diff options
10 files changed, 72 insertions, 64 deletions
diff --git a/tensorflow/examples/how_tos/reading_data/fully_connected_preloaded.py b/tensorflow/examples/how_tos/reading_data/fully_connected_preloaded.py index d0482568a2..7795248f82 100644 --- a/tensorflow/examples/how_tos/reading_data/fully_connected_preloaded.py +++ b/tensorflow/examples/how_tos/reading_data/fully_connected_preloaded.py @@ -75,7 +75,7 @@ def run_training(): eval_correct = mnist.evaluation(logits, labels) # Build the summary operation based on the TF collection of Summaries. - summary_op = tf.merge_all_summaries() + summary_op = tf.summary.merge_all() # Create a saver for writing training checkpoints. saver = tf.train.Saver() diff --git a/tensorflow/examples/how_tos/reading_data/fully_connected_preloaded_var.py b/tensorflow/examples/how_tos/reading_data/fully_connected_preloaded_var.py index d037b8731c..5325afbe60 100644 --- a/tensorflow/examples/how_tos/reading_data/fully_connected_preloaded_var.py +++ b/tensorflow/examples/how_tos/reading_data/fully_connected_preloaded_var.py @@ -81,7 +81,7 @@ def run_training(): eval_correct = mnist.evaluation(logits, labels) # Build the summary operation based on the TF collection of Summaries. - summary_op = tf.merge_all_summaries() + summary_op = tf.summary.merge_all() # Create a saver for writing training checkpoints. saver = tf.train.Saver() diff --git a/tensorflow/examples/image_retraining/retrain.py b/tensorflow/examples/image_retraining/retrain.py index d52a23fd15..4f06cb8add 100644 --- a/tensorflow/examples/image_retraining/retrain.py +++ b/tensorflow/examples/image_retraining/retrain.py @@ -647,17 +647,17 @@ def add_input_distortions(flip_left_right, random_crop, random_scale, return jpeg_data, distort_result -def variable_summaries(var, name): +def variable_summaries(var): """Attach a lot of summaries to a Tensor (for TensorBoard visualization).""" with tf.name_scope('summaries'): mean = tf.reduce_mean(var) - tf.scalar_summary('mean/' + name, mean) + tf.summary.scalar('mean', mean) with tf.name_scope('stddev'): stddev = tf.sqrt(tf.reduce_mean(tf.square(var - mean))) - tf.scalar_summary('stddev/' + name, stddev) - tf.scalar_summary('max/' + name, tf.reduce_max(var)) - tf.scalar_summary('min/' + name, tf.reduce_min(var)) - tf.histogram_summary(name, var) + tf.summary.scalar('stddev', stddev) + tf.summary.scalar('max', tf.reduce_max(var)) + tf.summary.scalar('min', tf.reduce_min(var)) + tf.summary.histogram('histogram', var) def add_final_training_ops(class_count, final_tensor_name, bottleneck_tensor): @@ -695,23 +695,23 @@ def add_final_training_ops(class_count, final_tensor_name, bottleneck_tensor): with tf.name_scope(layer_name): with tf.name_scope('weights'): layer_weights = tf.Variable(tf.truncated_normal([BOTTLENECK_TENSOR_SIZE, class_count], stddev=0.001), name='final_weights') - variable_summaries(layer_weights, layer_name + '/weights') + variable_summaries(layer_weights) with tf.name_scope('biases'): layer_biases = tf.Variable(tf.zeros([class_count]), name='final_biases') - variable_summaries(layer_biases, layer_name + '/biases') + variable_summaries(layer_biases) with tf.name_scope('Wx_plus_b'): logits = tf.matmul(bottleneck_input, layer_weights) + layer_biases - tf.histogram_summary(layer_name + '/pre_activations', logits) + tf.summary.histogram('pre_activations', logits) final_tensor = tf.nn.softmax(logits, name=final_tensor_name) - tf.histogram_summary(final_tensor_name + '/activations', final_tensor) + tf.summary.histogram('activations', final_tensor) with tf.name_scope('cross_entropy'): cross_entropy = tf.nn.softmax_cross_entropy_with_logits( logits, ground_truth_input) with