CPU training don't support multi-processor ?

[coderboyisongithub]

Tinytorch doesn't support multicore utilization while training on cpu? or there is option disabled by default.

[keith2018]

Currently, the CPU version of operators are all simple naive implementations, without SIMD or multithreading optimizations, for example, GEMM is implemented with just simple for-loops. At the moment, I am mainly optimizing the CUDA version (for LLM inference), and may optimize the CPU version later on.