from builtins import range from collections import namedtuple from datetime import datetime import csv import math import time import tensorflow.python.platform import tensorflow as tf FLAGS = tf.app.flags.FLAGS tf.app.flags.DEFINE_integer('batch_size', 128, """Batch size.""") tf.app.flags.DEFINE_integer('num_batches', 100, """Number of batches to run.""") tf.app.flags.DEFINE_boolean('forward_only', False, """Only run the forward pass.""") tf.app.flags.DEFINE_boolean('forward_backward_only', False, """Only run the forward-forward pass.""") tf.app.flags.DEFINE_string('data_format', 'NCHW', """The data format for Convnet operations. Can be either NHWC or NCHW. """) tf.app.flags.DEFINE_string('csv_file', '', """File to output timing information to in csv format. If not file is passed in, csv file will not be cteated. """) parameters = [] conv_counter = 1 pool_counter = 1 affine_counter = 1 TimingEntry = namedtuple( 'TimingEntry', ['info_string', 'timestamp', 'num_batches', 'mean', 'sd']) def _conv(inpOp, nIn, nOut, kH, kW, dH, dW, padType): global conv_counter global parameters name = 'conv' + str(conv_counter) conv_counter += 1 with tf.name_scope(name) as scope: kernel = tf.Variable(tf.truncated_normal([kH, kW, nIn, nOut], dtype=tf.float32, stddev=1e-1), name='weights') if FLAGS.data_format == 'NCHW': strides = [1, 1, dH, dW] else: strides = [1, dH, dW, 1] conv = tf.nn.conv2d(inpOp, kernel, strides, padding=padType, data_format=FLAGS.data_format) biases = tf.Variable(tf.constant(0.0, shape=[nOut], dtype=tf.float32), trainable=True, name='biases') bias = tf.reshape(tf.nn.bias_add(conv, biases, data_format=FLAGS.data_format), conv.get_shape()) conv1 = tf.nn.relu(bias, name=scope) parameters += [kernel, biases] return conv1 def _affine(inpOp, nIn, nOut): global affine_counter global parameters name = 'affine' + str(affine_counter) affine_counter += 1 with tf.name_scope(name) as scope: kernel = tf.Variable(tf.truncated_normal([nIn, nOut], dtype=tf.float32, stddev=1e-1), name='weights') biases = tf.Variable(tf.constant(0.0, shape=[nOut], dtype=tf.float32), trainable=True, name='biases') affine1 = tf.nn.relu_layer(inpOp, kernel, biases, name=name) parameters += [kernel, biases] return affine1 def _mpool(inpOp, kH, kW, dH, dW): global pool_counter global parameters name = 'pool' + str(pool_counter) pool_counter += 1 if FLAGS.data_format == 'NCHW': ksize = [1, 1, kH, kW] strides = [1, 1, dH, dW] else: ksize = [1, kH, kW, 1] strides = [1, dH, dW, 1] return tf.nn.max_pool(inpOp, ksize=ksize, strides=strides, padding='VALID', data_format=FLAGS.data_format, name=name) def loss(logits, labels): batch_size = tf.size(labels) labels = tf.expand_dims(labels, 1) indices = tf.expand_dims(tf.range(0, batch_size, 1), 1) concated = tf.concat([indices, labels], 1) onehot_labels = tf.sparse_to_dense( concated, tf.stack([batch_size, 1000]), 1.0, 0.0) cross_entropy = tf.nn.softmax_cross_entropy_with_logits( logits=logits, labels=onehot_labels, name='xentropy') loss = tf.reduce_mean(cross_entropy, name='xentropy_mean') return loss def inference(images): conv1 = _conv (images, 3, 64, 11, 11, 4, 4, 'VALID') pool1 = _mpool(conv1, 3, 3, 2, 2) conv2 = _conv (pool1, 64, 192, 5, 5, 1, 1, 'SAME') pool2 = _mpool(conv2, 3, 3, 2, 2) conv3 = _conv (pool2, 192, 384, 3, 3, 1, 1, 'SAME') conv4 = _conv (conv3, 384, 256, 3, 3, 1, 1, 'SAME') conv5 = _conv (conv4, 256, 256, 3, 3, 1, 1, 'SAME') pool5 = _mpool(conv5, 3, 3, 2, 2) resh1 = tf.reshape(pool5, [-1, 256 * 6 * 6]) affn1 = _affine(resh1, 256 * 6 * 6, 4096) affn2 = _affine(affn1, 4096, 4096) affn3 = _affine(affn2, 4096, 1000) return affn3 def time_tensorflow_run(session, target, info_string): num_steps_burn_in = 10 total_duration = 0.0 total_duration_squared = 0.0 if not isinstance(target, list): target = [target] target_op = tf.group(*target) for i in range(FLAGS.num_batches + num_steps_burn_in): start_time = time.time() _ = session.run(target_op) duration = time.time() - start_time if i > num_steps_burn_in: if not i % 10: print ('%s: step %d, duration = %.3f' % (datetime.now(), i - num_steps_burn_in, duration)) total_duration += duration total_duration_squared += duration * duration mn = total_duration / FLAGS.num_batches vr = total_duration_squared / FLAGS.num_batches - mn * mn sd = math.sqrt(vr) print ('%s: %s across %d steps, %.3f +/- %.3f sec / batch' % (datetime.now(), info_string, FLAGS.num_batches, mn, sd)) return TimingEntry(info_string, datetime.now(), FLAGS.num_batches, mn, sd) def store_data_in_csv(timing_entries): with open(FLAGS.csv_file, 'wb') as csvfile: writer = csv.writer(csvfile) for timing_entry in timing_entries: writer.writerow( [timing_entry.info_string, timing_entry.timestamp, timing_entry.num_batches, timing_entry.mean, timing_entry.sd]) def run_benchmark(): global parameters timing_entries = [] with tf.Graph().as_default(): # Generate some dummy images. image_size = 224 # Note that our padding definition is slightly different the cuda-convnet. # In order to force the model to start with the same activations sizes, # we add 3 to the image_size and employ VALID padding above. if FLAGS.data_format == 'NCHW': image_shape = [FLAGS.batch_size, 3, image_size + 3, image_size + 3] else: image_shape = [FLAGS.batch_size, image_size + 3, image_size + 3, 3] images = tf.Variable(tf.random_normal(image_shape, dtype=tf.float32, stddev=1e-1)) labels = tf.Variable(tf.ones([FLAGS.batch_size], dtype=tf.int32)) # Build a Graph that computes the logits predictions from the # inference model. last_layer = inference(images) # Build an initialization operation. init = tf.global_variables_initializer() # Start running operations on the Graph. sess = tf.Session('') sess.run(init) run_forward = True run_forward_backward = True if FLAGS.forward_only and FLAGS.forward_backward_only: raise ValueError("Cannot specify --forward_only and " "--forward_backward_only at the same time.") if FLAGS.forward_only: run_forward_backward = False elif FLAGS.forward_backward_only: run_forward = False if run_forward: # Run the forward benchmark. timing_entries.append(time_tensorflow_run(sess, last_layer, "Forward")) if run_forward_backward: # Add a simple objective so we can calculate the backward pass. objective = loss(last_layer, labels) # Compute the gradient with respect to all the parameters. grad = tf.gradients(objective, parameters) # Run the backward benchmark. timing_entries.append(time_tensorflow_run(sess, grad, "Forward-backward")) if FLAGS.csv_file: store_data_in_csv(timing_entries) def main(_): run_benchmark() if __name__ == '__main__': tf.app.run()