VXNlcjcyNDYzNTc= | User | false |
huggingface/diffusers | 2,345,044,954 | I_kwDOHa8MBc6Lxova | 8,465 | https://github.com/huggingface/diffusers/issues/8465 | https://api.github.com/repos/huggingface/diffusers/issues/8465 | StableCascadeUNet.from_single_file not working error runwayml/stable-diffusion-v1-5 does not appear to have a file named config.json. | i want to try this stable cascade fine tuned model https://civitai.com/models/326354/altcascade
script:
```
import torch
from math import ceil
from diffusers import (
StableCascadeDecoderPipeline,
StableCascadePriorPipeline,
StableCascadeUNet
)
# from diffusers.models import StableCascadeUNet
... | open | null | false | 4 | [
"stale",
"single_file"
] | [
"DN6"
] | 2024-06-10T23:47:47Z | 2024-09-14T15:09:57Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | s9anus98a | 151,509,142 | U_kgDOCQfYlg | User | false |
huggingface/diffusers | 2,345,548,569 | I_kwDOHa8MBc6LzjsZ | 8,469 | https://github.com/huggingface/diffusers/issues/8469 | https://api.github.com/repos/huggingface/diffusers/issues/8469 | ip-adapter-full-face_sd15 is producing corrupt / broken faces | ### Describe the bug
I'm trying to use the the following IP Adapter using diffusers (ip-adapter-full-face_sd15.bin):
https://huggingface.co/h94/IP-Adapter/blob/main/models/ip-adapter-full-face_sd15.bin
However, as soon as I provide a human face as a reference image, the results look broken (see example below).
... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-06-11T06:51:20Z | 2024-06-12T05:59:24Z | 2024-06-12T05:59:24Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jroessler | 23,525,054 | MDQ6VXNlcjIzNTI1MDU0 | User | false |
huggingface/diffusers | 2,346,752,440 | I_kwDOHa8MBc6L4Jm4 | 8,472 | https://github.com/huggingface/diffusers/issues/8472 | https://api.github.com/repos/huggingface/diffusers/issues/8472 | Questions of pre-training LoRA with other modules simutaneously | ### Describe the bug
Sorry for asking questions unrelated to diffusers... I am trying to jountly train an ELLA model with a UNET LoRA (rank: 64). However, I confronted a weird situation, the ELLA model worked well while the LoRA crashed and made the output like random noise. Have you guys witnessed this phenomenon?
... | closed | completed | true | 0 | [
"bug"
] | [] | 2024-06-11T15:42:07Z | 2024-06-18T15:34:38Z | 2024-06-18T15:34:38Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | George0726 | 38,740,075 | MDQ6VXNlcjM4NzQwMDc1 | User | false |
huggingface/diffusers | 2,346,856,064 | I_kwDOHa8MBc6L4i6A | 8,474 | https://github.com/huggingface/diffusers/issues/8474 | https://api.github.com/repos/huggingface/diffusers/issues/8474 | `.from_single_file` breaking in v28.2 if model_index.json is empty, unintuitive logs | ### Describe the bug
I've been using diffuser for couple months, primarily for `StableDiffusionPipeline` and `StableDiffusionXLPipeline` image inference using a downloaded .safetensors
A while back I upgraded to v28.2 and `.from_single_file` complained (refer to logs)
I did some debugging and concluded that for ... | closed | completed | false | 13 | [
"bug",
"stale"
] | [] | 2024-06-11T16:39:05Z | 2024-12-04T16:46:51Z | 2024-12-04T16:46:51Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | underscore-en | 52,004,879 | MDQ6VXNlcjUyMDA0ODc5 | User | false |
huggingface/diffusers | 2,347,158,267 | I_kwDOHa8MBc6L5sr7 | 8,477 | https://github.com/huggingface/diffusers/issues/8477 | https://api.github.com/repos/huggingface/diffusers/issues/8477 | Error in Research Projects Consistency Training Script(DistributedDataParallel Error) line 1198 | ### Model/Pipeline/Scheduler description
ConsistencyModelPipeline
In diffusers/examples/research_projects /consistency_training/ example, When I use multi_gpu then there is this error:- Traceback (most recent call last): File "/kaggle/working/train_cm_ct_unconditional.py", line 1438, in <module> main(args) ... | closed | completed | false | 2 | [] | [] | 2024-06-11T19:43:18Z | 2024-07-17T13:33:13Z | 2024-07-17T13:33:13Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | KetanMann | 117,455,191 | U_kgDOBwA5Vw | User | false |
huggingface/diffusers | 2,348,770,588 | I_kwDOHa8MBc6L_2Uc | 8,482 | https://github.com/huggingface/diffusers/issues/8482 | https://api.github.com/repos/huggingface/diffusers/issues/8482 | Support for stable-diffusion-3-medium model | **Is your feature request related to a problem? Please describe.**
No.
**Describe the solution you'd like.**
Support SD3 medium model (https://huggingface.co/stabilityai/stable-diffusion-3-medium).
**Describe alternatives you've considered.**
It seems like this PR(https://github.com/huggingface/diffusers/issue... | closed | completed | false | 20 | [] | [] | 2024-06-12T13:22:47Z | 2024-07-01T10:48:56Z | 2024-07-01T10:48:56Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | rupeshs | 3,255,994 | MDQ6VXNlcjMyNTU5OTQ= | User | false |
huggingface/diffusers | 2,349,081,434 | I_kwDOHa8MBc6MBCNa | 8,485 | https://github.com/huggingface/diffusers/issues/8485 | https://api.github.com/repos/huggingface/diffusers/issues/8485 | Support for SD3 LoRA | **Is your feature request related to a problem? Please describe.**
No.
**Describe the solution you'd like.**
When will you release the LoRA finetuning script for SD3? It would be great to try it out, especially since Stability pointed out that the finetuning capabilities are one of its standout features.
| closed | completed | false | 1 | [] | [] | 2024-06-12T15:33:37Z | 2024-06-12T16:03:32Z | 2024-06-12T16:03:31Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | boraykasap | 73,942,207 | MDQ6VXNlcjczOTQyMjA3 | User | false |
huggingface/diffusers | 2,349,571,552 | I_kwDOHa8MBc6MC53g | 8,488 | https://github.com/huggingface/diffusers/issues/8488 | https://api.github.com/repos/huggingface/diffusers/issues/8488 | StableDiffusion3Pipeline `from_pretrained` Exception | ### Describe the bug
The pipe fails to load on component 5 using `from_pretrained`. Since the diffusers weights were just updated an hour ago, possibly the conversion is broken again like Cascade was?
I was typing this out in the #8483 but it was merged before I could submit it, lol.
### Reproduction
```pyt... | closed | completed | false | 20 | [
"bug"
] | [] | 2024-06-12T19:52:30Z | 2024-12-30T22:25:49Z | 2024-06-12T23:01:30Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Beinsezii | 39,478,211 | MDQ6VXNlcjM5NDc4MjEx | User | false |
huggingface/diffusers | 2,349,625,364 | I_kwDOHa8MBc6MDHAU | 8,493 | https://github.com/huggingface/diffusers/issues/8493 | https://api.github.com/repos/huggingface/diffusers/issues/8493 | Error(s) in loading state_dict for SD3Transformer2DModel | ### Describe the bug
```
RuntimeError: Error(s) in loading state_dict for SD3Transformer2DModel:
Missing key(s) in state_dict: "pos_embed.pos_embed", "transformer_blocks.14.attn.to_add_out.weight", "transformer_blocks.14.attn.to_add_out.bias", "transformer_blocks.14.ff.net.0.proj.weight", "transformer_blocks... | closed | completed | false | 15 | [
"bug"
] | [] | 2024-06-12T20:27:50Z | 2024-06-12T22:59:38Z | 2024-06-12T21:39:29Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | eliphatfs | 23,738,781 | MDQ6VXNlcjIzNzM4Nzgx | User | false |
huggingface/diffusers | 2,349,664,990 | I_kwDOHa8MBc6MDQre | 8,494 | https://github.com/huggingface/diffusers/issues/8494 | https://api.github.com/repos/huggingface/diffusers/issues/8494 | SD3 - unexpected keyword argument 'force_hooks' | ### Describe the bug
Get the following error while executing the demo code:
Code:
import torch
from diffusers import StableDiffusion3Pipeline
pipe = StableDiffusion3Pipeline.from_pretrained("stabilityai/stable-diffusion-3-medium-diffusers", torch_dtype=torch.float16)
pipe = pipe.to("mps")
image = pipe(... | closed | completed | false | 3 | [
"bug"
] | [] | 2024-06-12T20:54:13Z | 2024-06-13T17:40:31Z | 2024-06-12T21:15:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TEGNoV | 32,719,570 | MDQ6VXNlcjMyNzE5NTcw | User | false |
huggingface/diffusers | 2,349,695,643 | I_kwDOHa8MBc6MDYKb | 8,495 | https://github.com/huggingface/diffusers/issues/8495 | https://api.github.com/repos/huggingface/diffusers/issues/8495 | SD3 - data did not match any variant of untagged enum PyPreTokenizerTypeWrapper | ### Describe the bug
Error while execution of demo code:
Traceback (most recent call last):
File "/Users/fp/Library/CloudStorage/OneDrive-Personal/1. Projekte/Code/Adobe StableDiffusion/mlx/mydiffusers.py", line 34, in <module>
pipe = StableDiffusion3Pipeline.from_pretrained("stabilityai/stable-diffusion-... | closed | completed | false | 9 | [
"bug"
] | [] | 2024-06-12T21:17:05Z | 2024-06-13T10:00:36Z | 2024-06-13T10:00:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TEGNoV | 32,719,570 | MDQ6VXNlcjMyNzE5NTcw | User | false |
huggingface/diffusers | 2,349,796,826 | I_kwDOHa8MBc6MDw3a | 8,500 | https://github.com/huggingface/diffusers/issues/8500 | https://api.github.com/repos/huggingface/diffusers/issues/8500 | SD3 Token Limit? | ### Describe the bug
From the SD3 research paper, it seems that the new context window (token limit) should be 77 + 77 = 154? However, submitting a prompt longer than 77 tokens gives a truncation warning.
