RomanNum3ral
/
cai
mirror of https://github.com/aliasrobotics/cai.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

debug

This commit is contained in:
lidia9 2025-05-08 16:12:10 +02:00
parent 4f3be1d6f3
commit 8820ab3be8
1 changed files with 2 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
ci/benchmarks/.benchmarks.yml
View File

@ -36,7 +36,7 @@ benchmarks-test-seceval:
script:
- pip3 install -e .
- pip install -r benchmarks/seceval/eval/requirements.txt
- python3 benchmarks/seceval/eval/eval.py --dataset_file benchmarks/seceval/eval/datasets/questions-2.json --output_dir benchmarks/seceval/eval/outputs --backend ollama --models ollama/qwen2.5:14b
- python3 benchmarks/seceval/eval/eval.py --dataset_file benchmarks/seceval/eval/datasets/questions-2.json --output_dir benchmarks/seceval/eval/outputs --backend openrouter --models openrouter/qwen/qwen3-32b:free
benchmarks-test-cybermetric:
<<: *run_benchmarks
@ -46,9 +46,4 @@ benchmarks-test-cybermetric:
script:
- pip3 install -e .
- |
if [ -z "$OPENROUTER_API_KEY" ]; then
echo "Error: OPENROUTER_API_KEY is not set. Please configure it in GitLab CI/CD variables."
exit 1
fi
- python3 benchmarks/cybermetric/CyberMetric_evaluator.py --model_name openrouter/qwen/qwen3-32b:free --file_path benchmarks/cybermetric/CyberMetric-2-v1.json --api_key "$OPENROUTER_API_KEY"
- python3 benchmarks/cybermetric/CyberMetric_evaluator.py --model_name ollama/qwen2.5:14b --file_path benchmarks/cybermetric/CyberMetric-2-v1.json