search

txie-93 / cdvae

An SE(3)-invariant autoencoder for generating the periodic structure of materials [ICLR 2022]
MIT License
211 stars 85 forks source link

Unable to run CDVAE or training script #53

Open snme opened 7 months ago

snme commented 7 months ago

On multiple systems (GPU and CPU), I've tried:

  1. Cloning the repository
  2. Creating the conda environment (both with the cpu-only yaml and the gpu yaml)
  3. Changing .env.template to .env and editing the directories listed in it
  4. Trying to run the training script with python cdvae/run.py data=perov expname=perov.

No matter how I try to configure my system, I always get this error upon trying to run the training process: image

This means that I'm not able to train or use CDVAE at all. Can someone explain this dependency issue, and how to get around it? Thank you so much!

confymacs commented 6 months ago

Hi snme,

I encountered this problem when I tried to install pytorch-lightning==1.3.8 and I guess it was triggered by some incompatibility of the version of torchmetrics. The problem disappeared when I installed pytorch-lightning == 2.1.2 and torchmetrics==1.2.1.

stargolike commented 6 months ago

i also meet this problem. you shouldpip uninstall torchmetrics==0.9.0.dev0 pip install torchmetrics==0.5 now i resolve this problem