### Reproduction
any prompt with > 77 tokens
### Logs
_No response_
### System Info
diffusers source
### Wh... | closed | completed | false | 16 | [
"bug"
] | [] | 2024-06-12T22:46:48Z | 2024-07-11T06:48:41Z | 2024-07-11T06:48:41Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | neuron-party | 96,799,331 | U_kgDOBcUKYw | User | false |
huggingface/diffusers | 2,349,810,712 | I_kwDOHa8MBc6MD0QY | 8,501 | https://github.com/huggingface/diffusers/issues/8501 | https://api.github.com/repos/huggingface/diffusers/issues/8501 | Allow configuring `shift=` for SD3 dynamically | **Is your feature request related to a problem? Please describe.**
Allow passing `shift=` per inference call (like timesteps) on the pipeline, for flow matching scheduler, or allow `set_shift()` etc. on the scheduler. This seems to be the key to getting good results with SD3 https://x.com/bfitzgerald242/status/1801018... | closed | completed | false | 11 | [] | [
"hlky"
] | 2024-06-12T22:59:42Z | 2024-12-17T21:40:10Z | 2024-12-17T21:40:10Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | isidentical | 47,358,913 | MDQ6VXNlcjQ3MzU4OTEz | User | false |
huggingface/diffusers | 2,349,862,046 | I_kwDOHa8MBc6MEAye | 8,502 | https://github.com/huggingface/diffusers/issues/8502 | https://api.github.com/repos/huggingface/diffusers/issues/8502 | SD3 medium model Generate 512x512 image | ### Describe the bug
When trying to generate image from diffusers I just notice the image is kinda flop and idid try saved the image and the resolution is 512x512
<img width="1280" alt="Screenshot 2024-06-13 at 8 53 08" src="https://github.com/huggingface/diffusers/assets/131154780/13fd884d-becf-4886-a94f-f4d0da7af5e... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-06-12T23:55:10Z | 2024-06-13T09:21:47Z | 2024-06-13T09:21:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | pandora523 | 131,154,780 | U_kgDOB9FDXA | User | false |
huggingface/diffusers | 2,349,868,156 | I_kwDOHa8MBc6MECR8 | 8,503 | https://github.com/huggingface/diffusers/issues/8503 | https://api.github.com/repos/huggingface/diffusers/issues/8503 | SD3 and Gradient checkpointing gives error and crashes | ### Describe the bug
**Describe the bug**
Activating --gradient_checkpointing in either Lora or DB scripts for SD3 causes: TypeError: layer_norm(): argument 'input' (position 1) must be Tensor, not tuple, which crashes the run, without it, LoRA runs fine at about 20GB vram usage batch size 1 with AdamW8bit
![ima... | closed | completed | false | 7 | [
"bug",
"stale"
] | [] | 2024-06-13T00:02:07Z | 2024-11-18T20:31:46Z | 2024-11-18T20:31:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bluvoll | 23,042,093 | MDQ6VXNlcjIzMDQyMDkz | User | false |
huggingface/diffusers | 2,349,878,321 | I_kwDOHa8MBc6MEEwx | 8,504 | https://github.com/huggingface/diffusers/issues/8504 | https://api.github.com/repos/huggingface/diffusers/issues/8504 | Allow user to pick which Text Encoders to train with SD3-Medium. | **Is your feature request related to a problem? Please describe.**
Kinda yes, it seems both training scripts for LoRA and Dreambooth load all 3 Text Encoders and Tokenizers which leads to high vram usage, 20GB on Rank 32 Lora, batch size 1, with AdamW8Bit
**Describe the solution you'd like.**
Allow users to train ... | open | null | false | 2 | [
"stale"
] | [] | 2024-06-13T00:14:19Z | 2024-09-14T15:09:39Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | bluvoll | 23,042,093 | MDQ6VXNlcjIzMDQyMDkz | User | false |
huggingface/diffusers | 2,349,942,264 | I_kwDOHa8MBc6MEUX4 | 8,507 | https://github.com/huggingface/diffusers/issues/8507 | https://api.github.com/repos/huggingface/diffusers/issues/8507 | [SD3] train_dreambooth_sd3.py --train_text_encoder fails | ### Describe the bug
`encode_prompt()` is called with a parameter that is not defined in code (`text_input_ids_list`)
### Reproduction
`accelerate launch train_dreambooth_sd3.py [args] --train_text_encoder`
### Logs
```shell
Traceback (most recent call last):
File "/root/diffusers/examples/dreamboot... | open | null | false | 3 | [
"bug",
"stale"
] | [] | 2024-06-13T01:25:53Z | 2024-09-14T15:09:36Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | float-trip | 102,226,344 | U_kgDOBhfZqA | User | false |
huggingface/diffusers | 2,349,989,392 | I_kwDOHa8MBc6MEf4Q | 8,508 | https://github.com/huggingface/diffusers/issues/8508 | https://api.github.com/repos/huggingface/diffusers/issues/8508 | Support for train_text_to_image_sd3.py | null | closed | completed | false | 1 | [] | [] | 2024-06-13T02:13:12Z | 2024-06-13T03:07:51Z | 2024-06-13T03:07:51Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CanvaChen | 18,375,123 | MDQ6VXNlcjE4Mzc1MTIz | User | false |
huggingface/diffusers | 2,350,009,678 | I_kwDOHa8MBc6MEk1O | 8,509 | https://github.com/huggingface/diffusers/issues/8509 | https://api.github.com/repos/huggingface/diffusers/issues/8509 | controlnet++ | ### Model/Pipeline/Scheduler description
https://huggingface.co/huchenlei/ControlNet_plus_plus_collection_fp16/tree/main
https://huggingface.co/huchenlei/ControlNet_plus_plus_collection_fp16/discussions/1
```
Traceback (most recent call last):
File "/home/zznet/workspace/ai-pipe/prod-ai-pipe-replace-model/ru... | closed | not_planned | false | 4 | [
"stale"
] | [] | 2024-06-13T02:33:11Z | 2024-11-14T22:22:11Z | 2024-11-14T22:22:11Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | crapthings | 1,147,704 | MDQ6VXNlcjExNDc3MDQ= | User | false |
huggingface/diffusers | 1,409,881,502 | I_kwDOHa8MBc5UCRWe | 851 | https://github.com/huggingface/diffusers/issues/851 | https://api.github.com/repos/huggingface/diffusers/issues/851 | Found dtype Float but expected Half | ### Describe the bug
I am trying to use the "Training on a 8 GB GPU" Dreambooth instructions. I set up the accelerate config and everything else to the best of my knowledge. I have it running all the way to displaying the training steps screen, but then I get an error "Found dtype Float but expected Half." I would app... | closed | completed | false | 1 | [
"bug"
] | [] | 2022-10-14T21:25:23Z | 2022-10-15T02:34:37Z | 2022-10-15T02:34:37Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | brodieferguson | 15,876,037 | MDQ6VXNlcjE1ODc2MDM3 | User | false |
huggingface/diffusers | 2,350,170,059 | I_kwDOHa8MBc6MFL_L | 8,511 | https://github.com/huggingface/diffusers/issues/8511 | https://api.github.com/repos/huggingface/diffusers/issues/8511 | SD3 Dreambooth/Lora training scripts don't actually unload the text encoders | ### Describe the bug
In the SD3 Dreambooth training script https://github.com/huggingface/diffusers/blob/7f51f286a5397cb3e5c5a25693681aa4955e6241/examples/dreambooth/train_dreambooth_sd3.py#L1349 and similarly in the Lora training script https://github.com/huggingface/diffusers/blob/7f51f286a5397cb3e5c5a25693681aa4955... | closed | completed | false | 6 | [
"bug"
] | [] | 2024-06-13T05:14:18Z | 2024-06-16T19:52:34Z | 2024-06-16T19:52:34Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | spacepxl | 143,970,342 | U_kgDOCJTQJg | User | false |
huggingface/diffusers | 2,350,309,109 | I_kwDOHa8MBc6MFt71 | 8,514 | https://github.com/huggingface/diffusers/issues/8514 | https://api.github.com/repos/huggingface/diffusers/issues/8514 | EMAModel causes InstructP2P parallel fine-tuning error | ### Describe the bug
[The official given example] works well if I set `args.pretrained_model_name_or_path="runwayml/stable-diffusion-v1-5"`.
When I try to fine-tune instructpix2pix with `args.pretrained_model_name_or_path="timbrooks/instruct-pix2pix"`(https://github.com/huggingface/diffusers/blob/main/examples/in... | open | reopened | false | 2 | [
"bug",
"stale"
] | [] | 2024-06-13T06:52:49Z | 2024-09-14T15:09:30Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | liming-ai | 73,068,772 | MDQ6VXNlcjczMDY4Nzcy | User | false |
huggingface/diffusers | 2,350,312,615 | I_kwDOHa8MBc6MFuyn | 8,515 | https://github.com/huggingface/diffusers/issues/8515 | https://api.github.com/repos/huggingface/diffusers/issues/8515 | Duplicate code: SD3 JointAttnProcessor2_0 | **What API design would you like to have changed or added to the library? Why?**
SD3's JointAttnProcessor2_0. It has a duplicate code.
**What use case would this enable or better enable? Can you give us a code example?**
In 1135 of diffusers.models.attention_processor.py
https://github.com/huggingface/diffuse... | closed | completed | false | 0 | [] | [] | 2024-06-13T06:54:16Z | 2024-06-13T07:52:36Z | 2024-06-13T07:52:36Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Prgckwb | 55,102,558 | MDQ6VXNlcjU1MTAyNTU4 | User | false |
huggingface/diffusers | 2,350,376,944 | I_kwDOHa8MBc6MF-fw | 8,518 | https://github.com/huggingface/diffusers/issues/8518 | https://api.github.com/repos/huggingface/diffusers/issues/8518 | [SD3] [torch.compile] torch._dynamo.exc.Unsupported: call_function UserDefinedClassVariable() [] {'sample': TensorVariable()} | ### Describe the bug
I just try official example code using torch.compile to speed up SD3 inference, just encounter this error. From error message, the error seems in VAE side, and error disappear if I just delete `pipe.vae.decode = torch.compile(pipe.vae.decode, mode="max-autotune", fullgraph=True)` .
### Reproducti... | closed | completed | false | 9 | [
"bug"
] | [] | 2024-06-13T07:29:22Z | 2024-06-29T13:12:24Z | 2024-06-29T13:12:23Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luciennnnnnn | 20,135,317 | MDQ6VXNlcjIwMTM1MzE3 | User | false |
huggingface/diffusers | 1,409,987,905 | I_kwDOHa8MBc5UCrVB | 852 | https://github.com/huggingface/diffusers/issues/852 | https://api.github.com/repos/huggingface/diffusers/issues/852 | diffusers 0.5.1 crashs on MPS with unexpected keyword argument 'device_map' error | ### Describe the bug
Trying to run Stable Diffusion code with the default scheduler and it simply crashes out every single time with the following error:
```
Traceback (most recent call last):
File "/Users/fahim/Code/Python/sd-gui/test.py", line 6, in <module>
pipe = StableDiffusionPipeline.from_pretrained... | closed | completed | false | 5 | [
"bug"
] | [] | 2022-10-15T00:37:46Z | 2022-10-17T00:26:23Z | 2022-10-16T22:52:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | FahimF | 181,110 | MDQ6VXNlcjE4MTExMA== | User | false |
huggingface/diffusers | 2,350,578,869 | I_kwDOHa8MBc6MGvy1 | 8,523 | https://github.com/huggingface/diffusers/issues/8523 | https://api.github.com/repos/huggingface/diffusers/issues/8523 | playground cuda error | https://github.com/huggingface/diffusers/blob/896fb6d8d7c10001eb2a92568be7b4bd3d5ddea3/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_img2img.py#L728
When I initialize without using `to('cuda')`, the model exists in the cpu, so here self.device gets the handle on the cpu, refer to the code ... | open | null | false | 15 | [
"stale"
] | [] | 2024-06-13T08:54:32Z | 2024-09-14T15:09:27Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Honey-666 | 76,274,942 | MDQ6VXNlcjc2Mjc0OTQy | User | false |
huggingface/diffusers | 2,350,727,123 | I_kwDOHa8MBc6MHT_T | 8,526 | https://github.com/huggingface/diffusers/issues/8526 | https://api.github.com/repos/huggingface/diffusers/issues/8526 | SD3 - MPS - Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective. | ### Describe the bug
While executing the code I get the following remark for MPS:
Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective.
### Reproduction
import torch
from diffusers import StableDiffusion3Pipeline
pipe = StableDiffusion3Pipeline.from_pretrained("stabilityai... | closed | completed | false | 4 | [
"bug"
] | [] | 2024-06-13T10:02:41Z | 2024-07-01T11:14:06Z | 2024-07-01T11:14:06Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | TEGNoV | 32,719,570 | MDQ6VXNlcjMyNzE5NTcw | User | false |
huggingface/diffusers | 2,350,753,824 | I_kwDOHa8MBc6MHagg | 8,527 | https://github.com/huggingface/diffusers/issues/8527 | https://api.github.com/repos/huggingface/diffusers/issues/8527 | how to add controlnet in sd3! | I currently use inpainting controlnet in sdxl because it uses unet to easily support controlnet. And I am curious about how to add controlnet in sd3 with transforms model structure. | closed | completed | false | 6 | [] | [] | 2024-06-13T10:14:38Z | 2024-08-24T04:20:28Z | 2024-06-29T13:19:08Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | appleyang123 | 166,299,531 | U_kgDOCemHiw | User | false |
huggingface/diffusers | 2,350,875,636 | I_kwDOHa8MBc6MH4P0 | 8,529 | https://github.com/huggingface/diffusers/issues/8529 | https://api.github.com/repos/huggingface/diffusers/issues/8529 | sd3 clip-merge in diffusers | how to do clip merge in diffusers? | closed | completed | false | 4 | [] | [] | 2024-06-13T11:16:02Z | 2024-06-20T03:42:25Z | 2024-06-20T03:42:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | s9anus98a | 151,509,142 | U_kgDOCQfYlg | User | false |
huggingface/diffusers | 1,410,013,262 | I_kwDOHa8MBc5UCxhO | 853 | https://github.com/huggingface/diffusers/issues/853 | https://api.github.com/repos/huggingface/diffusers/issues/853 | if use 8gb vram train seting convert, Should ckpt conversion also be modified? | hi there, Thanks to you, I finished my 8gb vram training well
but I have one problem that Web UI is can`t load My train Model ckpt
I use this ckpt convert py
`https://gist.github.com/jachiam/8a5c0b607e38fcc585168b90c686eb05`
if use my train model Web UI I'm sending out a log like this
```
To create a public l... | closed | completed | false | 3 | [] | [] | 2022-10-15T02:20:28Z | 2022-10-27T06:42:42Z | 2022-10-27T06:42:42Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | A-Polyana | 58,426,735 | MDQ6VXNlcjU4NDI2NzM1 | User | false |
huggingface/diffusers | 2,350,929,316 | I_kwDOHa8MBc6MIFWk | 8,530 | https://github.com/huggingface/diffusers/issues/8530 | https://api.github.com/repos/huggingface/diffusers/issues/8530 | Conversion script for SD3 Diffusers Lora to original | Will you be providing a script to convert SD3 lora trained with diffusers to the original version, similar to convert_state_dict_to_kohya? Thanks. | closed | completed | false | 2 | [] | [] | 2024-06-13T11:43:27Z | 2024-06-29T13:19:46Z | 2024-06-29T13:19:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | alarazin | 60,730,378 | MDQ6VXNlcjYwNzMwMzc4 | User | false |
huggingface/diffusers | 2,351,089,510 | I_kwDOHa8MBc6MIsdm | 8,532 | https://github.com/huggingface/diffusers/issues/8532 | https://api.github.com/repos/huggingface/diffusers/issues/8532 | Progressive Memory Increase on loading lora weights | ### Describe the bug
When invoking pipeline.load_lora_weights(path), an incremental memory increase is observed over multiple iterations, indicating a potential memory management issue. Although pipeline.unload_lora_weights() is called following each load operation, it does not consistently return the memory usage to ... | closed | completed | false | 13 | [
"bug"
] | [] | 2024-06-13T12:58:46Z | 2024-10-16T02:33:57Z | 2024-10-16T02:33:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | danishboy000 | 35,116,832 | MDQ6VXNlcjM1MTE2ODMy | User | false |
huggingface/diffusers | 2,351,242,381 | I_kwDOHa8MBc6MJRyN | 8,534 | https://github.com/huggingface/diffusers/issues/8534 | https://api.github.com/repos/huggingface/diffusers/issues/8534 | [SD3] the training script of SD3 dreambooth has wrong logit-normal weighting | ### Describe the bug
I guess the implementation of logit-normal weighting is wrong https://github.com/huggingface/diffusers/blob/896fb6d8d7c10001eb2a92568be7b4bd3d5ddea3/examples/dreambooth/train_dreambooth_lora_sd3.py#L1490.