tf.name_scope('total'): cross_entropy_mean = tf.reduce_mean(cross_entropy) - tf.scalar_summary('cross entropy', cross_entropy_mean) + tf.summary.scalar('cross_entropy', cross_entropy_mean) with tf.name_scope('train'): train_step = tf.train.GradientDescentOptimizer(FLAGS.learning_rate).minimize( @@ -738,7 +738,7 @@ def add_evaluation_step(result_tensor, ground_truth_tensor): tf.argmax(ground_truth_tensor, 1)) with tf.name_scope('accuracy'): evaluation_step = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) - tf.scalar_summary('accuracy', evaluation_step) + tf.summary.scalar('accuracy', evaluation_step) return evaluation_step @@ -792,7 +792,7 @@ def main(_): evaluation_step = add_evaluation_step(final_tensor, ground_truth_input) # Merge all the summaries and write them out to /tmp/retrain_logs (by default) - merged = tf.merge_all_summaries() + merged = tf.summary.merge_all() train_writer = tf.train.SummaryWriter(FLAGS.summaries_dir + '/train', sess.graph) validation_writer = tf.train.SummaryWriter(FLAGS.summaries_dir + '/validation') diff --git a/tensorflow/examples/tutorials/mnist/fully_connected_feed.py b/tensorflow/examples/tutorials/mnist/fully_connected_feed.py index 38ae88ee5b..4dbd43527b 100644 --- a/tensorflow/examples/tutorials/mnist/fully_connected_feed.py +++ b/tensorflow/examples/tutorials/mnist/fully_connected_feed.py @@ -139,7 +139,7 @@ def run_training(): eval_correct = mnist.evaluation(logits, labels_placeholder) # Build the summary Tensor based on the TF collection of Summaries. - summary = tf.merge_all_summaries() + summary = tf.summary.merge_all() # Add the variable initializer Op. init = tf.initialize_all_variables() diff --git a/tensorflow/examples/tutorials/mnist/mnist.py b/tensorflow/examples/tutorials/mnist/mnist.py index 4960a55a7c..e97a6c48ef 100644 --- a/tensorflow/examples/tutorials/mnist/mnist.py +++ b/tensorflow/examples/tutorials/mnist/mnist.py @@ -118,7 +118,7 @@ def training(loss, learning_rate): train_op: The Op for training. """ # Add a scalar summary for the snapshot loss. - tf.scalar_summary(loss.op.name, loss) + tf.summary.scalar('loss', loss) # Create the gradient descent optimizer with the given learning rate. optimizer = tf.train.GradientDescentOptimizer(learning_rate) # Create a variable to track the global step. diff --git a/tensorflow/examples/tutorials/mnist/mnist_with_summaries.py b/tensorflow/examples/tutorials/mnist/mnist_with_summaries.py index f1eff29718..0597d5149b 100644 --- a/tensorflow/examples/tutorials/mnist/mnist_with_summaries.py +++ b/tensorflow/examples/tutorials/mnist/mnist_with_summaries.py @@ -40,7 +40,6 @@ def train(): fake_data=FLAGS.fake_data) sess = tf.InteractiveSession() - # Create a multilayer model. # Input placeholders @@ -50,7 +49,7 @@ def train(): with tf.name_scope('input_reshape'): image_shaped_input = tf.reshape(x, [-1, 28, 28, 1]) - tf.image_summary('input', image_shaped_input, 10) + tf.summary.image('input', image_shaped_input, 10) # We can't initialize these variables to 0 - the network will get stuck. def weight_variable(shape): @@ -63,17 +62,17 @@ def train(): initial = tf.constant(0.1, shape=shape) return tf.Variable(initial) - def variable_summaries(var, name): - """Attach a lot of summaries to a Tensor.""" + def variable_summaries(var): + """Attach a lot of summaries to a Tensor (for TensorBoard visualization).""" with tf.name_scope('summaries'): mean = tf.reduce_mean(var) - tf.scalar_summary('mean/' + name, mean) + tf.summary.scalar('mean', mean) with tf.name_scope('stddev'): stddev = tf.sqrt(tf.reduce_mean(tf.square(var - mean))) - tf.scalar_summary('stddev/' + name, stddev) - tf.scalar_summary('max/' + name, tf.reduce_max(var)) - tf.scalar_summary('min/' + name, tf.reduce_min(var)) - tf.histogram_summary(name, var) + tf.summary.scalar('stddev', stddev) + tf.summary.scalar('max', tf.reduce_max(var)) + tf.summary.scalar('min', tf.reduce_min(var)) + tf.summary.histogram('histogram', var) def nn_layer(input_tensor, input_dim, output_dim, layer_name, act=tf.nn.relu): """Reusable code for making a simple neural net layer. @@ -87,22 +86,22 @@ def train(): # This Variable will hold the state of the weights for the layer with tf.name_scope('weights'): weights = weight_variable([input_dim, output_dim]) - variable_summaries(weights, layer_name + '/weights') + variable_summaries(weights) with tf.name_scope('biases'): biases = bias_variable([output_dim]) - variable_summaries(biases, layer_name + '/biases') + variable_summaries(biases) with tf.name_scope('Wx_plus_b'): preactivate = tf.matmul(input_tensor, weights) + biases - tf.histogram_summary(layer_name + '/pre_activations', preactivate) + tf.summary.histogram('pre_activations', preactivate) activations = act(preactivate, name='activation') - tf.histogram_summary(layer_name + '/activations', activations) + tf.summary.histogram('activations', activations) return activations hidden1 = nn_layer(x, 784, 500, 'layer1') with tf.name_scope('dropout'): keep_prob = tf.placeholder(tf.float32) - tf.scalar_summary('dropout_keep_probability', keep_prob) + tf.summary.scalar('dropout_keep_probability', keep_prob) dropped = tf.nn.dropout(hidden1, keep_prob) # Do not apply softmax activation yet, see below. @@ -122,7 +121,7 @@ def train(): diff = tf.nn.softmax_cross_entropy_with_logits(y, y_) with tf.name_scope('total'): cross_entropy = tf.reduce_mean(diff) - tf.scalar_summary('cross entropy', cross_entropy) + tf.summary.scalar('cross_entropy', cross_entropy) with tf.name_scope('train'): train_step = tf.train.AdamOptimizer(FLAGS.learning_rate).minimize( @@ -133,10 +132,10 @@ def train(): correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1)) with tf.name_scope('accuracy'): accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) - tf.scalar_summary('accuracy', accuracy) + tf.summary.scalar('accuracy', accuracy) # Merge all the summaries and write them out to /tmp/mnist_logs (by default) - merged = tf.merge_all_summaries() + merged = tf.summary.merge_all() train_writer = tf.train.SummaryWriter(FLAGS.summaries_dir + '/train', sess.graph) test_writer = tf.train.SummaryWriter(FLAGS.summaries_dir + '/test') diff --git a/tensorflow/g3doc/how_tos/distributed/index.md b/tensorflow/g3doc/how_tos/distributed/index.md index b5ad1991b2..96f5a548ee 100644 --- a/tensorflow/g3doc/how_tos/distributed/index.md +++ b/tensorflow/g3doc/how_tos/distributed/index.md @@ -226,7 +226,7 @@ def main(_): loss, global_step=global_step) saver = tf.train.Saver() - summary_op = tf.merge_all_summaries() + summary_op = tf.summary.merge_all() init_op = tf.initialize_all_variables() # Create a "supervisor", which oversees the training process. diff --git a/tensorflow/g3doc/how_tos/meta_graph/index.md b/tensorflow/g3doc/how_tos/meta_graph/index.md index e2af0d49b9..a7bce5101c 100644 --- a/tensorflow/g3doc/how_tos/meta_graph/index.md +++ b/tensorflow/g3doc/how_tos/meta_graph/index.md @@ -219,7 +219,7 @@ Here are some of the typical usage models: name="xentropy") loss = tf.reduce_mean(cross_entropy, name="xentropy_mean") - tf.scalar_summary(loss.op.name, loss) + tf.summary.scalar('loss', loss) # Creates the gradient descent optimizer with the given learning rate. optimizer = tf.train.GradientDescentOptimizer(0.01) diff --git a/tensorflow/g3doc/how_tos/summaries_and_tensorboard/index.md b/tensorflow/g3doc/how_tos/summaries_and_tensorboard/index.md