model_info.py

import os
import json


format_string = {
    'lstm': '''
        {0},
        {1[training_accuracy]},
        {1[best_validation_accuracy]},
        {1[best_epoch]},
        {2[model][seq2seq_encoder][hidden_size]},
        {2[model][seq2vec_encoder][hidden_size]},
        {2[model][regularizer][0][1][alpha]}''',
    'boe':
        '''{0},
        {1[training_accuracy]},
        {1[best_validation_accuracy]},
        {1[best_epoch]},
        {2[model][seq2seq_encoder][hidden_size]},
        {2[model][regularizer][0][1][alpha]}'''
}


def print_infos(model):
    model_dir = f'trained-{model}'

    print("name,training_acc,validation_acc,best_epoch,hidden1,hidden2,reg")
    for directory in os.listdir(model_dir):
        metrics_path = os.path.join(model_dir, directory, 'metrics.json')
        config_path = os.path.join(model_dir, directory, 'config.json')

        if not os.path.exists(metrics_path):
            continue  # Not fully trained

        with open(metrics_path) as metrics_file, \
            open(config_path) as config_file:
            metrics = json.load(metrics_file)
            config = json.load(config_file)

            print(format_string[model].replace(" ", "").replace("\n", "")
                  .format(directory, metrics, config))


print_infos('lstm')