Martin Thoma Martin Thoma - 16 days ago 5
Python Question

How can I access restored weights of a model?

I would like to be able to access the restored weights of a model.

I know how to get the weights directly after training:

def weight_variable(shape):
initial = tf.truncated_normal(shape, stddev=0.1)
return tf.Variable(initial, name='weights')

with tf.variable_scope('conv1') as scope:
W_conv1 = weight_variable([5, 5, 1, 32])
[train]
weights_ = W_conv1.eval()
print(weights_)


However, the following does not work:

saver.restore(sess, model_checkpoint_path)
with tf.variable_scope('conv1') as scope_conv:
W_conv1 = tf.get_variable('weights', shape=[5, 5, 1, 32])
weights_ = W_conv1.eval()
print(weights_)


It gives

Model restored.
Traceback (most recent call last):
File "./tf_mnist.py", line 122, in <module>
weights_ = W_conv1.eval()
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variables.py", line 445, in eval
return self._variable.eval(session=session)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 559, in eval
return _eval_using_default_session(self, feed_dict, self.graph, session)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 3761, in _eval_using_default_session
return session.run(tensors, feed_dict)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 717, in run
run_metadata_ptr)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 915, in _run
feed_dict_string, options, run_metadata)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 965, in _do_run
target_list, options, run_metadata)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 985, in _do_call
raise type(e)(node_def, op, message)
tensorflow.python.framework.errors.FailedPreconditionError: Attempting to use uninitialized value conv1/weights_1
[[Node: conv1/weights_1/_52 = _Send[T=DT_FLOAT, client_terminated=false, recv_device="/job:localhost/replica:0/task:0/cpu:0", send_device="/job:localhost/replica:0/task:0/gpu:0", send_device_incarnation=1, tensor_name="edge_4_conv1/weights_1", _device="/job:localhost/replica:0/task:0/gpu:0"](conv1/weights_1)]]
[[Node: conv1/weights_1/_53 = _Recv[_start_time=0, client_terminated=false, recv_device="/job:localhost/replica:0/task:0/cpu:0", send_device="/job:localhost/replica:0/task:0/gpu:0", send_device_incarnation=1, tensor_name="edge_4_conv1/weights_1", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/cpu:0"]()]]


How can I access the restored model weights?

Complete code



Execute the following twice to see what I mean. The first time, the model will be trained and saved to disk (takes about 5 minutes after the data is downloaded). The second run restores the model.

#!/usr/bin/env python

"""MNIST with Tensorflow."""

from tensorflow.examples.tutorials.mnist import input_data
import tensorflow as tf

import os

epochs = 1000
model_checkpoint_path = 'checkpoints/mnist_tf_model.ckpt'


def weight_variable(shape):
initial = tf.truncated_normal(shape, stddev=0.1)
return tf.Variable(initial, name='weights')


def bias_variable(shape):
initial = tf.constant(0.1, shape=shape)
return tf.Variable(initial, name='biases')


def conv2d(x, W):
return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')


def max_pool_2x2(x):
return tf.nn.max_pool(x, ksize=[1, 2, 2, 1],
strides=[1, 2, 2, 1], padding='SAME')

mnist = input_data.read_data_sets('MNIST_data', one_hot=True)

with tf.Session() as sess:
x = tf.placeholder(tf.float32, shape=[None, 784])
y_ = tf.placeholder(tf.float32, shape=[None, 10])
x_image = tf.reshape(x, [-1, 28, 28, 1])

with tf.variable_scope('conv1') as scope:
W_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1, name='ReLU1')
h_pool1 = max_pool_2x2(h_conv1)

with tf.variable_scope('conv2') as scope:
W_conv2 = weight_variable([5, 5, 32, 64])
b_conv2 = bias_variable([64])
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2, name='ReLU2')
h_pool2 = max_pool_2x2(h_conv2)

with tf.variable_scope('fc1'):
W_fc1 = weight_variable([7 * 7 * 64, 1024])
b_fc1 = bias_variable([1024])

h_pool2_flat = tf.reshape(h_pool2, [-1, 7 * 7 * 64])
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1) + b_fc1)

with tf.variable_scope('dropout'):
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)

with tf.variable_scope('softmax'):
W_fc2 = weight_variable([1024, 10])
b_fc2 = bias_variable([10])

y_conv = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2) + b_fc2)

cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y_conv),
reduction_indices=[1]))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

# Add ops to save and restore all the variables.
saver = tf.train.Saver()

sess.run(tf.initialize_all_variables())
if not os.path.isfile(model_checkpoint_path):
for i in range(epochs):
batch = mnist.train.next_batch(50)
train_step.run(feed_dict={x: batch[0], y_: batch[1], keep_prob: 0.5})

# Save the variables to disk.
save_path = saver.save(sess, model_checkpoint_path)
print("Model saved in file: %s" % save_path)
weights_ = W_conv1.eval()
print(weights_)
else:
saver.restore(sess, model_checkpoint_path)
print("Model restored.")
summary_writer = tf.train.SummaryWriter('summary_dir', sess.graph)

with tf.variable_scope('conv1') as scope_conv:
W_conv1 = tf.get_variable('weights', shape=[5, 5, 1, 32])
weights_ = W_conv1.eval()
print(weights_)

Answer

You should not mix tf.get_variable and tf.Variable.

You should define variables at the beginning as:

def weight_variable(shape):
    initial = tf.truncated_normal(shape, stddev=0.1)
    return tf.get_variable('weights', initializer=initial)

and then read them like this:

with tf.variable_scope('conv1', reuse=True) as scope_conv:
        W_conv1 = tf.get_variable('weights', shape=[5, 5, 1, 32])
        weights_ = W_conv1.eval()
        print(weights_)

note the reuse=True argument.

Comments