Closed
Description
At the end of a test, guidellm cancels all pending requests. This causes the reported throughput to be lower depending on the portion of the overall test that the canceled requests cover. Instead we should record some kind of partial completions for the tokens completed within the deadline.