如何正确保存 keras 模型以便能够使用 hub.Module() 加载？

Question

I am attempting to retrain inception v3 on a new image set.

When I try to save the model I receive an error.

I have tried:

    tf.keras.models.save_model(model, filename)

and

    model.save(filename)

and

    tf.contrib.saved_model.save_keras_model(model, filename)       

All give me a similar error, Module has no ' name '

I have attached my code relevant to the problem.

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

import argparse
import os
import sys

import numpy as np
import tensorflow as tf
import tensorflow_hub as hub
import matplotlib.pyplot as plt

FLAGS = None

def create_model(m, img_data):
    # load feature extractor (inception_v3)
    features_extractor_layer = tf.keras.layers.Lambda(m, input_shape=img_data.image_shape)

    # make pre-trained layers un-trainable
    features_extractor_layer.trainable = False

    print(features_extractor_layer.name)

    # add new activation layer to train to our classes
    model = tf.keras.Sequential([
        features_extractor_layer,
        tf.keras.layers.Dense(img_data.num_classes, activation='softmax')
    ])

    model.compile(optimizer='adam',
                  loss='categorical_crossentropy',
                  metrics=['accuracy'])

    return model

def get_and_gen_images(module):
    """
    get images from image directory or url

    :param module: module (to get required image size info
    :return: batched image data
    """
    data_name = os.path.splitext(os.path.basename(FLAGS.image_dir_or_url))[0]
    print("data: ", data_name)

    # download images to cache if not already
    if FLAGS.image_dir_or_url.startswith('https://'):
        data_root = tf.keras.utils.get_file(data_name,
                                            FLAGS.image_dir_or_url,
                                            untar=True,
                                            cache_dir=os.getcwd())
    else:   # specify directory with images
        data_root = tf.keras.utils.get_file(data_name,
                                            FLAGS.image_dir_or_url)

    # get image size for specific module
    image_size = hub.get_expected_image_size(module)


    # TODO: this is where to add noise, rotations, shifts, etc.
    image_generator = tf.keras.preprocessing.image.ImageDataGenerator(rescale=1. / 255, validation_split=0.2)

    # create image stream
    train_image_data = image_generator.flow_from_directory(str(data_root),
                                                           target_size=image_size,
                                                           batch_size=FLAGS.batch_size,
                                                           subset='training')

    validation_image_data = image_generator.flow_from_directory(str(data_root),
                                                                target_size=image_size,
                                                                batch_size=FLAGS.batch_size,
                                                                subset='validation')

    return train_image_data, validation_image_data

# load module (will download from url or directory_
module = hub.Module(FLAGS.tfhub_module)

# generate image stream
train_image_data, validation_image_data = get_and_gen_images(module)

model = create_model(module, train_image_data)
model.summary()

file = FLAGS.saved_model_dir + "/modelname.h5"

model.save(file)

This should save a ".h5" model file, but I receive a naming error:

Traceback (most recent call last):
  File "/home/raphy/projects/vmi/tf_cpu/retrain.py", line 305, in <module>
    tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)
  File "/home/raphy/.local/lib/python3.6/site-packages/tensorflow/python/platform/app.py", line 125, in run
    _sys.exit(main(argv))
  File "/home/raphy/projects/vmi/tf_cpu/retrain.py", line 205, in main
    model.save(file)
  File "/home/raphy/.local/lib/python3.6/site-packages/tensorflow/python/keras/engine/sequential.py", line 319, in save
    save_model(self, filepath, overwrite, include_optimizer)
  File "/home/raphy/.local/lib/python3.6/site-packages/tensorflow/python/keras/engine/saving.py", line 105, in save_model
    'config': model.get_config()
  File "/home/raphy/.local/lib/python3.6/site-packages/tensorflow/python/keras/engine/sequential.py", line 326, in get_config
    'config': layer.get_config()
  File "/home/raphy/.local/lib/python3.6/site-packages/tensorflow/python/keras/layers/core.py", line 756, in get_config
    function = self.function.__name__
AttributeError: 'Module' object has no attribute '__name__'

I want to save the model in the format of the tf_hub models.

Answer 1

As specified in Hosting TF Hub TF Org Link ,

If you are interested in hosting your own repository of models that are loadable with the tensorflow_hub library, your HTTP distribution service should follow the following protocol.

In other words, you cannot load any Model using TF Hub, but you can load only the Modules present in the TF Hub Modules Site .

If you want to Load your Saved Model, you can do it using tf.saved_model.load .

But if you want to do it using TF Hub, please refer this link .

Also, mentioning the instructions below just in case if the link doesn't work:

Hosting your own models :

TensorFlow Hub provides an open repository of trained models at thub.dev . The tensorflow_hub library can load models from this repository and other HTTP based repositories of machine learning models. In particular the protocol allows to use the URL identifying the model both for the documentation of the model and the endpoint to fetch the model.

If you are interested in hosting your own repository of models that are loadable with the tensorflow_hub library, your HTTP distribution service should follow the following protocol.

Protocol:

When a URL such as https://example.com/model is used to identify a model to load or instantiate, the model resolver will attempt to download a compressed tarball from the URL after appending a query parameter? tf-hub-format=compressed .

The query param is to be interpreted as a comma separated list of the model formats that the client is interested in. For now only the "compressed" format is defined.

The compressed format indicates that the client expects a tar.gz archive with the model contents. The root of the archive is the root of the model directory and should contain a SavedModel, as in this example:

# Create a compressed model from a SavedModel directory.
$ tar -cz -f model.tar.gz --owner=0 --group=0 -C /tmp/export-model/ .

# Inspect files inside a compressed model
$ tar -tf model.tar.gz
./
./variables/
./variables/variables.data-00000-of-00001
./variables/variables.index
./assets/
./saved_model.pb

Tarballs for use with the deprecated hub.Module() API from TF1 will also contain a./tfhub_module.pb file. The hub.load() API for TF2 SavedModels ignores such a file.

The tensorflow_hub library expects that model URLs are versioned and that the model content of a given version is immutable, so that it can be cached indefinitely.