Comments (4)
I don't know if same issue, but same error, with hidiffusers.
Running hidiffusion in an xyz script renders a couple of images, then breaks. Reloading the ckpt enables the same number of images be generated before breaking again.
Running same prompt and parameters without hidiffusers yields no errors.
PS C:\ai\automatic> .\webui.bat --medvram --debug
Using VENV: C:\ai\automatic\venv
17:43:37-160033 INFO Starting SD.Next
17:43:37-163505 INFO Logger: file="C:\ai\automatic\sdnext.log" level=DEBUG size=65 mode=create
17:43:37-164994 INFO Python 3.10.11 on Windows
17:43:37-347574 INFO Version: app=sd.next updated=2024-05-18 hash=ef736488 branch=dev
url=https://github.com/vladmandic/automatic/tree/dev
17:43:37-938624 INFO Latest published version: e081f232d2648d79659b9a53070cecb544b85b3d 2024-05-07T19:11:45Z
17:43:37-953007 INFO Platform: arch=AMD64 cpu=Intel64 Family 6 Model 165 Stepping 5, GenuineIntel system=Windows
release=Windows-10-10.0.22631-SP0 python=3.10.11
17:51:46-784318 ERROR Processing: RuntimeError
╭───────────────────────────────────────── Traceback (most recent call last) ──────────────────────────────────────────╮
│ C:\ai\automatic\modules\processing_diffusers.py:118 in process_diffusers │
│ │
│ 117 │ │ # base_args['image'] = set_latents(p) │
│ ❱ 118 │ │ output = shared.sd_model(**base_args) # pylint: disable=not-callable │
│ 119 │ │ if isinstance(output, dict): │
│ │
│ C:\ai\automatic\venv\lib\site-packages\torch\utils\_contextlib.py:115 in decorate_context │
│ │
│ 114 │ │ with ctx_factory(): │
│ ❱ 115 │ │ │ return func(*args, **kwargs) │
│ 116 │
│ │
│ C:\ai\automatic\venv\lib\site-packages\diffusers\pipelines\stable_diffusion_xl\pipeline_stable_diffusion_xl.py:1174 │
│ │
│ 1173 │ │ │ │ │ added_cond_kwargs["image_embeds"] = image_embeds │
│ ❱ 1174 │ │ │ │ noise_pred = self.unet( │
│ 1175 │ │ │ │ │ latent_model_input, │
│ │
│ C:\ai\automatic\venv\lib\site-packages\torch\nn\modules\module.py:1511 in _wrapped_call_impl │
│ │
│ 1510 │ │ else: │
│ ❱ 1511 │ │ │ return self._call_impl(*args, **kwargs) │
│ 1512 │
│ │
│ C:\ai\automatic\venv\lib\site-packages\torch\nn\modules\module.py:1561 in _call_impl │
│ │
│ 1560 │ │ │ │
│ ❱ 1561 │ │ │ result = forward_call(*args, **kwargs) │
│ 1562 │ │ │ if _global_forward_hooks or self._forward_hooks: │
│ │
│ C:\ai\automatic\venv\lib\site-packages\accelerate\hooks.py:166 in new_forward │
│ │
│ 165 │ │ else: │
│ ❱ 166 │ │ │ output = module._old_forward(*args, **kwargs) │
│ 167 │ │ return module._hf_hook.post_forward(module, output) │
│ │
│ C:\ai\automatic\venv\lib\site-packages\diffusers\models\unets\unet_2d_condition.py:1281 in forward │
│ │
│ 1280 │ │ │ if hasattr(upsample_block, "has_cross_attention") and upsample_block.has_cross_attention: │
│ ❱ 1281 │ │ │ │ sample = upsample_block( │
│ 1282 │ │ │ │ │ hidden_states=sample, │
│ │
│ C:\ai\automatic\venv\lib\site-packages\torch\nn\modules\module.py:1511 in _wrapped_call_impl │
│ │
│ 1510 │ │ else: │
│ ❱ 1511 │ │ │ return self._call_impl(*args, **kwargs) │
│ 1512 │
│ │
│ C:\ai\automatic\venv\lib\site-packages\torch\nn\modules\module.py:1520 in _call_impl │
│ │
│ 1519 │ │ │ │ or _global_forward_hooks or _global_forward_pre_hooks): │
│ ❱ 1520 │ │ │ return forward_call(*args, **kwargs) │
│ 1521 │
│ │
│ C:\ai\automatic\modules\hidiffusion\hidiffusion.py:1682 in forward │
│ │
│ 1681 │ │ │ │ │
│ ❱ 1682 │ │ │ │ hidden_states = torch.cat([hidden_states, res_hidden_states], dim=1) │
│ 1683 │
╰──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 128 but got size 64 for tensor number 1 in the list.
from automatic.
fixed.
hidiffusion had "unapply" function, but it was not unapplying all modifications that hidiffusion did, so each time it was going deeper and deeper into its own hole.
from automatic.
It might be related. You don't even need xyz script. Just running three separate generations will give the error on the third one. I've reported this in discord yesterday. No comments there yet, so it's good to have that issue also here.
from automatic.
fixed. hidiffusion had "unapply" function, but it was not unapplying all modifications that hidiffusion did, so each time it was going deeper and deeper into its own hole.
I confirm, it is fixed for me. Thank you!
from automatic.
Related Issues (20)
- [Feature]: Networks/LoRA: add a link back to the source (Civitai etc.) HOT 1
- [Issue]: Playground v2.5 get wrong result HOT 1
- [Issue]: Uncaught exception occurred: type=<class 'KeyError'> value='branch' HOT 1
- [Feature]: Lumina Model Support
- [Issue]: SDNext Dev freezes at 6%. (Win11-AMD-DirectML) HOT 1
- [Feature]: WebUI outputs to use in API requests HOT 2
- [Issue]: Negative prompt not being handled in control endpoint HOT 1
- [Issue]: Selecting Control model shows error in log HOT 1
- [Issue]: EN Styles Breaks due to '.json' file HOT 2
- [Issue]: Face Hires with batch counts > 1 causes (sporadically) exceptions HOT 1
- [Feature]: allow FaceID together with T2I adapter HOT 1
- [Issue]: FaceID in txt2img input prompt ignore partially? HOT 1
- [Issue]: running SD3 HOT 13
- [Issue]: API IP Adapter start value error HOT 1
- [Extension]: ADetailer not being "enabled" via UI HOT 2
- [Issue]: Control endpoint ignores init images. HOT 1
- [Issue]: Pydantic error on Apple M1 HOT 4
- [Feature]: Add the ability to select Extra Networks (loras at this point) to xyz script HOT 5
- [Feature]: Option to Force GC after every operation. HOT 6
- [Issue]: SDXL model swapping freezes the computer (ROCm on AMD RX 7900 XTX) HOT 3
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from automatic.