This project demonstrates benchmarking of AI agents for date fruit classification using LLaMA 3.1-8B-instant model via the Groq API. The implementation includes comprehensive data analysis, feature extraction, model evaluation, and performance benchmarking within a Jupyter Notebook environment.
-
AI and ML Libraries:
groq: API connection to LLaMA 3.1-8B-instant modellanggraph: For building agent workflow graphssklearn: For data preprocessing, scaling, and evaluation metrics
-
Data Processing:
pandas: For dataset manipulation and analysisnumpy: For numerical operationsmatplotlib&seaborn: For data visualization and benchmark reporting
-
Utilities:
dotenv: For secure API key managementre: For text processing with regular expressionsjson: For benchmark data storagedatetime: For timestamping benchmark results
The project uses the Date Fruit Dataset (Date_Fruit_Datasets.xlsx) containing features such as:
- Area, Perimeter, Major/Minor Axis measurements
- Eccentricity, Solidity, Convex Area
- Texture and color features
- Classification labels (BERHI, DEGLET, DOKOL, etc.)
A sophisticated agent that processes and analyzes date fruit features:
- Feature preprocessing and scaling
- Analysis of fruit characteristics
- Classification into fruit categories
- Comprehensive reporting
Metrics tracked and visualized:
- Latency (processing time)
- Model response analysis
- Classification accuracy
- Feature importance
- Performance charts and metrics visualizations
- Benchmark summaries
- Classification distribution reports
- Feature analysis documentation
The benchmarking shows:
- Average analysis time: ~2-3 seconds for feature analysis
- Classification latency: ~4-5 seconds per sample
- Varying performance based on sample complexity
- Model accuracy evaluation against ground truth
- Benchmark JSON files in
reports/benchmark/ - Visualization charts in
reports/charts/ - Comprehensive analysis reports in
reports/
