Fully Sharded Data Parallel: Faster AI Training with Fewer GPUs (engineering.fb.com) 3 points by TheGuyWhoCodes 5y ago ↗ HN
2 comments
[ 0.21 ms ] story [ 21.1 ms ] thread