:description: Learn how to use PyTorch's varlen_attn API for efficient variable length attention without padding. Complete tutorial with code examples for training Transformers with packed sequences. ...