Dependencies seem to be bugged with latest xFormers?

[bhowiebkr]

I was running into issues with the install that I think is due to the version of xFormers pip is pulling. Went to try these versions in https://github.com/ExponentialML/AnimateDiff-MotionDirector/issues/2#issuecomment-1924654048

I always seem to run into issues with conda so probably next time I have to use this i'll use virtualenv

could be good idea to make a Release and freeze the dependencies to make it easier to get up and running

After getting it working Here's the short list from the main dependencies:

diffusers==0.11.1
torch==1.13.1+cu117
pyav==12.0.2
torchvision==0.14.1+cu117
transformers==4.38.1
xformers==0.0.16
imageio-ffmpeg==0.4.9
imageio==2.34.0
decord==0.6.0
loralib==0.1.2
gdown==5.1.0
einops==0.7.0
omegaconf==2.3.0
safetensors==0.4.2
gradio==4.19.2
wandb==0.16.3
lion-pytorch==0.1.2

I haven't ran the code yet, but this got me past the requirements.txt

[ShoryuKyzan]

This also helped me get things installed, but had to comment out torch/torchvision and do the below:

On windows, got torch and torchvision by using the following

pip install torch==1.13.1+cu117 torchvision==0.14.1+cu117 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu117

Off top of my head, I dunno how to get that into the requirements.txt (use different index)

I'm a bit of a beginner with cuda versions. I'm assuming I may need 11.7 installed as well to get this working?

[tonynoce]

I support freezing the requirements