Hey! This benchmark by Outerbounds made me curious about the nvidia triton inference server. • does BentoML still integrate with it? • If you deploy a bento using the integration is there reason to believe it performs close to, or as well as using the triton inference server by itself? Source: https://outerbounds.com/blog/the-many-ways-to-deploy-a-model/

I guess it also depends on the resource type the model consumes (GPU/CPU), in this graph they presented a comparison for serving language models with tensorrt optimization.