Honestly, I’ve been using PyTorch’s DataLoader for how to split dataset into batches, and it’s been a lifesaver. You can set the batch size, shuffle, and even drop the last batch if it’s smaller.
For imbalanced datasets, I’ve had success with weighted sampling. It’s a bit more work, but it keeps the model from favoring the majority class. Also, don’t forget to monitor your GPU usage—batch size can make or break your training time.
For imbalanced datasets, I’ve had success with weighted sampling. It’s a bit more work, but it keeps the model from favoring the majority class. Also, don’t forget to monitor your GPU usage—batch size can make or break your training time.