Create "docker model bench MODEL"

[ericcurtin]

It should be able to output the Tokens per Second of any give model with 1, 2, 4 and 8 concurrent requests. Provide a hyperfine-like experience.

[ilopezluna]

@mdelapenya is also working in a model benchmarking tool using DMR behind the scenes

[mdelapenya]

Indeed, I'm doing this as a preparation for ContainerDays London in February: https://github.com/mdelapenya/generative-ai-with-testcontainers/tree/slms-benchmarks