v0.1.3

Added

• HuggingFace integration (seqpacker.hf_utils) with pack_dataset and pack_dataset_from_result for building packed datasets.Dataset objects directly -- no tensor roundtrip needed
• Training Integration docs page combining HF Trainer and PyTorch DataLoader workflows
• Training examples: examples/sft_trainer.py (TRL SFTTrainer) and examples/pytorch_training.py (custom training loop)

v0.1.2

Fixed

• Crate README code examples: corrected Pack field access (sequences not items), used_capacity() method, and StreamPacker::add() signature
• Crate README algorithms table: corrected ParallelOptimizedBestFitDecreasing enum variant name

Changed

• GitHub release workflow now sources release notes from CHANGELOG.md

v0.1.1

Full Changelog: v0.1.0...v0.1.1