Query regarding workload estimations ?

[narendradev]

can we use it for estimation of workloads across gpu's ( A100, H100, H200, B200 ) without actually having infrastructure,to simulate different infrastructure resource estimates for workload planning ?

[FogDong]

Hi, currently the llm-optimizer requires real infrastructure to generate accurate optimization results.

[narendradev]

Thank you for response, It would be great addition if it could provide estimates based on inputs like ( Number of users, concurrency , throughput etc ) to estimate infrastructure required