COAT: Compressing Optimizer states and Activation for Memory-Efficient FP8 Training

Devs

COAT: Compressing Optimizer states and Activation for Memory-Efficient FP8 Training | Read Paper on Bytez