diff --git a/.github/workflows/nightly-eval.yml b/.github/workflows/nightly-eval.yml index a39786611..809120c0c 100644 --- a/.github/workflows/nightly-eval.yml +++ b/.github/workflows/nightly-eval.yml @@ -27,9 +27,14 @@ jobs: bash scripts/ci_install_dependency.sh pip install --upgrade "evalplus[vllm] @ git+https://github.com/evalplus/evalplus" - - name: Nightly gsm8k and human eval Accuracy + - name: Test human eval timeout-minutes: 120 run: | cd test/srt python3 test_nightly_human_eval.py + + - name: Test gsm8k + timeout-minutes: 120 + run: | + cd test/srt python3 test_nightly_gsm8k_eval.py