Want to create an interactive transcript for this episode?
Podcast: Arxiv Papers
Episode: Adaptive Batch Size Schedules for Distributed Training of Language Models with Data and Model Parallelism