business_transaction_map/calculate_metrics.py

import argparse
import pandas as pd
from pathlib import Path
from sklearn.metrics import f1_score

from common.constants import DEVICE
from common.constants import DO_NORMALIZATION
from common.constants import COLUMN_LABELS_STR
from common.constants import COLUMN_TEXT
from components.faiss_vector_database import FaissVectorDatabase
from components.embedding_extraction import EmbeddingExtractor


def main(query):

    global database
    global model

    # Здесь можно добавить любую обработку текста
    cleaned_text = query.replace("\n", " ")
    cleaned_text = 'query: ' + cleaned_text
    query_tokens = model.query_tokenization(cleaned_text)
    query_embeds = model.query_embed_extraction(query_tokens.to(DEVICE), DO_NORMALIZATION)[0]
    query_embeds = query_embeds[None, :]

    answer = database.search_transaction_map(query_embeds, K_NEIGHBORS)
    return answer[0]


K_NEIGHBORS = 1


if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--input_file',
                        type=Path,
                        default='../../data/csv/Карта проводок clear_3_3.pkl',
                        help='path to csv file.')

    args = parser.parse_args()

    global_model_path = 'intfloat/multilingual-e5-base'

    # db_files_path = '../../data/csv/Карта проводок clear_3_3.pkl'
    df = pd.read_pickle(args.input_file)
    database = FaissVectorDatabase(args.input_file)
    model = EmbeddingExtractor(global_model_path, DEVICE)

    true, pred = [], []
    for ind, row in df.iterrows():
        answer = main(row[COLUMN_TEXT])
        true.append(row[COLUMN_LABELS_STR])
        pred.append(answer[COLUMN_LABELS_STR])
        if row[COLUMN_LABELS_STR] != answer[COLUMN_LABELS_STR]:
            print(f'True labels: {row[COLUMN_LABELS_STR]}-----Pred labels: {answer[COLUMN_LABELS_STR]}')
            print('Название таблицы', row['TableName'])
            print(row['DocName'], '---', answer['doc_name'])
            print('-----------------------------------------')
    print('macro', f1_score(true, pred, average='macro'))
    print('micro', f1_score(true, pred, average='micro'))