-
Notifications
You must be signed in to change notification settings - Fork 65
Add op (unbind) | feat(torchlib) #831
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
Conversation
Codecov Report
@@ Coverage Diff @@
## main #831 +/- ##
=======================================
Coverage 76.23% 76.23%
=======================================
Files 113 113
Lines 13312 13316 +4
Branches 1321 1321
=======================================
+ Hits 10148 10152 +4
Misses 2838 2838
Partials 326 326
|
if ( | ||
op.name.startswith("split") | ||
or op.name.startswith("chunk") | ||
or op.name.startswith("unbind") |
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.
Growing
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 suspect we can know look at the function signature for this?
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.
Oh yup. We can leverage on Opschema and get output dtype.
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 had a try yesterday but failed to find a clear solution.
Co-authored-by: Ti-Tai Wang <[email protected]>
core_ops.aten_unbind, | ||
).xfail( | ||
dtypes=[torch.float16], | ||
reason="fixme: SplitToSequence op inference failed. https://github.com/microsoft/onnxruntime/issues/16006", |
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.
nit: What is op inference? Is this the common issue of ORT missing impl for float16 dtype?
not blocking merge.
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.
My read is shape inference failed? I will test it on 1.15 and update the issue
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.
Yes. Updated the issue
Add unbind op into torch_lib functions.
Add tests as well.