Block Pruning For Faster Transformers
This work addresses the need for efficient transformer models in NLP tasks, offering a method that combines benefits of pruning and distillation, though it is incremental as it builds on existing structured pruning paradigms.
The paper tackles the problem of large, slow pre-trained models by introducing a block pruning approach that reduces model size and speeds up inference, achieving a 2.4x faster and 74% smaller BERT on SQuAD v1 with only a 1% drop in F1 score.
Pre-training has improved model accuracy for both classification and generation tasks at the cost of introducing much larger and slower models. Pruning methods have proven to be an effective way of reducing model size, whereas distillation methods are proven for speeding up inference. We introduce a block pruning approach targeting both small and fast models. Our approach extends structured methods by considering blocks of any size and integrates this structure into the movement pruning paradigm for fine-tuning. We find that this approach learns to prune out full components of the underlying model, such as attention heads. Experiments consider classification and generation tasks, yielding among other results a pruned model that is a 2.4x faster, 74% smaller BERT on SQuAD v1, with a 1% drop on F1, competitive both with distilled models in speed and pruned models in size.