Python config.cfg.is_training() Examples

def evaluation(model, supervisor, num_label):
    teX, teY, num_te_batch = load_data(cfg.dataset, cfg.batch_size, is_training=False)
    fd_test_acc = save_to()
    with supervisor.managed_session(config=tf.ConfigProto(allow_soft_placement=True)) as sess:
        supervisor.saver.restore(sess, tf.train.latest_checkpoint(cfg.logdir))
        tf.logging.info('Model restored!')

        test_acc = 0
        for i in tqdm(range(num_te_batch), total=num_te_batch, ncols=70, leave=False, unit='b'):
            start = i * cfg.batch_size
            end = start + cfg.batch_size
            acc = sess.run(model.accuracy, {model.X: teX[start:end], model.labels: teY[start:end]})
            test_acc += acc
        test_acc = test_acc / (cfg.batch_size * num_te_batch)
        fd_test_acc.write(str(test_acc))
        fd_test_acc.close()
        print('Test accuracy has been saved to ' + cfg.results + '/test_acc.csv') 

def evaluation(model, supervisor, num_label):
    teX, teY, num_te_batch = load_data(cfg.dataset, cfg.batch_size, is_training=False)
    fd_test_acc = save_to()
    with supervisor.managed_session(config=tf.ConfigProto(allow_soft_placement=True)) as sess:
        supervisor.saver.restore(sess, tf.train.latest_checkpoint(cfg.logdir))
        tf.logging.info('Model restored!')

        test_acc = 0
        for i in tqdm(range(num_te_batch), total=num_te_batch, ncols=70, leave=False, unit='b'):
            start = i * cfg.batch_size
            end = start + cfg.batch_size
            acc = sess.run(model.accuracy, {model.X: teX[start:end], model.labels: teY[start:end]})
            test_acc += acc
        test_acc = test_acc / (cfg.batch_size * num_te_batch)
        fd_test_acc.write(str(test_acc))
        fd_test_acc.close()
        print('Test accuracy has been saved to ' + cfg.results + '/test_acc.csv') 

def save_to(is_training):
    os.makedirs(os.path.join(cfg.results_dir, "activations"), exist_ok=True)
    os.makedirs(os.path.join(cfg.results_dir, "timelines"), exist_ok=True)

    if is_training:
        loss = os.path.join(cfg.results_dir, 'loss.csv')
        train_acc = os.path.join(cfg.results_dir, 'train_acc.csv')
        val_acc = os.path.join(cfg.results_dir, 'val_acc.csv')

        if os.path.exists(val_acc):
            os.remove(val_acc)
        if os.path.exists(loss):
            os.remove(loss)
        if os.path.exists(train_acc):
            os.remove(train_acc)

        fd_train_acc = open(train_acc, 'w')
        fd_train_acc.write('step,train_acc\n')
        fd_loss = open(loss, 'w')
        fd_loss.write('step,loss\n')
        fd_val_acc = open(val_acc, 'w')
        fd_val_acc.write('step,val_acc\n')
        fd = {"train_acc": fd_train_acc,
              "loss": fd_loss,
              "val_acc": fd_val_acc}
    else:
        test_acc = os.path.join(cfg.results_dir, 'test_acc.csv')
        if os.path.exists(test_acc):
            os.remove(test_acc)
        fd_test_acc = open(test_acc, 'w')
        fd_test_acc.write('test_acc\n')
        fd = {"test_acc": fd_test_acc}

    return(fd) 

def evaluate(model, data_loader):
    # Setting up model
    test_iterator = data_loader(cfg.batch_size, mode="test")
    inputs = data_loader.next_element["images"]
    labels = data_loader.next_element["labels"]
    model.create_network(inputs, labels)

