3 ms·
From what I've seen at FAANG companies, GPU clusters for ML researchers are similar. Basically a bunch of beefy nodes with 96 CPUs and 8 Nvidia A100 cards each
by sneed_chucker 3y ago
From what I've seen at FAANG companies, GPU clusters for ML researchers are similar.
Basically a bunch of beefy nodes with 96 CPUs and 8 Nvidia A100 cards each configured in a SLURM cluster. Users get SSH access and can run tmux, jupyter, or whatever else is needed for their workflow.