Add Julia Bindings to Torch backend

[daddydrac]

🚀 Feature

Usage of Julia to access Torch backend api’s.

Motivation

Array Programming is a paradigm, just like saying “Object Orientated Programming” or “Functional Programming”, along w/ a compiler that can handle Dispatch and mathematical types makes it a language perfect for deep learning

Pitch

Create Julia bindings to access torch backend while leveraging the capabilities of the Julia language itself.

Additional context

Include GPU support for NVIDIA/CUDA. (Julia has this)

[mrshenli]

Marking this as "triage review" to collect feedbacks and comments on the pitch.

[daddydrac]

@mrshenli How can I help contrib if this initiative gets approved?

[c0nn3r]

@joehoeller Take a look here. Zygote.jl already binds with PyTorch's backend to take the derivative of Python functions.

[daddydrac]

@c0nn3r my understanding is that I can’t do everything in that, as say, PyTorch or Torch Mobile. And getting full on NVIDIA CUDA support in that is problematic in portions — correct me if I am wrong please?

[ezyang]

We're probably not going to have in tree Julia bindings, but if you have changes you need to make to PyTorch to make bindings easier let us know.

[daddydrac]

What would be the technical justification for it?

…

On Mon, Feb 3, 2020 at 12:13 PM Edward Z. Yang ***@***.***> wrote: We're probably not going to have in tree Julia bindings, but if you have changes you need to make to PyTorch to make bindings easier let us know. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#32698?email_source=notifications&email_token=ABHVQHA4HT7HJFYJ7XAOA7DRBBND5A5CNFSM4KMSCZDKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEKU2V6Y#issuecomment-581544699>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABHVQHHRVLSNGW6QR7RZYSLRBBND5ANCNFSM4KMSCZDA> .