    # Create files to save evaluating results
    fd = save_to(is_training=False)
    saver = tf.train.Saver()

    config = tf.ConfigProto(allow_soft_placement=True)
    config.gpu_options.allow_growth = True
    with tf.Session(config=config) as sess:
        test_handle = sess.run(test_iterator.string_handle())
        saver.restore(sess, tf.train.latest_checkpoint(cfg.logdir))
        tf.logging.info('Model restored!')

        probs = []
        targets = []
        total_acc = 0
        n = 0
        while True:
            try:
                test_acc, prob, label = sess.run([model.accuracy, model.probs, labels], feed_dict={data_loader.handle: test_handle})
                probs.append(prob)
                targets.append(label)
                total_acc += test_acc
                n += 1
            except tf.errors.OutOfRangeError:
                break
        probs = np.concatenate(probs, axis=0)
        targets = np.concatenate(targets, axis=0).reshape((-1, 1))
        avg_acc = total_acc / n
        out_path = os.path.join(cfg.results_dir, 'prob_test.txt')
        np.savetxt(out_path, np.hstack((probs, targets)), fmt='%1.2f')
        print('Classification probability for each category has been saved to ' + out_path)
        fd["test_acc"].write(str(avg_acc))
        fd["test_acc"].close()
        out_path = os.path.join(cfg.results_dir, 'test_accuracy.txt')
        print('Test accuracy has been saved to ' + out_path) 

def main(_):
    model_list = ['baseline', 'vectorCapsNet', 'matrixCapsNet', 'convCapsNet']

    # Deciding which model to use
    if cfg.model == 'baseline':
        model = import_module(cfg.model).Model
    elif cfg.model in model_list:
        model = import_module(cfg.model).CapsNet
    else:
        raise ValueError('Unsupported model, please check the name of model:', cfg.model)

    # Deciding which dataset to use
    if cfg.dataset == 'mnist' or cfg.dataset == 'fashion-mnist':
        height = 28
        width = 28
        channels = 1
        num_label = 10
    elif cfg.dataset == 'smallNORB':
        num_label = 5
        height = 32
        width = 32
        channels = 1

    # Initializing model and data loader
    net = model(height=height, width=width, channels=channels, num_label=num_label)
    dataset = "capslayer.data.datasets." + cfg.dataset
    data_loader = import_module(dataset).DataLoader(path=cfg.data_dir,
                                                    splitting=cfg.splitting,
                                                    num_works=cfg.num_works)

    # Deciding to train or evaluate model
    if cfg.is_training:
        train(net, data_loader)
    else:
        evaluate(net, data_loader) 

def load_mnist(path, is_training):
    fd = open(os.path.join(cfg.dataset, 'train-images-idx3-ubyte'))
    loaded = np.fromfile(file=fd, dtype=np.uint8)
    trX = loaded[16:].reshape((60000, 28, 28, 1)).astype(np.float)

    fd = open(os.path.join(cfg.dataset, 'train-labels-idx1-ubyte'))
    loaded = np.fromfile(file=fd, dtype=np.uint8)
    trY = loaded[8:].reshape((60000)).astype(np.float)

    fd = open(os.path.join(cfg.dataset, 't10k-images-idx3-ubyte'))
    loaded = np.fromfile(file=fd, dtype=np.uint8)
    teX = loaded[16:].reshape((10000, 28, 28, 1)).astype(np.float)

    fd = open(os.path.join(cfg.dataset, 't10k-labels-idx1-ubyte'))
    loaded = np.fromfile(file=fd, dtype=np.uint8)
    teY = loaded[8:].reshape((10000)).astype(np.float)

