How was the 12 layer geneformer model trained?

#98
by ricomnl - opened

Hi there,

Thanks for your amazing work (love the almost daily new commits)! I was curious, how was the 12 layer geneformer model trained? Did you use the same training strategy / hyperparameters or what changed compared to how the 6 layer model was trained?

Cheers!

Thank you for your interest in Geneformer! Yes, it is the same strategy but different hyperparameters. You can load the training_args.bin to see the hyperparameters used:

import torch
training_args = torch.load(path_to_training_args)
training_args
ctheodoris changed discussion status to closed

Sign up or log in to comment