Building an ML library from scratch in Rust

Steps:

• Implement a base tensor class (pretty chill)
• implement macros for reduction and binary operators (chill, but annoying stuff with generic types / keeping clean code and lifetimes)
• tons of time dying on memory issues for keeping a clean computational graph
  • exploring refcells
  • simple problem of calling backward on the reference
  • looking at l2
• now, looking into implementing einsum with proc macro -> seems tough
• implement matmul / autodiff