| """CPU batch inference contract for organizer-run model evaluation.""" |
| import argparse |
|
|
| import numpy as np |
| import pandas as pd |
|
|
| from model import Model, validate_rows |
|
|
|
|
| def predict_file(input_path, output_path): |
| frame = pd.read_parquet(input_path) |
| if list(frame.columns) != ['id', 'smiles', 'sequence']: |
| raise ValueError('Expected input columns: id, smiles, sequence') |
| model = Model() |
| model.prepare(frame.sequence.to_numpy()) |
| predictions = [] |
| for start in range(0, len(frame), 128): |
| rows = validate_rows(frame.iloc[start:start + 128].to_dict('records')) |
| predictions.extend(model.predict(rows)) |
| if not np.isfinite(predictions).all(): |
| raise ValueError('Model returned non-finite predictions') |
| predictions = np.asarray(predictions) |
| if predictions.shape != (len(frame), 2): |
| raise ValueError('Model must return two logged predictions per input') |
| output = pd.DataFrame(predictions, columns=['log10_kcat', 'log10_km']) |
| output.insert(0, 'id', frame.id.to_numpy()) |
| output.to_csv(output_path, index=False) |
|
|
|
|
| if __name__ == '__main__': |
| parser = argparse.ArgumentParser() |
| parser.add_argument('--input', required=True) |
| parser.add_argument('--output', required=True) |
| args = parser.parse_args() |
| predict_file(args.input, args.output) |
|
|
