-
Notifications
You must be signed in to change notification settings - Fork 3.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[TOPI][CUDA] Schedule for pool_grad #3622
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM :-)
topi/python/topi/cuda/pooling.py
Outdated
out = op | ||
else: | ||
out = outs[0].op.output(0) | ||
s[op].set_scope('local') |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think this can be set automatically if it is compute_at a threadIdx
ff29150
to
e5e4040
Compare
@vinx13 can you get the pool_grad test from the current mode into numerical checking? In the current mode, we essentially copy and paste the grad function into the test function, and compare the two result. If our orginal concept is wrong, what we test against is likely to also be wrong. I think we should slowly replace basically all gradient test with numerical checking ones, and the more complex the grad is, the more crucial this is for us. |
* [TOPI][CUDA] Schedule for pool_grad * Relay test * Fix fused op * doc * Remove set scope local
* [TOPI][CUDA] Schedule for pool_grad * Relay test * Fix fused op * doc * Remove set scope local
Please review @junrushao1994 @MarisaKirisame @merrymercy @masahi