Pinned Loading

1. countdown-grpo countdown-grpo Public

   From-scratch GRPO (the RL method behind DeepSeek-R1) in pure PyTorch, replicating R1-Zero reasoning on the Countdown task

   Python 1

2. bbox-translator bbox-translator Public

   From-scratch discrete-token autoregressive transformer for drum-MIDI generation (unconditional + beatbox-conditioned)

   Python

3. Bridge Bridge Public

   Python