Open
Description
Description
It would be great if we could enable automatic performance stabilization so they user would not have to estimate max-requests or max-seconds and potentially still not find the point where performance plateaus and stabilizes. it would be great to add in some automatic logic for - Something along the lines of ensuring a combination across requests and time.
Acceptance Criteria
Add in automated logic to ensure GuideLLM is able to find the performance stabilization point when running a benchmark.
Proposal:
Can set max-seconds to accept an auto
flag, and we can have it measure when the benchmarks have stabilized, then will stop sending new requests and times out/stops benchmarking.
Metadata
Metadata
Assignees
Labels
Type
Projects
Status
Ready