Skip to content

README.md

Latest commit

 

History

History
46 lines (35 loc) · 2.43 KB

README.md

File metadata and controls

46 lines (35 loc) · 2.43 KB

Model Development (Kaggle Spaceship Titanic)

This example demonstrates using Weco to optimize a Python script designed for the Spaceship Titanic Kaggle competition. The goal is to improve the model's accuracy metric by optimizing the train.py

Setup

  1. Ensure you are in the examples/spaceship-titanic directory.
  2. Install the CLI and dependencies for the example: 
    pip install weco pandas numpy scikit-learn torch xgboost lightgbm catboost

Run Weco

Run the following command to start optimizing the model:

weco run --source train.py \
     --eval-command "python evaluate.py --data-dir ./data --seed 0" \
     --metric accuracy \
     --goal maximize \
     --steps 10 \
     --model o4-mini \
     --additional-instructions "Improve feature engineering, model choice and hyper-parameters." \
     --log-dir .runs/spaceship-titanic

Explanation

  • --source train.py: The script provides a baseline as root node and directly optimize the train.py.
  • --eval-command "python evaluate.py --data-dir ./data/ --seed 0": The weco agent will run the evaluate.py.
    • [optional] --data-dir: path to the train and test data.
    • [optional] --seed: Seed for reproduce the experiment.
  • --metric accuracy: The target metric Weco should optimize.
  • --goal maximize: Weco aims to increase the accuracy.
  • --steps 10: The number of optimization iterations.
  • --model o4-mini: The LLM driving the optimization.
  • --additional-instructions "Improve feature engineering, model choice and hyper-parameters.": A simple instruction for model improvement or you can put the path to comptition_description.md within the repo to feed the agent more detailed information.
  • --log-dir .runs/spaceship-titanic: Specifies the directory where Weco should save logs and results for this run.

Weco will iteratively update the feature engineering or modeling code within train.py guided by the evaluation method defined in evaluate.py

Next Steps

With model development covered, you might be curious to see how you can make your AI code run faster, saving you time and more importantly GPU credits. Check out our example on automating kernel engineering in CUDA and Triton, or dive into the CLI Reference.