README_RUN.md

🎯 READY TO RUN: Tests & Plots

✅ Everything is Ready!

All scripts are created and ready to execute. Here's what to do:

🐳 In Docker Container (Recommended)

# Navigate to project
cd /workspace/edgeeda-agent

# Run everything at once
./run_all.sh

That's it! The script will:

1. ✅ Run all agent tests
2. ✅ Generate comprehensive plots
3. ✅ Show summary of results

📝 What Gets Generated

Test Results

• Console output with ✅/❌ for each test
• Summary: "X/Y tests passed"

Plots (in runs/plots/)

• trials.csv - All data in CSV format
• success_rate.png - How success rate improves over time
• runtime_distribution.png - Histogram of how long trials take
• return_code_distribution.png - Breakdown of success/failure codes
• knob_*.png - Analysis of each knob's values across trials
• learning_curve.png - Best reward progression (if rewards available)
• area_vs_wns.png - Pareto front visualization (if metrics available)

🔍 Quick Check

Before running, verify:

# Check database exists
ls -la runs/experiment.sqlite

# Check scripts are executable
ls -la run_all.sh run_tests_simple.py generate_plots.py

🚨 If Something Goes Wrong

1. "No such file or directory"

   • Make sure you're in /workspace/edgeeda-agent
   • Check: pwd

2. "Permission denied"

   • Fix: chmod +x run_all.sh

3. "Module not found"

   • You're not in Docker - enter container first
   • Or install: pip3 install -e .

4. "Database not found"

   • You need to run experiments first
   • Run: edgeeda tune --config configs/gcd_nangate45.yaml --budget 5

📤 Getting Results Out

# Copy plots to host (from host machine)
docker cp <CONTAINER_ID>:/workspace/edgeeda-agent/runs/plots ./runs/

# Or use shared volume if configured

🎉 That's It!

Just run ./run_all.sh in Docker and you're done!