evaluate inference speedup with radix KV cache

[michalharakal]

https://www.mintlify.com/sgl-project/mini-sglang/concepts/radix-cache
https://learn.deeplearning.ai/courses/efficient-inference-with-sglang-text-and-image-generation

[michalharakal]

SGLANG https://arxiv.org/pdf/2312.07104