index 9132132879..1f90214ae8 100644 --- a/tensorflow/g3doc/how_tos/summaries_and_tensorboard/index.md +++ b/tensorflow/g3doc/how_tos/summaries_and_tensorboard/index.md @@ -24,12 +24,12 @@ lifecycle for summary data within TensorBoard. First, create the TensorFlow graph that you'd like to collect summary data from, and decide which nodes you would like to annotate with -[summary operations](../../api_docs/python/train.md#summary-operations). +[summary operations](../../api_docs/python/summary.md). For example, suppose you are training a convolutional neural network for recognizing MNIST digits. You'd like to record how the learning rate varies over time, and how the objective function is changing. Collect these by -attaching [`scalar_summary`](../../api_docs/python/train.md#scalar_summary) ops +attaching [`scalar_summary`](../../api_docs/python/summary.md#scalar) ops to the nodes that output the learning rate and loss respectively. Then, give each `scalar_summary` a meaningful `tag`, like `'learning rate'` or `'loss function'`. @@ -37,18 +37,18 @@ function'`. Perhaps you'd also like to visualize the distributions of activations coming off a particular layer, or the distribution of gradients or weights. Collect this data by attaching -[`histogram_summary`](../../api_docs/python/train.md#histogram_summary) ops to +[`histogram_summary`](../../api_docs/python/summary.md#histogram) ops to the gradient outputs and to the variable that holds your weights, respectively. For details on all of the summary operations available, check out the docs on -[summary operations](../../api_docs/python/train.md#summary-operations). +[summary operations](../../api_docs/python/summary.md). Operations in TensorFlow don't do anything until you run them, or an op that depends on their output. And the summary nodes that we've just created are peripheral to your graph: none of the ops you are currently running depend on them. So, to generate summaries, we need to run all of these summary nodes. Managing them by hand would be tedious, so use -[`tf.merge_all_summaries`](../../api_docs/python/train.md#merge_all_summaries) +[`tf.summary.merge_all`](../../api_docs/python/summary.md#merge_all) to combine them into a single op that generates all the summary data. Then, you can just run the merged summary op, which will generate a serialized @@ -79,17 +79,17 @@ training. The code below is an excerpt; full source is [here](https://www.tensorflow.org/code/tensorflow/examples/tutorials/mnist/mnist_with_summaries.py). ```python -def variable_summaries(var, name): - """Attach a lot of summaries to a Tensor.""" +def variable_summaries(var): + """Attach a lot of summaries to a Tensor (for TensorBoard visualization).""" with tf.name_scope('summaries'): mean = tf.reduce_mean(var) - tf.scalar_summary('mean/' + name, mean) + tf.summary.scalar('mean', mean) with tf.name_scope('stddev'): stddev = tf.sqrt(tf.reduce_mean(tf.square(var - mean))) - tf.scalar_summary('stddev/' + name, stddev) - tf.scalar_summary('max/' + name, tf.reduce_max(var)) - tf.scalar_summary('min/' + name, tf.reduce_min(var)) - tf.histogram_summary(name, var) + tf.summary.scalar('stddev', stddev) + tf.summary.scalar('max', tf.reduce_max(var)) + tf.summary.scalar('min', tf.reduce_min(var)) + tf.summary.histogram('histogram', var) def nn_layer(input_tensor, input_dim, output_dim, layer_name, act=tf.nn.relu): """Reusable code for making a simple neural net layer. @@ -103,31 +103,42 @@ def nn_layer(input_tensor, input_dim, output_dim, layer_name, act=tf.nn.relu): # This Variable will hold the state of the weights for the layer with tf.name_scope('weights'): weights = weight_variable([input_dim, output_dim]) - variable_summaries(weights, layer_name + '/weights') + variable_summaries(weights) with tf.name_scope('biases'): biases = bias_variable([output_dim]) - variable_summaries(biases, layer_name + '/biases') + variable_summaries(biases) with tf.name_scope('Wx_plus_b'): preactivate = tf.matmul(input_tensor, weights) + biases - tf.histogram_summary(layer_name + '/pre_activations', preactivate) - activations = act(preactivate, 'activation') - tf.histogram_summary(layer_name + '/activations', activations) + tf.summary.histogram('pre_activations', preactivate) + activations = act(preactivate, name='activation') + tf.summary.histogram('activations', activations) return activations hidden1 = nn_layer(x, 784, 500, 'layer1') with tf.name_scope('dropout'): keep_prob = tf.placeholder(tf.float32) - tf.scalar_summary('dropout_keep_probability', keep_prob) + tf.summary.scalar('dropout_keep_probability', keep_prob) dropped = tf.nn.dropout(hidden1, keep_prob) -y = nn_layer(dropped, 500, 10, 'layer2', act=tf.nn.softmax) +# Do not apply softmax activation yet, see below. +y = nn_layer(dropped, 500, 10, 'layer2', act=tf.identity) with tf.name_scope('cross_entropy'): - diff = y_ * tf.log(y) + # The raw formulation of cross-entropy, + # + # tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(tf.softmax(y)), + # reduction_indices=[1])) + # + # can be numerically unstable. + # + # So here we use tf.nn.softmax_cross_entropy_with_logits on the + # raw outputs of the nn_layer above, and then average across + # the batch. + diff = tf.nn.softmax_cross_entropy_with_logits(y, y_) with tf.name_scope('total'): - cross_entropy = -tf.reduce_mean(diff) - tf.scalar_summary('cross entropy', cross_entropy) + cross_entropy = tf.reduce_mean(diff) +tf.summary.scalar('cross_entropy', cross_entropy) with tf.name_scope('train'): train_step = tf.train.AdamOptimizer(FLAGS.learning_rate).minimize( @@ -138,10 +149,10 @@ with tf.name_scope('accuracy'): correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1)) with tf.name_scope('accuracy'): accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) - tf.scalar_summary('accuracy', accuracy) +tf.summary.scalar('accuracy', accuracy) # Merge all the summaries and write them out to /tmp/mnist_logs (by default) -merged = tf.merge_all_summaries() +merged = tf.summary.merge_all() train_writer = tf.train.SummaryWriter(FLAGS.summaries_dir + '/train', sess.graph) test_writer = tf.train.SummaryWriter(FLAGS.summaries_dir + '/test') diff --git a/tensorflow/g3doc/how_tos/supervisor/index.md b/tensorflow/g3doc/how_tos/supervisor/index.md index 98941634ca..d82f7cbe1e 100644 --- a/tensorflow/g3doc/how_tos/supervisor/index.md +++ b/tensorflow/g3doc/how_tos/supervisor/index.md @@ -130,7 +130,7 @@ For example this code runs the summary op every 100 steps in the training loop: ```python ...create graph... my_train_op = ... - my_summary_op = tf.merge_all_summaries() + my_summary_op = tf.summary.merge_all() sv = tf.Supervisor(logdir="/my/training/directory", summary_op=None) # Do not run the summary service @@ -317,7 +317,7 @@ constructor: `tf.GraphKeys.SUMMARY_OP` [graph collection](../../api_docs/python/framework#Graph.add_to_collection). If the collection is empty the supervisor creates an op that aggregates all - summaries in the graph using `tf.merge_all_summaries()`. + summaries in the graph using `tf.summary.merge_all()`. Passing `None` disables the summary service. @@ -404,5 +404,3 @@ Checkpoint recovery is controlled by the following keyword arguments to the ready op the first time, to initialize local variables and tables. * `saver`: (see above). Saver object used to load the checkpoint. - - |