Maybe the throughput between TGI and vLLM should be updated

[zhaoyang-star]

I just notice that TGI has support Paged Attention by integrating vLLM in TGI. The PR#516 has merged and be available in TGI v0.9.2.

[zhuohan123]

Hi! We are testing the performance of TGI and will update the performance results afterward. You can track the progress at #381