-
Notifications
You must be signed in to change notification settings - Fork 789
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
This is a fix to get stable_txt2img working on an M1 Mac. #36
base: main
Are you sure you want to change the base?
Conversation
- Install Stable Diffusion as per https://medium.com/gft-engineering/macbook-m1-how-to-install-and-run-stable-diffusion-7bfb2f802b1a - Install PyTorch by running `conda install pytorch torchvision torchaudio -c pytorch-nightly` Running with more than one sample seems to break it so I'm just running multiple itterations to get the regularization images. `python scripts/stable_txt2img.py --ddim_eta 0.0 --n_samples 1 --n_iter 200 --scale 10.0 --ddim_steps 50 --ckpt ~/Downloads/sd-v1-4-full-ema.ckpt --prompt "a photo of a <class>`
How long is it taking you to train the models this way? |
I can't get this repo (not the lstein one mentionned by OP) to train on M1. |
@swankwc I haven't gotten to training yet. ATM his patch is just for I'm having problems getting
I can keep using this PR to track that work or open a new one. Opinions? |
@Sorrow do you have your work somewhere? You seem to have gotten further than me. Maybe we can collaborate. Here's my WIP, it's very rough ATM https://github.com/beettlle/Dreambooth-Stable-Diffusion/tree/m1-training-fix |
Renamed PR explain scope of work better. |
@beettlle, I've been able to get it up and running on my Macbook Pro with some modifications using your code. It's linked here if you'd like to take a look: https://github.com/SujeethJinesh/DreamBoothMac |
That's awesome @SujeethJinesh ! Let me reset my env and I'll try it tomorrow. |
@SujeethJinesh I'm still getting the following error with your branch. Any ideas?
|
is there any progress (in @SujeethJinesh's built)? i cant even generate the regularization images on MPS as it doesnt support double precision floats, but sd requires them |
I tried running @SujeethJinesh 's repo and got the same error as you @HannesGitH . Some additional things I was to install the following
|
My latest attempt to fix was to perform the cast as follows
but that gave me this error
|
@SujeethJinesh open a PR SujeethJinesh/DreamBoothMac#3 to fix the float64 error. I was able to get around the error
I changed the size of the image to 256 x 256 that did the trick. It unblocks me for now, but it would be good to figure out a better solution. I will try to fix later. |
In my latest attempt I tried to perform training
I get the following error
My best guess is that the pytorch lighting version specified (1.5.9) does have this feature https://lightning.ai/docs/pytorch/stable/accelerators/mps_basic.html @SujeethJinesh how did you get this work? |
@alberto-salinas would you mind trying the following from 's site in your environment to see if MPS is supported?
Output should be: |
conda install pytorch torchvision torchaudio -c pytorch-nightly
Running with more than one sample seems to break it so I'm just running multiple itterations to get the regularization images:
python scripts/stable_txt2img.py --ddim_eta 0.0 --n_samples 1 --n_iter 200 --scale 10.0 --ddim_steps 50 --ckpt ~/Downloads/sd-v1-4-full-ema.ckpt --prompt "a photo of a <class>