Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

FlatTo1D in copy Tensor is a problem for huge array! #9304

Closed
starimpact opened this issue Jan 4, 2018 · 2 comments
Closed

FlatTo1D in copy Tensor is a problem for huge array! #9304

starimpact opened this issue Jan 4, 2018 · 2 comments
Labels
Backend Issues related to the backend of MXNet Call for Contribution NDArray

Comments

@starimpact
Copy link
Contributor

in ndarray Copy function:

        mshadow::Copy(to->FlatTo1D<cpu, DType>(),
                      from.FlatTo1D<cpu, DType>());

FlatTo1D will cause a int of Shape overflow problem. Because the Shape value type is index_t which range is on int. Sometimes, we need to handle a very big data, which size will exceed the range of int.
So... I suggest the following code:

        mshadow::Copy(to->FlatTo2D<cpu, DType>(),
                      from.FlatTo2D<cpu, DType>());

yes....just convert to 2D shape, we will be able to handle much bigger very big array data.

how about it?
@piiswrong

@starimpact starimpact changed the title FlatTo1D in copy Tensor is a problem! FlatTo1D in copy Tensor is a problem for huge array! Jan 5, 2018
@karan6181
Copy link
Contributor

@sandeep-krishnamurthy Can you please add labels:

  • NDArray
  • Call for Contribution

Thanks!

@marcoabreu marcoabreu added Backend Issues related to the backend of MXNet and removed C++ Related to C++ labels Jul 17, 2018
@apeforest
Copy link
Contributor

This issue has been fixed and verified. @sandeep-krishnamurthy Please close this issue. Thanks!

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
Backend Issues related to the backend of MXNet Call for Contribution NDArray
Projects
None yet
Development

No branches or pull requests

5 participants