# What's the best way to store BERT training data (input IDs)

The tricky thing about the input IDs is what they're varying in length for each data sample, so regular hdf5 may not be ideal. Since Bert is so popular I am wondering if there's an established way to store the data.

• If you are using pytorch, you can put your data in pt files, using torch.save and torch.load – Astariul Nov 8 '19 at 0:50