Updated Xformers (markdown)

AUTOMATIC1111 2022-10-08 19:53:14 +03:00
parent 03cfb4c5d5
commit 468ec85373

@ -1,24 +1,23 @@
Xformers library is an optional way to speedup your image generation. Xformers library is an optional way to speedup your image generation.
There are not binaries for Windows except for one specific configuration, but you can build it yourself. A guide from an anonymous user: There are no binaries for Windows except for one specific configuration, but you can build it yourself.
A guide from an anonymous user, although I think it is for building on Linux:
GUIDE ON HOW TO BUILD XFORMERS GUIDE ON HOW TO BUILD XFORMERS
also includes how to uncuck yourself from sm86 restriction on voldy's new commit also includes how to uncuck yourself from sm86 restriction on voldy's new commit
1. go to the webui directory 1. go to the webui directory
2. source ./venv/bin/activate 2. `source ./venv/bin/activate`
3. cd repositories 3. `cd repositories`
3. git clone https://github.com/facebookresearch/xformers.git 3. `git clone https://github.com/facebookresearch/xformers.git`
4. cd xformers 4. `cd xformers`
5. git submodule update --init --recursive 5. `git submodule update --init --recursive`
6. pip install -r requirements.txt 6. `pip install -r requirements.txt`
7. pip install -e . 7. `pip install -e .`
If you encounter some error about torch not being built with your cuda version blah blah, then try: If you encounter some error about torch not being built with your cuda version blah blah, then try:
pip install setuptools==49.6.0 `pip install setuptools==49.6.0`
After step 7, just wait like 30 minutes for everything to build and you're done. 8. wait like 30 minutes for everything to build
9. launch with `--force-enable-xformers` commandline argument
Then go back to modules/sd_[hijack.py](http://hijack.py/) and search for this text and delete it:
```python
and torch.cuda.get_device_capability(shared.device) == (8, 6)
```