Ctrl+K
- gradient_accumulation_steps=2, per_device_eval_batch_size=8, per_device_train_batch_size=4, run_name=baseline
- gradient_accumulation_steps=2, per_device_train_batch_size=4, run_name=baseline
- per_device_eval_batch_size=8, per_device_train_batch_size=8, run_name=baseline
- per_device_train_batch_size=4, run_name=baseline