Yo! Batch size is such a balancing act, right? I’ve found that for how to split dataset into batches, starting with a batch size of 32 works well for most cases. But yeah, shuffling is key to avoid bias.
If you’re dealing with imbalanced data, check out the `imblearn` library. It’s got some sick tools for oversampling and undersampling. Also, don’t forget to normalize your data before batching—it makes a huge diff.
If you’re dealing with imbalanced data, check out the `imblearn` library. It’s got some sick tools for oversampling and undersampling. Also, don’t forget to normalize your data before batching—it makes a huge diff.