    # normalization and convert to a tensor [60000, 28, 28, 1]
    trX = tf.convert_to_tensor(trX / 255., tf.float32)

    # => [num_samples, 10]
    trY = tf.one_hot(trY, depth=10, axis=1, dtype=tf.float32)
    teY = tf.one_hot(teY, depth=10, axis=1, dtype=tf.float32)

    if is_training:
        return trX, trY
    else:
        return teX / 255., teY 

def get_batch_data():
    trX, trY = load_mnist(cfg.dataset, cfg.is_training)

    data_queues = tf.train.slice_input_producer([trX, trY])
    X, Y = tf.train.shuffle_batch(data_queues, num_threads=cfg.num_threads,
                                  batch_size=cfg.batch_size,
                                  capacity=cfg.batch_size * 64,
                                  min_after_dequeue=cfg.batch_size * 32,
                                  allow_smaller_final_batch=False)

    return(X, Y) 

def main(_):
	
	# get dataset info
	result = create_image_lists(cfg.images)
	max_iters = len(result["train"]) * cfg.epoch // cfg.batch_size
	
	tf.logging.info('Loading Graph...')
	model = DFN(max_iters, batch_size=cfg.batch_size, init_lr=cfg.init_lr, power=cfg.power, momentum=cfg.momentum, stddev=cfg.stddev, regularization_scale=cfg.regularization_scale, alpha=cfg.alpha, gamma=cfg.gamma, fl_weight=cfg.fl_weight)
	tf.logging.info('Graph loaded.')
	
	if cfg.is_training:
		
		if not tf.gfile.Exists(cfg.logdir):
			
			tf.gfile.MakeDirs(cfg.logdir)
		
		if not tf.gfile.Exists(cfg.models):
			
			tf.gfile.MakeDirs(cfg.models)
		
		if os.path.exists(cfg.log):
			
			os.remove(cfg.log)
		
		fd = open(cfg.log, "a")
		tf.logging.info('Start training...')
		fd.write('Start training...\n')
		train(result, model, cfg.logdir, cfg.train_sum_freq, cfg.val_sum_freq, cfg.save_freq, cfg.models, fd)
		tf.logging.info('Training done.')
		fd.write('Training done.')
		fd.close()
	
	else:
		
		if not tf.gfile.Exists(cfg.test_outputs):
			
			tf.gfile.MakeDirs(cfg.test_outputs)
		
		tf.logging.info('Start testing...')
		test(result, model, cfg.models, cfg.test_outputs)
		tf.logging.info('Testing done.') 

def save_to():
    if not os.path.exists(cfg.results):
        os.mkdir(cfg.results)
    if cfg.is_training:
        loss = cfg.results + '/loss.csv'
        train_acc = cfg.results + '/train_acc.csv'
        val_acc = cfg.results + '/val_acc.csv'

        if os.path.exists(val_acc):
            os.remove(val_acc)
        if os.path.exists(loss):
            os.remove(loss)
        if os.path.exists(train_acc):
            os.remove(train_acc)

        fd_train_acc = open(train_acc, 'w')
        fd_train_acc.write('step,train_acc\n')
        fd_loss = open(loss, 'w')
        fd_loss.write('step,loss\n')
        fd_val_acc = open(val_acc, 'w')
        fd_val_acc.write('step,val_acc\n')
        return(fd_train_acc, fd_loss, fd_val_acc)
    else:
        test_acc = cfg.results + '/test_acc.csv'
        if os.path.exists(test_acc):
            os.remove(test_acc)
        fd_test_acc = open(test_acc, 'w')
        fd_test_acc.write('test_acc\n')
        return(fd_test_acc) 

def main(_):
    tf.logging.info(' Loading Graph...')
    num_label = 10
    model = CapsNet()
    tf.logging.info(' Graph loaded')

    sv = tf.train.Supervisor(graph=model.graph, logdir=cfg.logdir, save_model_secs=0)

    if cfg.is_training:
        tf.logging.info(' Start training...')
        train(model, sv, num_label)
        tf.logging.info('Training done')
    else:
        evaluation(model, sv, num_label) 

def create_inputs():
    trX, trY = load_mnist(cfg.dataset, cfg.is_training)

