Fetching the paper…

Tempo: Accelerating Transformer-Based Model Training through Memory Footprint Reduction · Around