Consider that timestep is sampled uniformly first, weighting function should be a functi... | closed | completed | false | 3 | [
"bug"
] | [] | 2024-06-13T14:02:15Z | 2024-06-29T13:23:08Z | 2024-06-29T13:23:08Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luciennnnnnn | 20,135,317 | MDQ6VXNlcjIwMTM1MzE3 | User | false |
huggingface/diffusers | 2,351,423,089 | I_kwDOHa8MBc6MJ95x | 8,535 | https://github.com/huggingface/diffusers/issues/8535 | https://api.github.com/repos/huggingface/diffusers/issues/8535 | [SD3] pipe.enable_xformers_memory_efficient_attention | ### Describe the bug
> RuntimeError: The size of tensor a (154) must match the size of tensor b (2304) at non-singleton dimension 1
### Reproduction
```python
# StableDiffusion3Pipeline
pipe.enable_xformers_memory_efficient_attention()
```
### Logs
_No response_
### System Info
diffusers==0.29.0
python=3.10
... | closed | completed | false | 14 | [
"bug",
"stale"
] | [] | 2024-06-13T15:17:36Z | 2025-05-28T07:28:55Z | 2025-05-28T07:28:55Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CanvaChen | 18,375,123 | MDQ6VXNlcjE4Mzc1MTIz | User | false |
huggingface/diffusers | 2,351,486,254 | I_kwDOHa8MBc6MKNUu | 8,537 | https://github.com/huggingface/diffusers/issues/8537 | https://api.github.com/repos/huggingface/diffusers/issues/8537 | StableDiffusion3Pipeline.from_single_file error：module 'torch' has no attribute 'float8_e4m3fn' | ### Describe the bug
error below：
site-packages\safetensors\torch.py", line 313, in load_file
result[k] = f.get_tensor(k)
AttributeError: module 'torch' has no attribute 'float8_e4m3fn'
### Reproduction
import torch
from diffusers import StableDiffusion3Pipeline
```
pipe = StableDiffusion3Pipeline.from... | closed | completed | false | 9 | [
"bug"
] | [] | 2024-06-13T15:46:09Z | 2024-07-25T14:16:40Z | 2024-07-01T11:01:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhaoyun0071 | 35,762,050 | MDQ6VXNlcjM1NzYyMDUw | User | false |
huggingface/diffusers | 2,351,590,848 | I_kwDOHa8MBc6MKm3A | 8,539 | https://github.com/huggingface/diffusers/issues/8539 | https://api.github.com/repos/huggingface/diffusers/issues/8539 | SD3 SD3Transformer2DModel issues with setting cross-attention | ### Describe the bug
calling `set_attn_processor(attention)` using `AttnProcessor` or `AttnProcessor2_0`
on `SD3Transformer2DModel` executes without issues, but results in runtime error during inference:
### Reproduction
```py
def set_diffusers_attention(pipe):
def set_attn(pipe, attention):
... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-06-13T16:38:25Z | 2024-06-13T17:26:45Z | 2024-06-13T17:26:45Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 1,410,226,911 | I_kwDOHa8MBc5UDlrf | 854 | https://github.com/huggingface/diffusers/issues/854 | https://api.github.com/repos/huggingface/diffusers/issues/854 | Community pipeline to run Stable Diffusion with k-diffusion | **Describe the solution you'd like**
I'm making a pipeline to use k-diffusion with Diffusers.
The code run from start to finish, but it's giving pretty bad outputs.
I did try to undestand both codes and I'm currently stuck in what could be possible wrong. Perhaps someone can shed some light of what is wrong with my... | closed | completed | false | 6 | [
"stale"
] | [] | 2022-10-15T16:53:29Z | 2022-11-23T15:03:18Z | 2022-11-23T15:03:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | BurguerJohn | 32,472,975 | MDQ6VXNlcjMyNDcyOTc1 | User | false |
huggingface/diffusers | 2,351,827,512 | I_kwDOHa8MBc6MLgo4 | 8,540 | https://github.com/huggingface/diffusers/issues/8540 | https://api.github.com/repos/huggingface/diffusers/issues/8540 | SD3 dreambooth training - caching of embeddings and vae representations | **Is your feature request related to a problem? Please describe.**
The current implementation of the dreambooth trainings for both loras and finetuning is very memory intensive.
**Describe the solution you'd like.**
I would like the option to pre-cache the VAE representations of the images and the text encoder rep... | open | null | false | 2 | [
"stale"
] | [] | 2024-06-13T18:42:26Z | 2024-09-14T15:09:18Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | GavChap | 1,280,021 | MDQ6VXNlcjEyODAwMjE= | User | false |
huggingface/diffusers | 2,351,910,155 | I_kwDOHa8MBc6ML00L | 8,541 | https://github.com/huggingface/diffusers/issues/8541 | https://api.github.com/repos/huggingface/diffusers/issues/8541 | SD3: Repository Not Found for url: https://huggingface.co/api/models/stabilityai/stable-diffusion-3-medium-diffusers-diffusers | ### Describe the bug
It used to work, and now it doesn't work anymore?
SD3: Repository Not Found for url: https://huggingface.co/api/models/stabilityai/stable-diffusion-3-medium-diffusers-diffusers
### Reproduction
```
import torch
from huggingface_hub.commands.user import lo... | closed | completed | false | 1 | [
"bug"
] | [] | 2024-06-13T19:31:35Z | 2024-06-13T19:32:21Z | 2024-06-13T19:32:20Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | tin2tin | 1,322,593 | MDQ6VXNlcjEzMjI1OTM= | User | false |
huggingface/diffusers | 2,351,983,549 | I_kwDOHa8MBc6MMGu9 | 8,543 | https://github.com/huggingface/diffusers/issues/8543 | https://api.github.com/repos/huggingface/diffusers/issues/8543 | SD3 dreambooth example - gradient checkpointing fails | ### Describe the bug
When using the SD3 dreambooth LORA script with all default arguments except for gradient checkpointing, the script fails with an error that a tensor is expected, not a tuple. I have raised a PR to fix this:
https://github.com/huggingface/diffusers/pull/8542
### Reproduction
SD3 dreambooth LO... | closed | completed | false | 1 | [
"bug"
] | [] | 2024-06-13T20:19:40Z | 2024-06-13T21:05:28Z | 2024-06-13T21:05:28Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | RefractAI | 140,120,812 | U_kgDOCFoS7A | User | false |
huggingface/diffusers | 2,352,096,569 | I_kwDOHa8MBc6MMiU5 | 8,545 | https://github.com/huggingface/diffusers/issues/8545 | https://api.github.com/repos/huggingface/diffusers/issues/8545 | SD3 latents error 'tuple' object has no attribute 'to' issue on stable_diffusion_3/pipeline_stable_diffusion_3.py | script to produce:
```
image = pipe(prompt="", negative_prompt="", guidance_scale=1.,
num_inference_steps=20, latents=inv_latents)
```
inv_latents value print:
```
(tensor([[[[-0.8428, -0.8203, -0.8325, ..., -0.8325, -0.8706, -0.8521],
[-0.8511, -0.7632, -0.8467, ..., -... | open | null | false | 6 | [
"stale"
] | [] | 2024-06-13T21:29:14Z | 2024-09-14T15:09:15Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | s9anus98a | 151,509,142 | U_kgDOCQfYlg | User | false |
huggingface/diffusers | 2,352,167,241 | I_kwDOHa8MBc6MMzlJ | 8,546 | https://github.com/huggingface/diffusers/issues/8546 | https://api.github.com/repos/huggingface/diffusers/issues/8546 | SD3 missing support from-single-file | ### Describe the bug
`StableDiffusion3Pipeline` does implement `from_single_file` which correctly loads DiT and VAE.
however, it fails to deal with any of the text-encoders: TE1, TE2 and TE3.
