Y
Hacker News
new
|
ask
|
show
|
jobs
by
WhitneyLand
122 days ago
A bit misleading to say they take 14x less memory, no one is doing inference with 16-bit models.