Show HN: Hardware-software co-design for MoE models to bypass NCCL bottlenecks (github.com) 3 points by PJHkorea 1mo ago ↗ HN
[–] PJHkorea 1mo ago ↗ Imagine virtual memory, but instead of using your local storage, you’re pooling a whole bunch of network servers into a single address space
1 comment
[ 0.23 ms ] story [ 5.1 ms ] thread