benderthon

Set of utilities to work easier with Bender.

Currently there's support for TensorFlow and Caffe, but we are working on more stuff!

Works on Python 2.7.+ and 3.+, with Tensorflow 1.2+.

To install:

pip install benderthon

TensorFlow is required too. The simplest way to install it is:

pip install tensorflow

There are other ways, see Installing Tensorflow. Benderthon does not install it by default to let the usage of a custom installation.

tf-freeze

Utility to convert TensorFlow checkpoints into minimal frozen graphs.

Usage

From a checkpoint

To take the checkpoint checkpoint_path.ckpt, whose output is yielded by the node named Tanh, and save it to graph_with_weights.pb:

benderthon tf-freeze checkpoint_path.ckpt graph_with_weights.pb Tanh

From code

If you don't have a checkpoint or prefer to run it from code, this is the way to go. This is the same example as above but from code:

from benderthon import tf_freeze

// …

with tf.Session() as sess:
    // …

    tf_freeze.freeze(sess, 'graph_with_weights.pb', ['Tanh'])

Sample

The file sample.py contains a network example for MNIST dataset with 2 convolutional layers and 2 dense layers. If you run it, it will generate a minimal protobuf for with the weights frozen to run in Bender in output/mnist.pb:

./sample.py

The generated file occupies half the original checkpoints (26MB to 13MB).

The script will also generate checkpoints files with prefix checkpoints/mnist.ckpt. So you could have generated the protobuf from it:

benderthon tf-freeze checkpoints/mnist.ckpt output/mnist.pb Prediction

You can also get only the graph, which occupies just 13kB:

benderthon tf-freeze --no-weights checkpoints/mnist.ckpt output/mnist_only_graph.pb Prediction

To save the weights in a separate path for later processing:

benderthon tf-freeze --only-weights checkpoints/mnist.ckpt weights/ Prediction

caffe-freeze

This module cannot be accessed from the command line utility, it should be used from Python code, importing benderthon.caffe_freeze.

You need caffeflow package installed first:

pip install -e git://github.com/xmartlabs/caffeflow.git@4618f89#egg=caffeflow

Development

This utility is under development and the API is not stable. So, do not heavily rely on it.

To install locally you should do ./setup.py install, but first have pandoc and pypandoc installed.

License

Copyright 2017 Xmartlabs SRL.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

   http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.