Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 30 additions & 2 deletions .github/workflows/run-eval.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,8 @@ on:
branches:
- main
workflow_dispatch:
schedule:
- cron: '29 2 * * 1'

concurrency:
group: ${{ github.workflow }}-${{ github.ref || github.run_id }}
Expand All @@ -33,9 +35,35 @@ jobs:
run: |
bin/api/prepare.sh

- name: Get Available Model
id: get_model
env:
MODEL_URL: ${{ vars.MODEL_URL }}
MODEL_API_KEY: ${{ secrets.MODEL_API_KEY }}
run: |
echo "Fetching available models from $MODEL_URL/models"
RESPONSE=$(curl -s "$MODEL_URL/models" -H "Authorization: Bearer $MODEL_API_KEY")
echo "API Response: $RESPONSE"

DATA_LENGTH=$(echo "$RESPONSE" | jq -r '.data | length')
if [ "$DATA_LENGTH" = "null" ] || [ "$DATA_LENGTH" = "0" ]; then
echo "Error: No models available in the data array"
exit 1
fi

SELECTED_MODEL=$(echo "$RESPONSE" | jq -r '.data[0].id')

if [ "$SELECTED_MODEL" = "null" ] || [ -z "$SELECTED_MODEL" ]; then
echo "Error: Could not extract model ID from API response"
exit 1
fi

echo "Selected model: $SELECTED_MODEL"
echo "model_name=$SELECTED_MODEL" >> $GITHUB_OUTPUT

- name: Run API Evaluation Script
env:
MODEL_NAME: lfm-3b
MODEL_NAME: ${{ steps.get_model.outputs.model_name }}
MODEL_URL: ${{ vars.MODEL_URL }}
MODEL_API_KEY: ${{ secrets.MODEL_API_KEY }}
run: |
Expand Down Expand Up @@ -69,7 +97,7 @@ jobs:

- name: Run OpenAI Judge
env:
MODEL_NAME: lfm-3b
MODEL_NAME: ${{ steps.get_model.outputs.model_name }}
MODEL_URL: ${{ vars.MODEL_URL }}
MODEL_API_KEY: ${{ secrets.MODEL_API_KEY }}
run: |
Expand Down