- if loading `sd3_medium.safetensors` that is understandable as that model does have any TEs baked in
- if loading `sd3_... | closed | completed | false | 9 | [
"bug"
] | [] | 2024-06-13T22:35:24Z | 2024-06-19T18:23:37Z | 2024-06-19T18:23:37Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 2,352,400,459 | I_kwDOHa8MBc6MNshL | 8,547 | https://github.com/huggingface/diffusers/issues/8547 | https://api.github.com/repos/huggingface/diffusers/issues/8547 | Latent Consistency Model Possible Scaling Bug | ### Describe the bug
The LCM scheduler and LCM training scripts use the following formula for the $`c_{\mbox{out}}(t)`$ scaling (ignoring timestep scaling):
```math
c_{\mbox{out}}(t) = \frac{t}{\sqrt{t^2 + \sigma_{\mbox{data}}^2}}
```
However, the [original consistency models paper](https://arxiv.org/pdf/2303.... | closed | completed | false | 6 | [
"bug"
] | [] | 2024-06-14T03:13:46Z | 2024-10-30T09:22:04Z | 2024-06-14T05:17:22Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | dg845 | 58,458,699 | MDQ6VXNlcjU4NDU4Njk5 | User | false |
huggingface/diffusers | 2,352,423,234 | I_kwDOHa8MBc6MNyFC | 8,549 | https://github.com/huggingface/diffusers/issues/8549 | https://api.github.com/repos/huggingface/diffusers/issues/8549 | [SD3] Incorrect stochastic sampling implementation | ### Describe the bug
Ref to Algorithm 2 of [EDM](https://arxiv.org/abs/2206.00364), for a given sample $x_t$, noise is introduced to it reaching a higher noise level $\hat{t}$, then we evaluate network with $\hat{x_t}$, $\hat{t}$ as input. However, the current implementation evaluates network with $x_t$, $t$ as inpu... | open | null | false | 5 | [
"bug",
"stale"
] | [] | 2024-06-14T03:40:44Z | 2024-09-14T15:09:12Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luciennnnnnn | 20,135,317 | MDQ6VXNlcjIwMTM1MzE3 | User | false |
huggingface/diffusers | 2,352,520,901 | I_kwDOHa8MBc6MOJ7F | 8,551 | https://github.com/huggingface/diffusers/issues/8551 | https://api.github.com/repos/huggingface/diffusers/issues/8551 | StableDiffusionXLInpaintPipeline: image latent not transferred to correct device when using `device_map="balanced"` on VAE model | ### Describe the bug
Running StableDiffusionXLInpaintPipeline on multiple GPUs with `device_map="balanced"` on the VAE model, I got the following error (irrelevant part excluded)
```
File "/home1/uname/.cache/pypoetry/virtualenvs/project-350heM32-py3.11/lib/python3.11/site-packages/diffusers/pipelines/stable_dif... | open | null | false | 2 | [
"bug",
"stale"
] | [] | 2024-06-14T05:10:45Z | 2024-09-14T15:09:09Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | nghiatt90 | 3,022,104 | MDQ6VXNlcjMwMjIxMDQ= | User | false |
huggingface/diffusers | 2,352,621,559 | I_kwDOHa8MBc6MOif3 | 8,552 | https://github.com/huggingface/diffusers/issues/8552 | https://api.github.com/repos/huggingface/diffusers/issues/8552 | Segmentation fault (core dump) in the TextToVideoZeroPipeline inference | ### Describe the bug
Segmentation fault (core dump) in the TextToVideoZeroPipeline inference
### Reproduction
Following https://huggingface.co/docs/diffusers/v0.29.0/en/api/pipelines/text_to_video_zero#usage-example
```
import torch
from diffusers import TextToVideoZeroPipeline
model_id = "runwayml/stable-diff... | closed | completed | false | 4 | [
"bug"
] | [] | 2024-06-14T06:27:40Z | 2024-11-08T06:41:46Z | 2024-07-05T04:56:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | hyunW3 | 26,101,356 | MDQ6VXNlcjI2MTAxMzU2 | User | false |
huggingface/diffusers | 2,352,664,468 | I_kwDOHa8MBc6MOs-U | 8,553 | https://github.com/huggingface/diffusers/issues/8553 | https://api.github.com/repos/huggingface/diffusers/issues/8553 | When will train_text_to_image_lora_sd3 release? | Thank you so much for your excellent work and selfless open source spirit. I wonder that when will train_text_to_image_lora_sd3 release? | closed | completed | false | 5 | [] | [] | 2024-06-14T06:55:13Z | 2024-06-26T01:37:58Z | 2024-06-18T15:27:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Rody1911641 | 73,045,959 | MDQ6VXNlcjczMDQ1OTU5 | User | false |
huggingface/diffusers | 2,352,974,106 | I_kwDOHa8MBc6MP4ka | 8,555 | https://github.com/huggingface/diffusers/issues/8555 | https://api.github.com/repos/huggingface/diffusers/issues/8555 | [SD3] [Feature Request] FP8 and Availability of Text Encoder training | **Is your feature request related to a problem? Please describe.**
Current training scripts use a significant amount of VRAM. Since the T5 text encoder takes up most of it, I would like to be able to choose whether or not to learn here.
Also, an option to load models in FP8 and reduce VRAM usage would be appreciat... | closed | completed | false | 2 | [] | [] | 2024-06-14T09:41:50Z | 2024-06-29T13:14:36Z | 2024-06-29T13:14:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | turbo-boo | 59,400,127 | MDQ6VXNlcjU5NDAwMTI3 | User | false |
huggingface/diffusers | 2,352,982,979 | I_kwDOHa8MBc6MP6vD | 8,556 | https://github.com/huggingface/diffusers/issues/8556 | https://api.github.com/repos/huggingface/diffusers/issues/8556 | SD3 from_single_file get error：Cannot copy out of meta tensor; no data! | ### Describe the bug
Loading pipeline components...: 100%|██████████| 9/9 [00:01<00:00, 5.38it/s]
Traceback (most recent call last):
site-packages\torch\nn\modules\module.py", line 1143, in convert
return t.to(device, dtype if t.is_floating_point() or t.is_complex() else None, , non_blocking)
NotImplemented... | closed | completed | false | 4 | [
"bug"
] | [] | 2024-06-14T09:46:42Z | 2024-06-17T08:26:14Z | 2024-06-17T08:26:14Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zhaoyun0071 | 35,762,050 | MDQ6VXNlcjM1NzYyMDUw | User | false |
huggingface/diffusers | 2,353,226,865 | I_kwDOHa8MBc6MQ2Rx | 8,557 | https://github.com/huggingface/diffusers/issues/8557 | https://api.github.com/repos/huggingface/diffusers/issues/8557 | [SD3] Token limit in training | ### Describe the bug
I'm aware there is a PR #8506 for token limit in inference, however, the token limit in training is not processed. I guess token limit in training will affect performance.
### Reproduction
no
### Logs
_No response_
### System Info
no
### Who can help?
@yiyixuxu @asomoza | closed | completed | false | 7 | [
"bug"
] | [] | 2024-06-14T12:04:56Z | 2024-06-17T20:32:58Z | 2024-06-17T20:32:58Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luciennnnnnn | 20,135,317 | MDQ6VXNlcjIwMTM1MzE3 | User | false |
huggingface/diffusers | 1,410,249,944 | I_kwDOHa8MBc5UDrTY | 856 | https://github.com/huggingface/diffusers/issues/856 | https://api.github.com/repos/huggingface/diffusers/issues/856 | JAX / Flax pipeline: "Ran out of memory in memory space hbm." on TPU v2-8 | ### Describe the bug
I followed the blog post [here](https://huggingface.co/blog/stable_diffusion_jax) on `diffusers==0.5.1`. I am running inside a Colab TPU instance so I added [this snippet](https://github.com/google/jax#pip-installation-colab-tpu):
```py
import jax.tools.colab_tpu
jax.tools.colab_tpu.setup_tpu... | closed | completed | false | 2 | [
"bug"
] | [] | 2022-10-15T18:16:50Z | 2022-10-15T21:27:36Z | 2022-10-15T21:27:36Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | skirsten | 1,972,314 | MDQ6VXNlcjE5NzIzMTQ= | User | false |
huggingface/diffusers | 2,353,507,050 | I_kwDOHa8MBc6MR6rq | 8,560 | https://github.com/huggingface/diffusers/issues/8560 | https://api.github.com/repos/huggingface/diffusers/issues/8560 | SD3 training OOM on 4090 | torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 48.00 MiB. GPU 0 has a total capacty of 23.64 GiB of which 9.69 MiB is free. Process 20296 has 23.63 GiB memory in use. Of the allocated memory 22.81 GiB is allocated by PyTorch, and 362.31 MiB is reserved by PyTorch but unallocated. If reserved but una... | closed | completed | false | 23 | [] | [] | 2024-06-14T14:29:54Z | 2024-06-29T13:17:23Z | 2024-06-29T13:17:23Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | inspire-boy | 112,338,987 | U_kgDOBrIoKw | User | false |
huggingface/diffusers | 2,354,269,683 | I_kwDOHa8MBc6MU03z | 8,565 | https://github.com/huggingface/diffusers/issues/8565 | https://api.github.com/repos/huggingface/diffusers/issues/8565 | StableDiffusion3Pipeline `load_lora_weights` IndexError | ### Describe the bug
Using `load_lora_weights` with a single safetensor on StableDiffusion3Pipeline results in an IndexError
### Reproduction
StableDiffusion3Pipeline fails at `load_lora_weights` with IndexError on latest diffusers master
```python
import torch
from diffusers import StableDiffusion3Pipeline... | closed | completed | false | 5 | [
"bug"
] | [] | 2024-06-14T23:54:05Z | 2024-06-21T23:02:16Z | 2024-06-21T23:02:16Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Beinsezii | 39,478,211 | MDQ6VXNlcjM5NDc4MjEx | User | false |
huggingface/diffusers | 2,354,538,198 | I_kwDOHa8MBc6MV2bW | 8,567 | https://github.com/huggingface/diffusers/issues/8567 | https://api.github.com/repos/huggingface/diffusers/issues/8567 | Is the training objective for SD3 correct? | ### Describe the bug
The input to the `v` term is `noisy_model_input = sigmas * noise + (1.0 - sigmas) * model_input`, but by the [paper](https://arxiv.org/abs/2209.03003):
- `v(tx_... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-06-15T04:35:27Z | 2024-06-15T12:47:42Z | 2024-06-15T12:47:42Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | AmericanPresidentJimmyCarter | 110,263,573 | U_kgDOBpJ9FQ | User | false |
huggingface/diffusers | 2,354,823,757 | I_kwDOHa8MBc6MW8JN | 8,576 | https://github.com/huggingface/diffusers/issues/8576 | https://api.github.com/repos/huggingface/diffusers/issues/8576 | Tired tired of crash, random crash, random crash | ### Describe the bug
I'm training Lora sinc e approximately 1 year, I have three computers each with 1000Watts PSU and diffferent CPU (Intel and AMD) and rtx 3090 of different companies and between 64Gb and 128Gb of ram.
