Gradient Checkpointing and Mixed-Precision Training in mT5-XL Fine-Tuning for Zero-Shot XTREME Evaluation

Intermediate-task training---fine-tuning a pretrained model on an intermediate task before fine-tuning again on the target task---often improves model performance substantially on language understanding tasks in monolingual English settings. We investigate whether English intermediate-task training is still helpful on non-English target tasks. Using nine intermediate language-understanding tasks, we evaluate intermediate-task transfer in a zero-shot cross-lingual setting on the XTREME benchmark. We see large improvements from intermediate training on the BUCC and Tatoeba sentence retrieval tas

Research goal: What is the impact of gradient checkpointing or mixed-precision training on the computational efficiency of fine-tuning mT5-XL with English intermediate-task training for zero-shot XTREME evaluation?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.5/10.