    num_pre_threads = cfg.thread_per_gpu*cfg.num_gpu
    data_queue = tf.train.slice_input_producer([trX, trY], capacity=64*num_pre_threads)
    X, Y = tf.train.shuffle_batch(data_queue, num_threads=num_pre_threads,
                                  batch_size=cfg.batch_size_per_gpu*cfg.num_gpu,
                                  capacity=cfg.batch_size_per_gpu*cfg.num_gpu * 64,
                                  min_after_dequeue=cfg.batch_size_per_gpu*cfg.num_gpu * 32,
                                  allow_smaller_final_batch=False)

    return (X, Y) 

def save_to():
    if not os.path.exists(cfg.results):
        os.mkdir(cfg.results)
    if cfg.is_training:
        loss = cfg.results + '/loss.csv'
        train_acc = cfg.results + '/train_acc.csv'
        val_acc = cfg.results + '/val_acc.csv'

        if os.path.exists(val_acc):
            os.remove(val_acc)
        if os.path.exists(loss):
            os.remove(loss)
        if os.path.exists(train_acc):
            os.remove(train_acc)

        fd_train_acc = open(train_acc, 'w')
        fd_train_acc.write('step,train_acc\n')
        fd_loss = open(loss, 'w')
        fd_loss.write('step,loss\n')
        fd_val_acc = open(val_acc, 'w')
        fd_val_acc.write('step,val_acc\n')
        return(fd_train_acc, fd_loss, fd_val_acc)
    else:
        test_acc = cfg.results + '/test_acc.csv'
        if os.path.exists(test_acc):
            os.remove(test_acc)
        fd_test_acc = open(test_acc, 'w')
        fd_test_acc.write('test_acc\n')
        return(fd_test_acc) 

def main(_):
    tf.logging.info(' Loading Graph...')
    num_label = 10
    model = CapsNet()
    tf.logging.info(' Graph loaded')

    sv = tf.train.Supervisor(graph=model.graph, logdir=cfg.logdir, save_model_secs=0)

    if cfg.is_training:
        tf.logging.info(' Start training...')
        train(model, sv, num_label)
        tf.logging.info('Training done')
    else:
        evaluation(model, sv, num_label) 

def create_inputs():
    trX, trY = load_mnist(cfg.dataset, cfg.is_training)

    num_pre_threads = cfg.thread_per_gpu*cfg.num_gpu
    data_queue = tf.train.slice_input_producer([trX, trY], capacity=64*num_pre_threads)
    X, Y = tf.train.shuffle_batch(data_queue, num_threads=num_pre_threads,
                                  batch_size=cfg.batch_size_per_gpu*cfg.num_gpu,
                                  capacity=cfg.batch_size_per_gpu*cfg.num_gpu * 64,
                                  min_after_dequeue=cfg.batch_size_per_gpu*cfg.num_gpu * 32,
                                  allow_smaller_final_batch=False)

    return (X, Y) 

def train(model, supervisor, num_label):
    trX, trY, num_tr_batch, valX, valY, num_val_batch = load_data(cfg.dataset, cfg.batch_size, is_training=True)
    Y = valY[:num_val_batch * cfg.batch_size].reshape((-1, 1))

    fd_train_acc, fd_loss, fd_val_acc = save_to()
    config = tf.ConfigProto()
    config.gpu_options.allow_growth = True
    with supervisor.managed_session(config=config) as sess:
        print("\nNote: all of results will be saved to directory: " + cfg.results)
        for epoch in range(cfg.epoch):
            print('Training for epoch ' + str(epoch) + '/' + str(cfg.epoch) + ':')
            if supervisor.should_stop():
                print('supervisor stoped!')
                break
            for step in tqdm(range(num_tr_batch), total=num_tr_batch, ncols=70, leave=False, unit='b'):
                start = step * cfg.batch_size
                end = start + cfg.batch_size
                global_step = epoch * num_tr_batch + step

