Efficient Sequence Packing without Cross-contamination: Accelerating Large Language Models without Impacting Performance

Open in new window