eval:
  defaults:
    max_response_len: 16384
    top_p: 0.7
  datasets:
    - name: aime
      path: /root/aime-2024/aime-2024.jsonl
      rm_type: deepscaler
      n_samples_per_eval_prompt: 16
    - name: gpqa    # hf download --repo-type dataset zyzshishui0627/gpqa_diamond --local-dir /root/gpqa
      path: /root/gpqa/gpqa_eval.jsonl
      rm_type: gpqa
      n_samples_per_eval_prompt: 2
    - name: ifbench   # hf download --repo-type dataset zyzshishui0627/IFBench --local-dir /root/ifbench
      path: /root/ifbench/IFBench_eval.jsonl
      rm_type: ifbench
      n_samples_per_eval_prompt: 1