E engineering.fb.com Fully Sharded Data Parallel: Faster AI Training with Fewer GPUs 3 points TheGuyWhoCodes 5 years ago 2 comments
