3 ms·
Always figured it would be too slow. Distributed training on clusters is usually done with 1+ gb/s interconnects.
by humansareok1 3y ago
Always figured it would be too slow. Distributed training on clusters is usually done with 1+ gb/s interconnects.