patchy631 / time-to-first-token

git clone https://github.com/patchy631/time-to-first-token.git

A 10-week, 30-minutes-a-day roadmap for LLM inference serving and optimization. vLLM, SGLang, quantization, speculative decoding, benchmarking.