src/evaluate.py

"""
Evaluation script - test trained models with sample reactions.
"""
from transformers import pipeline
from config import FORWARD_MODEL_NAME, RETRO_MODEL_NAME


def main():
    """Test forward and retro models."""
    print("=" * 60)
    print("ORD Reaction Model Evaluation")
    print("=" * 60)
    
    print(f"\nLoading forward model: {FORWARD_MODEL_NAME}")
    forward_pipe = pipeline("text2text-generation", model=FORWARD_MODEL_NAME)
    
    print(f"Loading retro model: {RETRO_MODEL_NAME}")
    retro_pipe = pipeline("text2text-generation", model=RETRO_MODEL_NAME)
    
    # Test examples
    forward_examples = [
        "CCO.COC(=O)C",  # ethanol + methyl acetate
        "c1ccccc1.Br",   # benzene + bromine
        "CC(=O)O.CCN",   # acetic acid + ethylamine
    ]
    
    retro_examples = [
        "CCOC(=O)C",     # ethyl acetate
        "c1ccccc1Br",    # bromobenzene
        "CC(=O)NCC",     # N-ethylacetamide
    ]
    
    print("\n" + "=" * 60)
    print("FORWARD SYNTHESIS TESTS")
    print("=" * 60)
    
    for reactants in forward_examples:
        print(f"\nInput (Reactants): {reactants}")
        try:
            result = forward_pipe(reactants, max_length=128, num_beams=5)
            predicted = result[0]["generated_text"]
            print(f"Predicted (Products): {predicted}")
        except Exception as e:
            print(f"Error: {e}")
    
    print("\n" + "=" * 60)
    print("RETROSYNTHESIS TESTS")
    print("=" * 60)
    
    for product in retro_examples:
        print(f"\nInput (Product): {product}")
        try:
            result = retro_pipe(product, max_length=256, num_beams=10)
            predicted = result[0]["generated_text"]
            print(f"Predicted (Reactants): {predicted}")
        except Exception as e:
            print(f"Error: {e}")
    
    print("\n" + "=" * 60)
    print("Evaluation complete!")
    print("=" * 60)


if __name__ == "__main__":
    main()