NVIDIA Minitron: Compact Language Models via Pruning and Knowledge Distillation

知识蒸馏