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
Support for libtorch-1.3 #202
Support for libtorch-1.3 #202
Conversation
junjihashimoto
commented
Oct 12, 2019
•
edited
edited
- Add Dimname and DimnameList of new data-types
- Fix compilation-error
- Add Symbol of new data-type
- Upload binaries of libtorch-1.3
- Refine auto generated functions.
ab96610
to
4a64e90
Compare
Thanks for getting this going @junjihashimoto! Let us know if there’s aspects others can help with. We may need to tweak some of the examples and higher level apis depending on how the ffi gets updated. |
@austinvhuang |
hi @junjihashimoto! |
@tscholak |
@junjihashimoto @tscholak from what I understand the trajectory for libtorch is basically feature parity with the python api, which is great in terms of having as much functionality as possible. There might be some features where we'll have to make design decisions in terms of how they're interfaced though (or even to have high level interfaces at all). For namedtensor, it's not obvious to me what it looks like yet. But we can go ahead with the codegen bindings first and consider options for the functionality. |
@austinvhuang Oh, yes, for sure we can go ahead without figuring out named tensors! I was just surprised to find any evidence of named tensors in libtorch. |
@tscholak @austinvhuang |
sweet! thanks @junjihashimoto, what an achievement! |
@@ -25,13 +25,13 @@ spec = do | |||
(shape (asTensor v)) `shouldBe` [3] | |||
print (toDense x) | |||
-- When we call print for sparse tensor, it throws a exception. | |||
(print x) `shouldThrow` anyException | |||
print x -- `shouldThrow` anyException |
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.
doesn't throw anymore? did they implement something sparse-related that wasn't implemented before?
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.
No. Our understanding is right. They changed the implementation, but I do not find the change.
We can print out sparse-tensor.
Wow @junjihashimoto this is amazing. Considering how long it took to catch up to the last PTDC release and now it's down to a few days. Any idea what's going on with the OSX Circle CI build? Should CircleCI just be deprecated entirely? If CI passes, I'm good with going ahead and merging. |
@austinvhuang |