Xingyao Wang 41a8bb3cf1
[eval,fix]: metrics get carried across eval instances (#3072)
* fix: make max_budget_per_task optional in `run_agent_controller`

* update arg for each run infer

* fix: metrics logging carried along; reset llm metric with the agent;

---------

Co-authored-by: Graham Neubig <neubig@gmail.com>
2024-07-23 03:30:28 +00:00
..