New week, new Rust! What are you folks up to?
I'm choosing a tech stack for my AI assistant and am currently torn between Rust and Go.
Creating small custom architecture AI in Google Colab using Rust :v A port of my Python code :]
I use Pytorch C++ binding, but the binding library doesn't has features parity, so I have to implement some of them manually from scratch. Now I'm trying to fix a bunch of error :]
I know there is pure Rust library like Candle, etc, but based on my browsing they have lower performance than C++ ones that has many optimizations since long ago. I wanted to use JAX binding, but I can't find one that is well documented and maintained. So I use native Pytorch C++ binding. I want to test will it increase the training speed compared to Python + Pytorch, because now both the dataset and training orchestration in native, so it is worth inspecting :] plus I haven't wrote Rust code for GPU before this because my GPU is just potato :[ that is great learning opportunity, I can learn more about GPU programming in Rust
Continuing to work on my "tablestg" crate, after a lot of false starts I think it is now more promising ( it isn't remotely stable yet ). I am exploring using hash tables instead of BTree-like data structures. See Store in tablestg::store - Rust