I'm getting random crash while training so often, like 60% of time is spent retraining because o... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-06-15T12:31:10Z | 2024-06-29T13:17:36Z | 2024-06-29T13:17:36Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | axel578 | 44,734,791 | MDQ6VXNlcjQ0NzM0Nzkx | User | false |
huggingface/diffusers | 2,354,855,041 | I_kwDOHa8MBc6MXDyB | 8,577 | https://github.com/huggingface/diffusers/issues/8577 | https://api.github.com/repos/huggingface/diffusers/issues/8577 | Differential Diffusion with SD3 | **Is your feature request related to a problem? Please describe.**
One might wish to soft inpaint or edit pictures with differential diffusion with SD3.
**Describe the solution you'd like.**
I believe we can port diff-diff to SD3, by adding a new pipeline for it.
**Describe alternatives you've considered.**
Di... | closed | completed | false | 9 | [] | [] | 2024-06-15T13:27:39Z | 2024-06-29T03:12:40Z | 2024-06-29T03:12:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | exx8 | 8,540,180 | MDQ6VXNlcjg1NDAxODA= | User | false |
huggingface/diffusers | 2,355,027,846 | I_kwDOHa8MBc6MXt-G | 8,578 | https://github.com/huggingface/diffusers/issues/8578 | https://api.github.com/repos/huggingface/diffusers/issues/8578 | Cannot copy out of meta tensor; no data! with sd_turbo 1.5 | ### Describe the bug
Error on loading from single file. Might be related to this one https://huggingface.co/stabilityai/sd-turbo/discussions/1
### Reproduction
```
StableDiffusionPipeline.from_single_file(
'sd_turbo.safetensors',
cache_dir = 'models/checkpoints',
... | closed | completed | false | 3 | [
"bug"
] | [] | 2024-06-15T16:54:04Z | 2024-06-18T15:39:32Z | 2024-06-18T15:39:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | olegchomp | 11,017,531 | MDQ6VXNlcjExMDE3NTMx | User | false |
huggingface/diffusers | 2,355,255,958 | I_kwDOHa8MBc6MYlqW | 8,579 | https://github.com/huggingface/diffusers/issues/8579 | https://api.github.com/repos/huggingface/diffusers/issues/8579 | SD3 lora support non functional | ### Describe the bug
PR #8483 added support for SD3 lora loading, but it does so a) without any checks, b) documentation on required params.
For SD3, lora_lora_weights triggers new method `load_lora_into_transformer` which expects to extract values from `kwargs` and without having default values.
Documentation... | closed | completed | false | 11 | [
"bug"
] | [] | 2024-06-15T20:59:49Z | 2024-06-21T15:47:35Z | 2024-06-21T15:47:35Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | vladmandic | 57,876,960 | MDQ6VXNlcjU3ODc2OTYw | User | false |
huggingface/diffusers | 1,410,430,681 | I_kwDOHa8MBc5UEXbZ | 858 | https://github.com/huggingface/diffusers/issues/858 | https://api.github.com/repos/huggingface/diffusers/issues/858 | AttributeError: 'Parameter' object has no attribute 'named_children' | ### Describe the bug
Hi, friends, I meet a problem I hope to get your help.
When I run the code as follow:
`from diffusers import StableDiffusionPipeline
import torch
pipe = StableDiffusionPipeline.from_pretrained("PATH_OF_MODEL/stable-diffusion",device_map="auto",torch_type=torch.float16, revision="fp16")
... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2022-10-16T09:30:07Z | 2022-12-12T15:04:22Z | 2022-12-12T15:04:22Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | klyuhang9 | 39,409,233 | MDQ6VXNlcjM5NDA5MjMz | User | false |
huggingface/diffusers | 2,355,463,916 | I_kwDOHa8MBc6MZYbs | 8,581 | https://github.com/huggingface/diffusers/issues/8581 | https://api.github.com/repos/huggingface/diffusers/issues/8581 | ValueError: Checkpoint not supported because layer lora_unet_label_emb_0_0.alpha not supported. | script to produce error:
```
import torch
from diffusers import StableDiffusionXLPipeline
pipe = StableDiffusionXLPipeline.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16,
).to('cuda')
pipe.load_lora_weights("ostris/sdxl-sd1-vae-lora")
```
| closed | completed | false | 9 | [] | [] | 2024-06-16T06:17:22Z | 2025-11-04T02:57:00Z | 2024-06-16T17:20:48Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | s9anus98a | 151,509,142 | U_kgDOCQfYlg | User | false |
huggingface/diffusers | 2,355,839,589 | I_kwDOHa8MBc6Ma0Jl | 8,588 | https://github.com/huggingface/diffusers/issues/8588 | https://api.github.com/repos/huggingface/diffusers/issues/8588 | SD3 diffusers/scripts/convert_diffusers_to_original_sd3.py is missing | ### Describe the bug
The repository is missing conversion from diffusers to original for SD3: diffusers/scripts/convert_diffusers_to_original_sd3.py
### Reproduction
The repository is missing conversion from diffusers to original for SD3: diffusers/scripts/convert_diffusers_to_original_sd3.py
### Logs
_No response... | closed | completed | false | 5 | [
"bug"
] | [] | 2024-06-16T15:29:37Z | 2024-08-01T03:57:23Z | 2024-06-26T00:24:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | a-l-e-x-d-s-9 | 6,518,064 | MDQ6VXNlcjY1MTgwNjQ= | User | false |
huggingface/diffusers | 2,355,842,569 | I_kwDOHa8MBc6Ma04J | 8,589 | https://github.com/huggingface/diffusers/issues/8589 | https://api.github.com/repos/huggingface/diffusers/issues/8589 | [Bug] SD3 DreamBooth Training Unable to Use Prior Preservation | ### Describe the bug
Error will happen if `--with_prior_preservation` is used.
### Reproduction
```
#!/bin/sh
MODEL_NAME="stabilityai/stable-diffusion-3-medium-diffusers"
accelerate launch train_dreambooth_lora_sd3.py \
--pretrained_model_name_or_path=$MODEL_NAME \
--output_dir "./xxx-lora/sd3-lora/model... | closed | completed | false | 11 | [
"bug"
] | [] | 2024-06-16T15:36:51Z | 2024-06-29T13:18:34Z | 2024-06-29T13:18:33Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | DefinitlyEvil | 8,604,063 | MDQ6VXNlcjg2MDQwNjM= | User | false |
huggingface/diffusers | 1,410,435,125 | I_kwDOHa8MBc5UEYg1 | 859 | https://github.com/huggingface/diffusers/issues/859 | https://api.github.com/repos/huggingface/diffusers/issues/859 | TypeError: __init__() got an unexpected keyword argument 'device_map' | ### Describe the bug
Hi, friends, I meet a problem I hope to get your help.
When I run the code as follow:
`from diffusers import StableDiffusionPipeline
import torch
pipe = StableDiffusionPipeline.from_pretrained("PATH_OF_MODEL/stable-diffusion",device_map="auto",torch_type=torch.float16, revision="fp16")
... | closed | completed | false | 3 | [
"bug"
] | [] | 2022-10-16T09:49:17Z | 2022-10-16T22:52:18Z | 2022-10-16T22:52:18Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | klyuhang9 | 39,409,233 | MDQ6VXNlcjM5NDA5MjMz | User | false |
huggingface/diffusers | 2,356,010,827 | I_kwDOHa8MBc6Mbd9L | 8,593 | https://github.com/huggingface/diffusers/issues/8593 | https://api.github.com/repos/huggingface/diffusers/issues/8593 | SD3 model generates noise on the bfloat16 datatype on CPU. | ### Describe the bug
SD3 model generates noise on the bfloat16 datatype on CPU.
### Reproduction
```
import torch
from diffusers import StableDiffusion3Pipeline
pipe = StableDiffusion3Pipeline.from_pretrained("stabilityai/stable-diffusion-3-medium-diffusers", torch_dtype=torch.bfloat16, token=token)
pipe... | closed | completed | false | 7 | [
"bug"
] | [] | 2024-06-16T21:56:22Z | 2024-06-20T22:04:41Z | 2024-06-20T22:04:40Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | 0x6b64 | 105,229,704 | U_kgDOBkWtiA | User | false |
huggingface/diffusers | 2,356,058,289 | I_kwDOHa8MBc6Mbpix | 8,594 | https://github.com/huggingface/diffusers/issues/8594 | https://api.github.com/repos/huggingface/diffusers/issues/8594 | UnboundLocalError: tokenizer_one referenced before assignment in train_dreambooth_lora_sd3.py | ### Describe the bug
I encountered an issue when trying to run train_dreambooth_lora_sd3.py. The script throws an UnboundLocalError because tokenizer_one is referenced before it is assigned. This prevents the training process from starting.
### Reproduction
Clone the repository.
