Contribute this project to original project pytorch.

[talregev]

It is possible to contribute this project to original project pytorch?

[artyom-beilis]

It is possible to contribute this project to original project pytorch?

Good question. 1. Pytorch supports Out-of-Tree (OOT) backends. For example intel extension for pytorch is still managed as out of tree backend - so it isn't strictly needed. 2. This backend is still lags behind significantly in comparison to other OOT backends like Intel one. So I wish it would become maintained by core torch team but realistically I don't see it happen in near future. The best would be gather more developers to contribute to this project and push it forward. Artyom

[talregev]

I am with you. I am not sure I will able to contribute code, but I would like to push this project forward, and attract more developers to this project.

I also believe that deep learning should be open for all hardware that exist, and opencl is one of the ways to do it.
I still don't understand why Intel nor AMD puts some money to import library like you that opencl will support in pytorch and tensorflow. This will bring millions to them, and open deep learning for many other devices and platforms.

For start (if it not to late) to try to import the dlprimitives to vcpkg. meaning that programers will able to compile it and able to contribute changes. Maybe it will attract more devs to this project.
But I will need that you will push a version tag in the dlprimitives. I will open an issue there as well.

Thank you for your answer!

[artyom-beilis]

I still don't understand why Intel nor AMD puts some money to import library like you 

- Because AMD are chasing cuda and developed rcom/hip that I must admit works  - Intel decided to go with their own idea syncl that nobody supports Each company looks for its narrow point of view and spend money where they see profits. I can't blame them for that, but I think they are missing the big picture Thanks!