I think Rust is a promising language which is as fast as c++ and has much safety . Tensorflow has a rust-bind, I wish mxnet support rust api, its c binding has no cost.
i want to second this feature request!
although some rather outdated and unmaintained rust-bindings for mxnet already exist (mxnet-rs) and the utilization of mxnets c-bindings looks more ore less trival given rusts ffi capabilities (see: rust-bindgen), i would really like to see rust-bindings officially supported, reliable maintained and a little bit better documented.
beside the fact, that rust is becoming more and more popular as very fast and secure general purpose programming language, it's an exceptional efficient solution to generate webassambly components (see: https://github.com/rustwasm). that's a field of applications, where python based development approaches usually fail. but because of mxnets excellent low level binding capabilities resp. hybrid nature, DNN creation utilizing python based gluon and final deployment via rust resp. wasm, looks like a very attractive and promising combination to overcome this shortcomings.
we recently saw a lot of interest and more advanced efforts to solve this issue in the TVM project (tvm-rust, tvm.rs), which could be seen as another workaround to archive this goal, nevertheless i would see some kind of official rust support within the mxnet project as a very useful improvement.
Here's an attempt SunDoge/mxnet-rs and I'm looking for people to join forces.
Most helpful comment
i want to second this feature request!
although some rather outdated and unmaintained rust-bindings for mxnet already exist (mxnet-rs) and the utilization of mxnets c-bindings looks more ore less trival given rusts ffi capabilities (see: rust-bindgen), i would really like to see rust-bindings officially supported, reliable maintained and a little bit better documented.
beside the fact, that rust is becoming more and more popular as very fast and secure general purpose programming language, it's an exceptional efficient solution to generate webassambly components (see: https://github.com/rustwasm). that's a field of applications, where python based development approaches usually fail. but because of mxnets excellent low level binding capabilities resp. hybrid nature, DNN creation utilizing python based gluon and final deployment via rust resp. wasm, looks like a very attractive and promising combination to overcome this shortcomings.
we recently saw a lot of interest and more advanced efforts to solve this issue in the TVM project (tvm-rust, tvm.rs), which could be seen as another workaround to archive this goal, nevertheless i would see some kind of official rust support within the mxnet project as a very useful improvement.