Set up the environment as per the in... | closed | completed | false | 3 | [
"bug"
] | [] | 2024-06-16T23:53:24Z | 2024-06-20T12:12:25Z | 2024-06-20T12:12:25Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | C0nsumption | 98,207,055 | U_kgDOBdqFTw | User | false |
huggingface/diffusers | 2,356,423,548 | I_kwDOHa8MBc6MdCt8 | 8,596 | https://github.com/huggingface/diffusers/issues/8596 | https://api.github.com/repos/huggingface/diffusers/issues/8596 | Controlnet training method for playgroundV2.5 model | When I try to train a controlnet model for playgroundV2.5, but there were some issues happened.
The general training code in diffusers for controlnet sdxl is not fit for EDM model, training is ok, loss looks ok, but the result is broke, after 400 steps the image looks like snow pieces. although I tied to set the sched... | closed | completed | true | 0 | [] | [] | 2024-06-17T05:33:14Z | 2024-06-18T15:19:03Z | 2024-06-18T15:19:03Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | baizh0u | 26,198,796 | MDQ6VXNlcjI2MTk4Nzk2 | User | false |
huggingface/diffusers | 2,356,606,071 | I_kwDOHa8MBc6MdvR3 | 8,599 | https://github.com/huggingface/diffusers/issues/8599 | https://api.github.com/repos/huggingface/diffusers/issues/8599 | Nan appears during sd3 training | ### Describe the bug
I try to use the train_dreambooth_sd3.py for training. The loss and results were normal at first, but after two or three thousand steps, Nan appeared. Has anyone had the same problem? I'm training with fp16 precision, but setting the accuracy of VAE to fp32 as per the training file
### Reproduct... | closed | completed | false | 21 | [
"bug"
] | [] | 2024-06-17T07:30:59Z | 2024-10-25T08:54:57Z | 2024-06-29T13:13:12Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | jyy-1998 | 55,531,840 | MDQ6VXNlcjU1NTMxODQw | User | false |
huggingface/diffusers | 2,357,443,101 | I_kwDOHa8MBc6Mg7od | 8,604 | https://github.com/huggingface/diffusers/issues/8604 | https://api.github.com/repos/huggingface/diffusers/issues/8604 | [BUG] convert T5 text encoder to float16 results corrupted image | ### Describe the bug
I have tested PixArt-Sigma with following code, where I load text_encoder separately since I will fine-tune it in later. I found `T5EncoderModel.from_pretrained(torch_dtype=torch.float16)` is very different from `T5EncoderModel.from_pretrained().to(dtype=torch.float16)`, the later one produces c... | open | null | false | 15 | [
"bug",
"stale"
] | [] | 2024-06-17T14:05:38Z | 2024-12-14T15:05:22Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luciennnnnnn | 20,135,317 | MDQ6VXNlcjIwMTM1MzE3 | User | false |
huggingface/diffusers | 2,357,747,217 | I_kwDOHa8MBc6MiF4R | 8,605 | https://github.com/huggingface/diffusers/issues/8605 | https://api.github.com/repos/huggingface/diffusers/issues/8605 | global, eager model weight GPU unloading | **What API design would you like to have changed or added to the library? Why?**
Most people expect `diffusers` and `transformers` "models" to be "unloaded" so that they can "just" "run" a "big" "pipeline" using their "VRAM" so that it "fits."
In other words, author a mixin that keeps track of *all* weights in Hu... | open | null | false | 5 | [
"stale"
] | [] | 2024-06-17T16:25:08Z | 2024-12-13T15:05:54Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | doctorpangloss | 2,229,300 | MDQ6VXNlcjIyMjkzMDA= | User | false |
huggingface/diffusers | 2,358,766,263 | I_kwDOHa8MBc6Ml-q3 | 8,609 | https://github.com/huggingface/diffusers/issues/8609 | https://api.github.com/repos/huggingface/diffusers/issues/8609 | can not use custom_pipeline hd_painter with IP_adapter | ### Describe the bug
When I use IP_adapter and hd_painter at the same time.
it pop out the RuntimeError: mat1 and mat2 shapes cannot be multiplied (514x1280 and 1024x3072).
It should be successful since both IP_adapter and hd_painter didn't modify the output shape of attention layers.
From log, it seems that the ... | open | null | false | 6 | [
"bug",
"stale"
] | [] | 2024-06-18T03:04:03Z | 2024-09-14T15:08:56Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | linthmwan | 114,485,232 | U_kgDOBtLn8A | User | false |
huggingface/diffusers | 1,410,695,033 | I_kwDOHa8MBc5UFX95 | 861 | https://github.com/huggingface/diffusers/issues/861 | https://api.github.com/repos/huggingface/diffusers/issues/861 | Can't convert ckpt model to diffusers with convert_original_stable_diffusion_to_diffusers | ### Describe the bug
When I use the script convert_original_stable_diffusion_to_diffusers
with the command
python Output.py --checkpoint_path model.ckpt --dump_path hq-full/ --original_config_file config.yaml
It shows me
Traceback (most recent call last):
File "Output.py", line 643, in <module>
checkpoi... | closed | completed | false | 1 | [
"bug"
] | [] | 2022-10-17T00:37:17Z | 2022-10-22T16:21:19Z | 2022-10-22T16:21:19Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | izumedonabe | 42,296,908 | MDQ6VXNlcjQyMjk2OTA4 | User | false |
huggingface/diffusers | 2,359,126,966 | I_kwDOHa8MBc6MnWu2 | 8,612 | https://github.com/huggingface/diffusers/issues/8612 | https://api.github.com/repos/huggingface/diffusers/issues/8612 | Pixart-sigma is incompatible with mixed precision inference | ### Describe the bug
I found Pixart-Sigma is incompatible with mixed precision inference, loading models both in `float32` and in `float16` have similar problem.
I guess this problem may have some relationship with #8604
### Reproduction
```python
from diffusers import PixArtSigmaPipeline
import torch
fr... | closed | completed | false | 5 | [
"bug",
"stale"
] | [] | 2024-06-18T07:19:15Z | 2024-09-23T11:33:25Z | 2024-09-23T11:33:24Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Luciennnnnnn | 20,135,317 | MDQ6VXNlcjIwMTM1MzE3 | User | false |
huggingface/diffusers | 2,359,143,432 | I_kwDOHa8MBc6MnawI | 8,613 | https://github.com/huggingface/diffusers/issues/8613 | https://api.github.com/repos/huggingface/diffusers/issues/8613 | Getting OOM error when using "--caption_column="caption"". | ### Describe the bug
DREAMBOOTH ERROR:
torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 28.00 MiB. GPU |
DREAMBOOTH LORA ERROR:
torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 12.00 MiB. GPU
in dreambooth lora training i get the error on validation.
