How does jax.jit
decide when to use gradient checkpointing/recomputation?
#20252
Unanswered
zhixuan-lin
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
I read here that
jax.jit
does gradient checkpointing automatically. How does it decide when to use it for different operations? I'm noticing strange memory usage changes in my program with certain code changes (e.g., broadcasting a small array to a larger array dramatically reduces memory usage). I suspect that this is because it causesjax.jit
to make different decisions regarding gradient checkpointing, but I'm not sure.Beta Was this translation helpful? Give feedback.
All reactions