Am the only one hitting bug after bug after bug?

[dragetd]

Providing an api key breaks the csv renderer, the provided podman/docker command requires user permissions that shoudl be set via -e user=guidellm, the HTML output is ALWAYS broken, the console-table is lacking width and showing almost all info with an ellipsis...

And I sort of expected it to benchmark ALL models after it queries 'models', but instead if only benchmarks the first.

I do not really see the point of this utility.

I wanted a simple time-to-first-token, tps, prefix benchmark for all my models behind a local endpoint.

guidlellm is probably the completely wrong tool for this...?