Gradientzoo Python bindings
This is a Python library for Gradientzoo's API. Loading a pre-trained neural network is easy with Gradientzoo. Here's how easy it is to load a model with Tensorflow (full example below):
import tensorflow as tf
from gradientzoo import TensorflowGradientzoo
# (build MNIST graph here)
with tf.Session() as sess:
# Load latest weights from Gradientzoo
TensorflowGradientzoo('ericflo/mnist').load(sess)
# Graph is now ready to use!
Saving models is similarly straightforward:
import tensorflow as tf
from gradientzoo import TensorflowGradientzoo
# (build MNIST graph here)
with tf.Session() as sess:
for epoch in xrange(6):
# Train the model...
# Save the updated weights out to Gradientzoo
TensorflowGradientzoo('ericflo/mnist').save(sess)
Features
Supports saving models in Keras, variables in Tensorflow, and networks in Lasagne, and regular old files using Python with your framework of choice.
Installation
You don't need this source code unless you want to modify the package. If you just want to use the Gradientzoo Python bindings, you should run:
pip install --upgrade gradientzoo
or
easy_install --upgrade gradientzoo
See http://www.pip-installer.org/en/latest/index.html for instructions
on installing pip. If you are on a system with easy_install but not
pip, you can use easy_install instead. If you're not using virtualenv,
you may have to prefix those commands with sudo
. You can learn more
about virtualenv at http://www.virtualenv.org/
To install from source, run:
python setup.py install
Documentation
Please see http://python-gradientzoo.readthedocs.org/ for the most up-to-date documentation or visit a project page to see project-specific instructions, e.g. https://www.gradientzoo.com/ericflo/mnist
Contribute
- Issue Tracker: https://github.com/gradientzoo/python-gradientzoo/issues
- Source Code: https://github.com/gradientzoo/python-gradientzoo
Support
If you are having issues, please let us know at support@gradientzoo.com
Full Tensorflow Example
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data, mnist
from gradientzoo import TensorflowGradientzoo
learning_rate = 0.01
batch_size = 100
# Build MNIST graph
images_placeholder = tf.placeholder(tf.float32,
shape=(batch_size, mnist.IMAGE_PIXELS))
labels_placeholder = tf.placeholder(tf.int32, shape=(batch_size))
logits = mnist.inference(images_placeholder, 128, 32)
loss = mnist.loss(logits, labels_placeholder)
train_op = mnist.training(loss, learning_rate)
eval_correct = mnist.evaluation(logits, labels_placeholder)
# Start a Tensorflow session
with tf.Session() as sess:
# Load latest weights from Gradientzoo
TensorflowGradientzoo('ericflo/mnist').load(sess)
# Read in some data
data_sets = input_data.read_data_sets('data', False)
# Test the trained network on the dataset
true_count = 0
for step in xrange(data_sets.test.num_examples // batch_size):
images_feed, labels_feed = data_sets.test.next_batch(batch_size, False)
true_count += sess.run(eval_correct, feed_dict={
images_placeholder: images_feed,
labels_placeholder: labels_feed,
})
precision = true_count / float(data_sets.test.num_examples)
print('Num Examples: %d Num Correct: %d Precision: %0.04f' %
(data_sets.test.num_examples, true_count, precision))