### Reproduction
PAR... | closed | completed | false | 10 | [
"bug"
] | [] | 2024-06-18T07:28:15Z | 2024-06-29T13:10:03Z | 2024-06-29T13:10:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | abhicake2048 | 74,735,261 | MDQ6VXNlcjc0NzM1MjYx | User | false |
huggingface/diffusers | 2,359,556,649 | I_kwDOHa8MBc6Mo_op | 8,614 | https://github.com/huggingface/diffusers/issues/8614 | https://api.github.com/repos/huggingface/diffusers/issues/8614 | controlnet singlefile dont have config.json | ```py
controlnets = ControlNetModel.from_single_file(zzzz/diffusers_xl_depth_full.safetensors, torch_dtype=torch.float16)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "Z:\software\p... | closed | completed | false | 15 | [] | [] | 2024-06-18T10:50:13Z | 2024-06-21T13:15:49Z | 2024-06-18T11:51:46Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xalteropsx | 103,671,642 | U_kgDOBi3nWg | User | false |
huggingface/diffusers | 2,359,619,937 | I_kwDOHa8MBc6MpPFh | 8,615 | https://github.com/huggingface/diffusers/issues/8615 | https://api.github.com/repos/huggingface/diffusers/issues/8615 | Support `fuse_lora` on Stable Diffusion 3 | **Is your feature request related to a problem? Please describe.**
Currently Stable DIffusion 3 errors with `fuse_lora`
```
AttributeError: 'StableDiffusion3Pipeline' object has no attribute 'fuse_lora'
```
**Describe the solution you'd like.**
I would like that fusing LoRAs is supported by SD3
**Additional... | closed | completed | false | 0 | [] | [] | 2024-06-18T11:24:29Z | 2024-06-20T08:55:53Z | 2024-06-20T08:55:53Z | MEMBER | null | 20260407T133413Z | 2026-04-07T13:34:13Z | apolinario | 788,417 | MDQ6VXNlcjc4ODQxNw== | User | false |
huggingface/diffusers | 2,359,958,185 | I_kwDOHa8MBc6Mqhqp | 8,617 | https://github.com/huggingface/diffusers/issues/8617 | https://api.github.com/repos/huggingface/diffusers/issues/8617 | Classifier free guidance(CFG) on different prediction types and karras style schedulers | There is banch of repos [1](https://github.com/duxiaodan/intrinsic-lora/blob/b69770c53035b936eae3132354afb1e14d2043ea/rescale_cfg_pipeline_forward.py#L259), [2](https://github.com/Amblyopius/Stable-Diffusion-ONNX-FP16/blob/55d050957f70c90f14aed13d9854f7aa59ff1f70/pipeline_onnx_stable_diffusion_instruct_pix2pix.py#L361)... | open | null | false | 3 | [
"stale"
] | [] | 2024-06-18T14:05:48Z | 2024-09-14T15:08:50Z | null | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | maxs-kan | 50,169,754 | MDQ6VXNlcjUwMTY5NzU0 | User | false |
huggingface/diffusers | 2,360,102,367 | I_kwDOHa8MBc6MrE3f | 8,619 | https://github.com/huggingface/diffusers/issues/8619 | https://api.github.com/repos/huggingface/diffusers/issues/8619 | High Batch Size with SD3 Dreambooth Destabilizes Training | ### Describe the bug
I have been trying to train a slightly modified IP Adapter architecture for SD3 over the past few days and wrote the training script by copying the up to date weighting, noise and loss code from the train_dreamboothsd3.py script, while training nothing I could do would allow the model to train, af... | closed | completed | true | 1 | [
"bug"
] | [] | 2024-06-18T15:09:23Z | 2024-06-18T15:31:05Z | 2024-06-18T15:31:05Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | CiaraStrawberry | 13,116,982 | MDQ6VXNlcjEzMTE2OTgy | User | false |
huggingface/diffusers | 2,360,317,388 | I_kwDOHa8MBc6Mr5XM | 8,624 | https://github.com/huggingface/diffusers/issues/8624 | https://api.github.com/repos/huggingface/diffusers/issues/8624 | Running stable diffusion 3 medium : fused_layer_norm_cuda.cpython-310-x86_64-linux-gnu.so: undefined symbol | ### Describe the bug
Running diffusers version 0.29
```import torch
from diffusers import StableDiffusion3Pipeline
pipe = StableDiffusion3Pipeline.from_pretrained("stabilityai/stable-diffusion-3-medium-diffusers", torch_dtype=torch.float16)
```
immediately throws error:
ImportError: /usr/local/lib/python3.... | closed | completed | false | 2 | [
"bug"
] | [] | 2024-06-18T17:02:44Z | 2024-06-20T11:04:58Z | 2024-06-20T11:04:58Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | zagglez | 61,600,248 | MDQ6VXNlcjYxNjAwMjQ4 | User | false |
huggingface/diffusers | 2,360,420,373 | I_kwDOHa8MBc6MsSgV | 8,626 | https://github.com/huggingface/diffusers/issues/8626 | https://api.github.com/repos/huggingface/diffusers/issues/8626 | More thorough guidance for multiple IP adapter images/masks and a single IP Adapter | ### Describe the bug
I'm trying to use a single IP adapter with multiple IP adapter images and masks. This section of the docs gives an example of how I could do that: https://huggingface.co/docs/diffusers/v0.29.0/en/using-diffusers/ip_adapter#ip-adapter-masking
The docs provide the following code:
```python
fr... | closed | completed | false | 11 | [
"bug",
"stale"
] | [] | 2024-06-18T18:06:37Z | 2024-09-23T11:36:10Z | 2024-09-23T11:36:10Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | chrismaltais | 17,098,959 | MDQ6VXNlcjE3MDk4OTU5 | User | false |
huggingface/diffusers | 1,410,947,248 | I_kwDOHa8MBc5UGViw | 863 | https://github.com/huggingface/diffusers/issues/863 | https://api.github.com/repos/huggingface/diffusers/issues/863 | Inference fix needed (ASAP) | Has anyone done nerdy rodent's install? It works but I get this at the end and my shell is this as followed in his tutorial
export LD_LIBRARY_PATH=/usr/lib/wsl/lib:$LD_LIBRARY_PATH
export MODEL_NAME="CompVis/stable-diffusion-v1-4"
export INSTANCE_DIR="training"
export OUTPUT_DIR="classes"
accelerate launch tr... | closed | completed | false | 3 | [
"stale"
] | [
"patil-suraj"
] | 2022-10-17T05:53:43Z | 2022-11-16T20:28:57Z | 2022-11-16T20:28:57Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | wcarletsdrive | 110,745,584 | U_kgDOBpnX8A | User | false |
huggingface/diffusers | 2,361,358,872 | I_kwDOHa8MBc6Mv3oY | 8,633 | https://github.com/huggingface/diffusers/issues/8633 | https://api.github.com/repos/huggingface/diffusers/issues/8633 | Stable audio open diffusers version | Good afternoon. Recently a new model Stable audio open was released. Will it have a diffusers version and if so, approximately when will it be? Thanks in advance for your answer | closed | completed | false | 1 | [] | [] | 2024-06-19T05:35:25Z | 2024-06-29T10:00:59Z | 2024-06-29T10:00:59Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Dartvauder | 140,557,322 | U_kgDOCGC8Cg | User | false |
huggingface/diffusers | 2,361,827,578 | I_kwDOHa8MBc6MxqD6 | 8,637 | https://github.com/huggingface/diffusers/issues/8637 | https://api.github.com/repos/huggingface/diffusers/issues/8637 | ImportError: cannot import name 'PixArtSigmaPipeline' from 'diffusers' | ### Describe the bug
ImportError: cannot import name 'PixArtSigmaPipeline' from 'diffusers'
### Reproduction
diffusers 0.29.0
### Logs
_No response_
### System Info
diffusers ... | closed | completed | false | 0 | [
"bug"
] | [] | 2024-06-19T09:17:20Z | 2024-06-19T09:25:32Z | 2024-06-19T09:25:32Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | lonngxiang | 40,717,349 | MDQ6VXNlcjQwNzE3MzQ5 | User | false |
huggingface/diffusers | 2,362,134,903 | I_kwDOHa8MBc6My1F3 | 8,638 | https://github.com/huggingface/diffusers/issues/8638 | https://api.github.com/repos/huggingface/diffusers/issues/8638 | getting bug on diffuser example | 
modelp is "dreamshaper_8Inpainting.safetensors"
https://huggingface.co/docs/diffusers/v0.29.0/en/api/pipelines/controlnet_sdxl#diffusers.StableDiffusionXLControlNetInpaintPipeline
tried compel as well it... | closed | completed | false | 8 | [] | [] | 2024-06-19T11:33:50Z | 2024-06-29T13:07:02Z | 2024-06-29T13:07:02Z | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | xalteropsx | 103,671,642 | U_kgDOBi3nWg | User | false |
huggingface/diffusers | 2,362,544,266 | I_kwDOHa8MBc6M0ZCK | 8,641 | https://github.com/huggingface/diffusers/issues/8641 | https://api.github.com/repos/huggingface/diffusers/issues/8641 | Integrate Lumina-T2X | ### Model/Pipeline/Scheduler description
Lumina-T2X is a text-to-any generation model. Our model is capable of generating multiple modalities, most notably image generation. Currently, our image generation supports any resolution, achieving stunning results. We will unify various modalities and integrate them into d... | closed | completed | false | 6 | [] | [] | 2024-06-19T14:34:50Z | 2024-10-20T23:03:03Z | 2024-10-20T23:03:03Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | PommesPeter | 54,879,512 | MDQ6VXNlcjU0ODc5NTEy | User | false |
huggingface/diffusers | 2,363,591,952 | I_kwDOHa8MBc6M4Y0Q | 8,644 | https://github.com/huggingface/diffusers/issues/8644 | https://api.github.com/repos/huggingface/diffusers/issues/8644 | Loading T5 encoder separately with StableDiffusion3Pipeline causes meta tensor error on sequential/model cpu offload | ### Describe the bug
Attempting to load an SD3 checkpoint which includes clips from a single file, and supplying a T5 encoder loaded separately causes a meta tensor data copy error when calling ``enable_sequential_cpu_offload()`` or ``enable_model_cpu_offload()``
### Reproduction
```python
import diffusers
impor... | closed | completed | false | 5 | [
"bug"
] | [] | 2024-06-20T06:03:49Z | 2024-06-23T19:27:05Z | 2024-06-23T19:27:05Z | CONTRIBUTOR | null | 20260407T133413Z | 2026-04-07T13:34:13Z | Teriks | 14,919,098 | MDQ6VXNlcjE0OTE5MDk4 | User | false |
huggingface/diffusers | 2,363,633,229 | I_kwDOHa8MBc6M4i5N | 8,645 | https://github.com/huggingface/diffusers/issues/8645 | https://api.github.com/repos/huggingface/diffusers/issues/8645 | There is no create_diffusers_controlnet_model_from_ldm function in single_file_utils.py | ### Describe the bug
in diffusers/loaders/controlnet.py, there is
from .single_file_utils import (
create_diffusers_controlnet_model_from_ldm,
fetch_ldm_config_and_checkpoint,
)
But, there is no create_diffusers_controlnet_model_from_ldm function in single_file_utils.py
### Reproduction
from .single_... | open | null | false | 3 | [
"bug",
"stale",
"single_file"
] | [] | 2024-06-20T06:32:26Z | 2024-09-14T15:08:41Z | null | NONE | null | 20260407T133413Z | 2026-04-07T13:34:13Z | meng-yuan321 | 73,415,049 | MDQ6VXNlcjczNDE1MDQ5 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.