Show Your Work: Gemma2 Process Rewards
Google Tunix hackathon submission. GRPO and hybrid process rewards so Gemma2 2B shows its reasoning.
Projects
Qwen-style LM from scratch
Decoder-only transformer in PyTorch with GQA, RoPE, RMSNorm, and SwiGLU. Trained on Tiny Shakespeare.
ViT Image Classifier
Vision Transformer trained from scratch on CIFAR-10. Train, eval, inference, Gradio demo, Hub push.
BPE in Go
Byte-Pair Encoding from first principles: character tokens, pair counts, merges, and a growing subword vocab.