-
Notifications
You must be signed in to change notification settings - Fork 6.8k
Enable Large Tensor Support: Stage 1 #18625
Conversation
Hey @access2rohit , Thanks for submitting the PR
CI supported jobs: [sanity, miscellaneous, edge, centos-cpu, windows-gpu, windows-cpu, clang, unix-cpu, website, centos-gpu, unix-gpu] Note: |
@leezu @ChaiBapchya @josephevans Please review |
@mxnet-label-bot add [pr-awaiting-review] |
5447ef4
to
4c20ec8
Compare
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.
If its ready to merge, lets rename the PR title.
Looks good except that MKL_IF_AVAILABLE switch.
@@ -53,7 +53,8 @@ struct polyval_backward_p { | |||
DType igrad_p = 0; | |||
index_t j = x_size - 1; | |||
while (j >= 0) { | |||
igrad_p += pow(x_dptr[j], p_size - i - 1) * ograd_dptr[j]; | |||
igrad_p += pow(x_dptr[j], static_cast<DType>(p_size) - |
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.
just to confirm, there is a unittest coverage for this change?
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.
@@ -53,7 +53,8 @@ struct polyval_backward_p { | |||
DType igrad_p = 0; | |||
index_t j = x_size - 1; | |||
while (j >= 0) { | |||
igrad_p += pow(x_dptr[j], p_size - i - 1) * ograd_dptr[j]; | |||
igrad_p += pow(x_dptr[j], static_cast<DType>(p_size) - |
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.
This prevents Windows CI build failure
bf5a241
to
1b4e4b3
Compare
1b4e4b3
to
bee7e02
Compare
bee7e02
to
f633986
Compare
206c792
to
3e6b50f
Compare
04c78de
to
9e0796d
Compare
56ca8b4
to
5cdf8d8
Compare
@leezu would you review |
Description
This PR enables Large Tensor Support(LTS) by default on master for all platforms except for Miscellaneous CLang build and UNIX MKL blas builds with lapack for CI stages and build scripts for both dynamic and static builds(make as well as ninja).
DO NOT MERGE BEFORE THIS PR: #17882
Progress for Large Tensor Support tracked here: #17331
Checklist
Essentials
Please feel free to remove inapplicable items for your PR.
@leezu @ChaiBapchya @josephevans