title	MVPpred Lizard Performance Predictor
emoji	🦎
colorFrom	green
colorTo	blue
sdk	docker
app_port	8501

MVPpred: Lizard Performance Predictor

MVPpred predicts lizard performance traits from morphology measurements using trained machine-learning models.

Local Setup

1. Install uv

curl -LsSf https://astral.sh/uv/install.sh | sh

2. Install dependencies

uv add numpy pandas scikit-learn scipy xgboost joblib mlflow streamlit huggingface_hub

3. Activate environment

source .venv/bin/activate

Training Workflow

1. Find best K values for KNN imputation

uv run python scripts/find_best_k.py

This creates:

artifacts_inference/best_k.json
artifacts_inference/best_k_results.csv

Copy the selected K values into:

src/config.py

2. Run cross-validation evaluation

uv run python scripts/train_eval.py

3. Train final inference models

uv run python scripts/train_final.py

This saves model bundles into:

artifacts_inference/

Run the Streamlit App

uv run streamlit run scripts/app.py

MLflow

View training runs in the browser:

uv run mlflow ui -p 5003

Export results to CSV (saved to results/):

uv run python scripts/export_mlflow_results.py

Upload Artifacts to Hugging Face

uv run hf upload-large-folder wasicse/mvppred-artifacts ./artifacts_inference --repo-type model

Notes

• Use -1 for missing morphology values.
• Sparse targets may show lower confidence.
• Prediction bundles are downloaded from the Hugging Face model repository when missing locally.