train_eval_image_classifier.py

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import argparse
import os
import sys

def parse_args(check=True):
    print(sys.argv[1:]);
    parser = argparse.ArgumentParser()
    print('parser loaded')
    # train
    parser.add_argument('--dataset_name', type=str, default='quiz')
    parser.add_argument('--dataset_dir', type=str)
    parser.add_argument('--checkpoint_path', type=str)
    parser.add_argument('--model_name', type=str, default='inception_v4')
    parser.add_argument('--checkpoint_exclude_scopes', type=str, default='InceptionV4/Logits,InceptionV4/AuxLogits/Aux_logits')
    parser.add_argument('--train_dir', type=str)
    parser.add_argument('--learning_rate', type=float, default=0.001)
    parser.add_argument('--optimizer', type=str, default='rmsprop')
    parser.add_argument('--batch_size', type=int, default=32)
    parser.add_argument('--clone_on_cpu', type=bool, default=False)
    print('parser extra argument 1')
    # eval
    parser.add_argument('--dataset_split_name', type=str, default='validation')
    parser.add_argument('--eval_dir', type=str, default='validation')
    parser.add_argument('--max_num_batches', type=str, default='validation')

    print('parser extra argument 2')

    FLAGS, unparsed = parser.parse_known_args()

    print('parser extra argument finished')
    return FLAGS, unparsed


# train_cmd = 'python3 ./train_image_classifier.py  --dataset_name={dataset_name} --dataset_dir={dataset_dir} --checkpoint_path={checkpoint_path} --model_name={model_name} --checkpoint_exclude_scopes={checkpoint_exclude_scopes} --train_dir={train_dir} --learning_rate={learning_rate} --optimizer={optimizer} --batch_size={batch_size} --max_number_of_steps={max_number_of_steps} --clone_on_cpu={clone_on_cpu} '
train_cmd = 'python3 ./train_image_classifier.py  --dataset_name={dataset_name} --dataset_dir={dataset_dir} --model_name={model_name} --train_dir={train_dir} --learning_rate={learning_rate} --optimizer={optimizer} --batch_size={batch_size}'
eval_cmd = 'python3 ./eval_image_classifier.py --dataset_name={dataset_name} --dataset_dir={dataset_dir} --dataset_split_name={dataset_split_name} --model_name={model_name} --checkpoint_path={checkpoint_path}  --eval_dir={eval_dir} --batch_size={batch_size}  --max_num_batches={max_num_batches}'

if __name__ == '__main__':
    FLAGS, unparsed = parse_args()
    print('current working dir [{0}]'.format(os.getcwd()))
    w_d = os.path.dirname(os.path.abspath(__file__))
    print('change wording dir to [{0}]'.format(w_d))
    os.chdir(w_d)

    step_per_epoch = 70000 // FLAGS.batch_size
    for i in range(30):
        steps = int(step_per_epoch * (i + 1))
        # train 1 epoch
        print('################    train    ################')
        p = os.popen(train_cmd.format(**{'dataset_name': FLAGS.dataset_name, 'dataset_dir': FLAGS.dataset_dir,
                                         'checkpoint_path': FLAGS.checkpoint_path, 'model_name': FLAGS. model_name,
                                         'checkpoint_exclude_scopes': FLAGS.checkpoint_exclude_scopes, 'train_dir': FLAGS. train_dir,
                                         'learning_rate': FLAGS.learning_rate, 'optimizer': FLAGS.optimizer,
                                         'batch_size': FLAGS.batch_size, 'max_number_of_steps': steps, 'clone_on_cpu': FLAGS.clone_on_cpu}))
        for l in p:
            print(p.strip())

        # eval
        print('################    eval    ################')
        p = os.popen(eval_cmd.format(**{'dataset_name': FLAGS.dataset_name, 'dataset_dir': FLAGS.dataset_dir,
                                        'dataset_split_name': 'validation', 'model_name': FLAGS. model_name,
                                        'checkpoint_path': FLAGS.train_dir, 'batch_size': FLAGS.batch_size,
                                        'eval_dir': FLAGS. eval_dir, 'max_num_batches': FLAGS. max_num_batches}))
        for l in p:
            print(p.strip())