Search for a command to run...
The Automated LLM Speedrunning Benchmark: Reproducing NanoGPT Improvements