Wow, thanks for all the awesome tips, everyone! I tried using scikit-learn’s `train_test_split` and it worked like a charm. Also, shoutout to whoever mentioned `imblearn`—I’ve been playing around with SMOTE, and it’s been a game-changer for my imbalanced dataset.
Quick question though: how do you guys decide on the perfect batch size? I’ve been experimenting with 32 and 64, but I’m curious if there’s a rule of thumb or if it’s just trial and error. Also, anyone tried dynamic batching? Heard it’s a thing but not sure if it’s worth the effort.
Quick question though: how do you guys decide on the perfect batch size? I’ve been experimenting with 32 and 64, but I’m curious if there’s a rule of thumb or if it’s just trial and error. Also, anyone tried dynamic batching? Heard it’s a thing but not sure if it’s worth the effort.