CS336 Assignment
From-scratch implementation of core Transformer language model components following the Stanford CS336 assignment series.
This project is a from-scratch implementation of core components of a Transformer language model, following the Stanford CS336 assignment series.
I implemented modules such as linear layers, embeddings, RMSNorm, SwiGLU, rotary position embeddings, multi-head self-attention, training utilities, optimizer logic, learning-rate scheduling, gradient clipping, and checkpointing.