What's the multithreading story? tch-rs's (and libtorch's) lack of real threading support was causing problems for my async code - at least, I haven't found a w

What's the multithreading story? tch-rs's (and libtorch's) lack of real threading support was causing problems for my async code - at least, I haven't found a way to work around it yet, given my level of rust knowledge. Tangentially: I keep thinking an ML framework could be done based on Rust macros, so the automatic differentiation is done entirely at compile time. `let grad = autodiff!(x + y);` (The `autodiff` crate does this at runtime IIUC). Then the compiler's optimizations can apply to the static computation graph. I always thought that was the vision for TF in the early days, but clearly it moved in a more dynamic direction under PyTorch's influence.