my goal was to train a Language model (SLM) entirely from scratch so no ML libraries allowed . so i gathered what's needed to make it happen : from tensor manipulation (views, operations , allocations) the autograd ( a DAG that retains the previous operations and inputs in order that led into a cert
