DL Template

Integrates Pytorch lightning + Hydra + Neptune for LM finetuning

To finetune bert on SST2 dataset:

python main.py

To run without neptune (offline mode)

python main.py --offline

To debug on small part of the dataset

python main.py --debug

To override specific hydra config

python main.py --override fixed

This will start using the fixed lr scheduler instead of the linear_with_warmup mentioned in configs/config.yaml. Override can be comma separated config names that we want to override for each category (e.g., --override fixed,adamw,rtx_2080.