How to deploy a trained model?

[catastropiyush]

Hello,
I just started out with using NequIP.
$ nequip-train configs/example.yaml After running this I get the following files in the directory.

But how do I deploy the trained model for LAMMPS?
nequip-deploy build path/to/training/session/ path/to/deployed.pth
What should one put in path/to/training/session the directory that contains the .yaml file?

pair_style	nequip
pair_coeff	* * deployed.pth <NequIP type for LAMMPS type 1> <NequIP type for LAMMPS type 2> ...

What are LAMMPS type 1 and LAMMPS type 2 here?

I apologize if the questions are too basic.
Thank you for your time

[Linux-cpp-lisp]

Hi @catastropiyush ,

Thanks for your interest in nequip! The Allegro tutorial (https://github.com/mir-group/allegro?tab=readme-ov-file#tutorial) should address these questions.

[catastropiyush]

Thank you for replying, I was running the colab on a T4 GPU and I am getting the following error.
After running the evaluate test error command
!nequip-evaluate --train-dir results/silicon-tutorial/si --batch-size 1
The following error pops up
FileNotFoundError: [Errno 2] No such file or directory: 'results/silicon-tutorial/si/best_model.pth'
And the subsequent code after that does not run.

There is no best_model.pth also after running the training command

I might be missing something. Please have a look.

Thank you for your time and consideration,
Piyush

[catastropiyush]

Hello, @Linux-cpp-lisp can you please look into this?

Thank you for your time,
Piyush

[Linux-cpp-lisp]

If there is no best_model.pth, your training likely did not actually run a full epoch. Did you see it get through a validation after epoch in the logs?

[catastropiyush]

After running this
!rm -rf ./results !nequip-train allegro/configs/tutorial.yaml --equivariance-test

After that

I think the error is in the first train command.
Can you please look into it. @Linux-cpp-lisp
Thank you for your time,
Piyush