                if global_step % cfg.train_sum_freq == 0:
                    _, loss, train_acc, summary_str = sess.run([model.train_op, model.total_loss, model.accuracy, model.train_summary])
                    assert not np.isnan(loss), 'Something wrong! loss is nan...'
                    supervisor.summary_writer.add_summary(summary_str, global_step)

                    fd_loss.write(str(global_step) + ',' + str(loss) + "\n")
                    fd_loss.flush()
                    fd_train_acc.write(str(global_step) + ',' + str(train_acc / cfg.batch_size) + "\n")
                    fd_train_acc.flush()
                else:
                    sess.run(model.train_op)

                if cfg.val_sum_freq != 0 and (global_step) % cfg.val_sum_freq == 0:
                    val_acc = 0
                    for i in range(num_val_batch):
                        start = i * cfg.batch_size
                        end = start + cfg.batch_size
                        acc = sess.run(model.accuracy, {model.X: valX[start:end], model.labels: valY[start:end]})
                        val_acc += acc
                    val_acc = val_acc / (cfg.batch_size * num_val_batch)
                    fd_val_acc.write(str(global_step) + ',' + str(val_acc) + '\n')
                    fd_val_acc.flush()

            if (epoch + 1) % cfg.save_freq == 0:
                supervisor.saver.save(sess, cfg.logdir + '/model_epoch_%04d_step_%02d' % (epoch, global_step))

        fd_val_acc.close()
        fd_train_acc.close()
        fd_loss.close() 

def train(model, supervisor, num_label):
    trX, trY, num_tr_batch, valX, valY, num_val_batch = load_data(cfg.dataset, cfg.batch_size, is_training=True)
    Y = valY[:num_val_batch * cfg.batch_size].reshape((-1, 1))

    fd_train_acc, fd_loss, fd_val_acc = save_to()
    config = tf.ConfigProto()
    config.gpu_options.allow_growth = True
    with supervisor.managed_session(config=config) as sess:
        print("\nNote: all of results will be saved to directory: " + cfg.results)
        for epoch in range(cfg.epoch):
            print("Training for epoch %d/%d:" % (epoch, cfg.epoch))
            if supervisor.should_stop():
                print('supervisor stoped!')
                break
            for step in tqdm(range(num_tr_batch), total=num_tr_batch, ncols=70, leave=False, unit='b'):
                start = step * cfg.batch_size
                end = start + cfg.batch_size
                global_step = epoch * num_tr_batch + step

                if global_step % cfg.train_sum_freq == 0:
                    _, loss, train_acc, summary_str = sess.run([model.train_op, model.total_loss, model.accuracy, model.train_summary])
                    assert not np.isnan(loss), 'Something wrong! loss is nan...'
                    supervisor.summary_writer.add_summary(summary_str, global_step)

                    fd_loss.write(str(global_step) + ',' + str(loss) + "\n")
                    fd_loss.flush()
                    fd_train_acc.write(str(global_step) + ',' + str(train_acc / cfg.batch_size) + "\n")
                    fd_train_acc.flush()
                else:
                    sess.run(model.train_op)

                if cfg.val_sum_freq != 0 and (global_step) % cfg.val_sum_freq == 0:
                    val_acc = 0
                    for i in range(num_val_batch):
                        start = i * cfg.batch_size
                        end = start + cfg.batch_size
                        acc = sess.run(model.accuracy, {model.X: valX[start:end], model.labels: valY[start:end]})
                        val_acc += acc
                    val_acc = val_acc / (cfg.batch_size * num_val_batch)
                    fd_val_acc.write(str(global_step) + ',' + str(val_acc) + '\n')
                    fd_val_acc.flush()

            if (epoch + 1) % cfg.save_freq == 0:
                supervisor.saver.save(sess, cfg.logdir + '/model_epoch_%04d_step_%02d' % (epoch, global_step))

        fd_val_acc.close()
        fd_train_acc.close()
        fd_loss.close()