TensorNetwork Support real and imag at the backend level

Sep 28 '20 12:09 alewis

Hi @alewis, Any ideas on how to tackle this?

Oct 01 '20 17:10 ahmedkrmn

For every backend except PyTorch, I think .real and .imag are harmless when applied to real input, so in those cases we can just wrap the relevant function. PyTorch will require some special handling. I guess a simple try / except would work?

On Thu, Oct 1, 2020 at 1:03 PM Ahmed Karaman [email protected] wrote:

Hi @alewis https://github.com/alewis, Any ideas on how to tackle this?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/google/TensorNetwork/issues/841#issuecomment-702271385, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAINHBS7YWOUO7H6JCJ7CHTSISY5TANCNFSM4R4PC7TA .

Oct 01 '20 17:10 alewis

Hey, Is this issue still still open?

Oct 08 '20 13:10 MikeMerz

Yep, this one is, though I don't know whether @ahmedkrmn is working on it?

Oct 08 '20 14:10 alewis

@MikeMerz, feel free to work on it if you want to. I've been busy lately so I still haven't started.

Oct 10 '20 12:10 ahmedkrmn

@ahmedkrmn Thanks mate. @alewis Just a thought: 1)For numpy and tensorflow backends, it's possible to just to call numpy and tensor's .real and .imag? as the input is of the same type.

Oct 10 '20 16:10 MikeMerz

Yes, that should work in all cases except PyTorch and Symmetric. For PyTorch you'll need to do some additional type checking because calling . real and .imag throws an error for non complex input in that case.

On Sat, Oct 10, 2020, 12:26 MikeMerz [email protected] wrote:

@ahmedkrmn https://github.com/ahmedkrmn Thanks mate. Just a though: 1)For numpy and tensorflow backends, it's possible to just to call numpy and tensor's .real and .imag? as the input is of the same type.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/google/TensorNetwork/issues/841#issuecomment-706574826, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAINHBTSARWHKVUX6PP6W6TSKCDMNANCNFSM4R4PC7TA .

Oct 10 '20 16:10 alewis

@alewis For PyTorch, is it better to raise an excpetion if the input is not complex or treat it as 0j?

Oct 10 '20 17:10 MikeMerz

I think the latter, yeah.

On Sat, Oct 10, 2020, 13:22 MikeMerz [email protected] wrote:

For PyTorch, is it better to raise an excpetion if the input is not complex or treat it as 0j?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/google/TensorNetwork/issues/841#issuecomment-706583344, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAINHBTRMYYG76K67CQM4JLSKCJ65ANCNFSM4R4PC7TA .

Oct 10 '20 20:10 alewis

@alewis Hey, I unintentionally pushed the changes of this Issue to the ones_like and zero_like Pull request

Oct 11 '20 17:10 MikeMerz

Hmm, do you think you can revert it and split them up?

On Sun, Oct 11, 2020, 13:39 MikeMerz [email protected] wrote:

@alewis https://github.com/alewis Hey, I unintentionally pushed the changes of this Issue to the ones_like and zero_like Pull request

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/google/TensorNetwork/issues/841#issuecomment-706740113, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAINHBV34EXQEYSG6AXYDFLSKHUWTANCNFSM4R4PC7TA .

Oct 11 '20 19:10 alewis

@alewis after much tries, I re-uploaded the changes for the ones_like & zero_like issue , had to close the last request

Oct 12 '20 14:10 MikeMerz

Heyy @MikeMerz , @alewis I am a new contributor can i take up thi s?? and can someone provide little heads-up or description about this issue??

Aug 20 '23 06:08 AnuravModak