Description
Hi Mohammed and Yeqing, this is really exciting work! Thanks for sharing the code and providing nice documentation. I encountered a very minor issue running it which I thought I'd mention.
To install Genie I created a fresh venv and installed the dependencies using pip install -e .. I then downloaded and cleaned the SCOPe dataset as instructed in the README, and created a sub-directory in runs and a configuration file. I then started training but got an error. Here's the stack trace:
Traceback (most recent call last):
File "/home/joel/git/genie/genie/train.py", line 70, in <module>
main(args)
File "/home/joel/git/genie/genie/train.py", line 44, in main
trainer = Trainer(
File "/home/joel/git/genie/venv/lib/python3.10/site-packages/pytorch_lightning/utilities/argparse.py", line 69, in insert_env_defaults
return fn(self, **kwargs)
TypeError: Trainer.__init__() got an unexpected keyword argument 'gpus'
I believe the source of the issue is that pip installed v2.0.1 of PyTorch Lightning and according to this discussion Trainer may now expect the keyword arguments devices and accelerator instead of gpus. I was able to start training by replacing this line with these lines:
devices=gpus,
accelerator='gpu',
Is this issue reproducible on your end? If the issue is indeed genuine I'd be glad to submit a PR to fix it.