Hey! I feel you on the GPU struggles. For how to split dataset into batches, I usually start with a batch size of 64 and adjust based on my GPU’s memory. Shuffling is a must, but I also like to use random seeds for reproducibility.
If you’re dealing with imbalanced data, stratified sampling is great, but you can also try SMOTE for oversampling. It’s in the `imblearn` library and works wonders.
If you’re dealing with imbalanced data, stratified sampling is great, but you can also try SMOTE for oversampling. It’s in the `imblearn` library and works wonders.