grid_search.py

import os
import subprocess
import utils.project_functions as pf

from tqdm import tqdm

root_path = os.path.dirname(os.path.abspath(__file__))
data_folder = os.path.join(root_path, 'data')
config_folder = os.path.join(root_path, 'config_files')

grid_configs_path = os.path.join(config_folder , 'grid_configs.yml')
current_config_file_path = os.path.join(config_folder , 'CurrentTest.yml')

metrics_file = os.path.join(data_folder, 'grid_search_metrics.csv')
current_training_output = os.path.join(data_folder, 'current_training_output.txt')

config_combinations, models_to_test = pf.get_configuration_combinations(grid_configs_path)
total_configs = len(config_combinations)
print('Número de configurações: ', total_configs)
input('Pressione \'Enter\' para continuar...\n')

pf.delete_old_grid_data(root_path)

best_metrics = {model: float('-inf') for model in models_to_test}

for config, ID in tqdm(zip(config_combinations, range(len(config_combinations))), total = total_configs):
    config['testID'] = 'GridTest_' + str(ID)
    
    pf.save_config(config, current_config_file_path)
    
    with open(current_training_output, 'w+') as output_file:
        subprocess.run(["python3", "model_training.py"], 
                    check=True,
                    stdout=output_file)
    
    best_metrics = pf.update_best_results(metrics_file, 
                                    best_metrics,
                                    config,
                                    config_folder,
                                    monitor = 'val_SDR',
                                    mode = 'max')