-
Notifications
You must be signed in to change notification settings - Fork 6.8k
[v1.x] add large matrix tests for linalg ops: det, inverse, trsm, trmm #18744
Conversation
Hey @mseth10 , Thanks for submitting the PR
CI supported jobs: [windows-gpu, centos-cpu, sanity, unix-cpu, miscellaneous, edge, website, windows-cpu, centos-gpu, unix-gpu, clang] Note: |
The nightly large tensor tests are already broken: http://jenkins.mxnet-ci.amazon-ml.com/blue/organizations/jenkins/NightlyTestsForBinaries/detail/master/752/pipeline/96 Adding more tests may not help. Should they be fixed and enabled as part of the PR runs? As this will become a default feature, I don't think testing nightly is sufficient (especially as the test only runs 30min). |
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
aaa6120
to
fce478f
Compare
@access2rohit please review |
@leezu This is currently for MXnet1.x branch and not master. For master more tests will be added. In the meantime since we are removing NDarray's completely it makes sense to disabled them from nightly. |
b56fad4
to
7dc2004
Compare
@access2rohit made the changes, can you please approve |
Also I don't see a |
b76a34a
to
303e5ab
Compare
7ec9749
to
889bfd2
Compare
All unit tests pass. Here are the results using pytest:
|
Added backward invocation @ChaiBapchya |
I see missing shape checks, can you add them? ... rest LGTM! |
LGTM thanks Manu for the tweaks |
Added shape checks. Can you please approve? @access2rohit |
@mseth10 Did you run the tests again after making these changes ? Did they all pass ? |
Yes @access2rohit |
@mxnet-bot run ci [unix-gpu] |
Jenkins CI successfully triggered : [unix-gpu] |
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.
Clean code and Tests run successfully.... LGTM!
@mxnet-bot run ci [unix-gpu] |
Jenkins CI successfully triggered : [unix-gpu] |
apache#18744) * add linalg large matrix tests * add batch inputs linalg tests * reducing bsize to 1 to save time * move matrix generator to utils * passing mat size as arg * import util fn * fix sanity * add mx * call backward * merge fn * update grad value * refactor tests * add mx * add shape check Co-authored-by: Ubuntu <ubuntu@ip-172-31-41-26.us-west-2.compute.internal>
Description
In this PR, we add large matrix (Input Size > 2^32 -1) tests for the following linear algebra ops: det, inverse, trsm, trmm
We test for shape and numerical correctness of forward and backward passes, with normal and batched inputs.
Here's the pytest output: