Skip to content

[Needs someone to complete] torch.einsum()#2008

Closed
vlasenkov wants to merge 1 commit intopytorch:masterfrom
vlasenkov:einsum
Closed

[Needs someone to complete] torch.einsum()#2008
vlasenkov wants to merge 1 commit intopytorch:masterfrom
vlasenkov:einsum

Conversation

@vlasenkov
Copy link
Contributor

@vlasenkov vlasenkov commented Jul 7, 2017

resolves #1889

@soumith
Copy link
Collaborator

soumith commented Jul 7, 2017

this is worth adding in tensor.py, we want it to be available not just under autograd, but also for Tensors.

@rockt transpiled TensorFlow code here: https://gist.github.com/rockt/15ee013889d65342088e9260a377dc8f to get an einsum going in PyTorch, but we were hesitant to commit that exact code as-is because we weren't sure about licensing.
You can however lift the tests etc. out of it.

@vlasenkov
Copy link
Contributor Author

Are you talking about torch/tensor.py? It defines only base tensor class and does not contain any functional API, adding einsum here seems strange...
And how is torch/functional.py related to autograd?

@soumith
Copy link
Collaborator

soumith commented Jul 13, 2017

oh, nevermind i just misread the PR. I thought it went into nn/functional.py

@vlasenkov vlasenkov changed the title [WIP] torch.einsum() [Needs someone to complete] torch.einsum() Aug 10, 2017
@giorgiop
Copy link

Is this PR still on progress?

@t-vi
Copy link
Collaborator

t-vi commented Apr 18, 2018

I think this could be closed. (@fmassa ?)

@soumith soumith closed this Apr 18, 2018
xwang233 pushed a commit to xwang233/pytorch that referenced this pull request Oct 7, 2022
* Allow non-root trivial reductions

Fixes pytorch#2008

Co-authored-by: Christian Sarofeen <csarofeen@nvidia.com>
IvanYashchuk pushed a commit to IvanYashchuk/pytorch that referenced this pull request Oct 12, 2022
* Allow non-root trivial reductions (pytorch#2037)

* Allow non-root trivial reductions

Fixes pytorch#2008

Co-authored-by: Christian Sarofeen <csarofeen@nvidia.com>

* Test file cleanup (pytorch#2040)

* Move test_gpu.cpp to test_gpu1.cpp

* Split test_gpu1.cpp to test_gpu1.cpp, test_gpu2.cpp and test_gpu3.cpp.

Each file should be up to 10K LoC. New tests should be added to
test_gpu3.cpp until it gets 10K LoC.

* format

* fix merge

* format

Co-authored-by: Christian Sarofeen <csarofeen@nvidia.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Support for einsum notation

5 participants