Y
Hacker News
new
|
ask
|
show
|
jobs
by
nyrikki
853 days ago
I have a hunch they made their decision to train off that particular type of A* traces to avoid an exponential number of embeddings.