Viewing profile — jfrankle

whyyy

Honestly, just a matter of having the time to clean everything up and get it out. The ancillary code, model cards, etc. take a surprising amount of time.

We've come to accept that it's an impossible problem at this point. Instead, we're getting good at automatically detecting hardware failures and rapidly restarting runs on fewer no…

It's an indictment of the A100 node that died on us yesterday, leaving us with 248 GPUs in the particular cluster where we were running the experiments :( It turns out that, in the…

This is an illuminating (and notably rigorous) read for anyone interested in neural network sparsity and compression. But - equally importantly - it's a valuable read for anyone in…

(Author of paper here) This is approximately my growing suspicion.