def train(): with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) images, labels = cifar10.distorted_inputs() logits = cifar10_resnet(images) loss = cifar10.loss(logits, labels) train_op = cifar10.train(loss, global_step) summary_op = tf.merge_all_summaries() init = tf.initialize_all_variables() sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) tf.train.start_queue_runners(sess=sess) for step in xrange(FLAGS.max_steps): _, loss_value = sess.run([train_op, loss]) if step % 10 == 0: print 'step %d, loss = %.3f' % (step, loss_value)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference6(images) # Calculate loss. loss = cifar10.loss(logits, labels) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) # Create a saver. saver = tf.train.Saver(tf.all_variables()) # Build the summary operation based on the TF collection of Summaries. summary_op = tf.merge_all_summaries() # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) # Start the queue runners. tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, sess.graph) for step in xrange(FLAGS.max_steps): start_time = time.time() _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print (format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) if step % 100 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.contrib.framework.get_or_create_global_step() # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() print('Finished getting images & labels') # Build a Graph that computes the logits predictions from the # inference model. logits = modified_inference(images) print('Finished building inference graph') # Calculate loss. loss = cifar10.loss(logits, labels) print('Finished building loss graph') # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) print('Finished building train graph') class _LoggerHook(tf.train.SessionRunHook): """Logs loss and runtime.""" def begin(self): self._step = -1 def before_run(self, run_context): self._step += 1 self._start_time = time.time() return tf.train.SessionRunArgs(loss) # Asks for loss value. def after_run(self, run_context, run_values): duration = time.time() - self._start_time loss_value = run_values.results if self._step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ( '%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print(format_str % (datetime.now(), self._step, loss_value, examples_per_sec, sec_per_batch)) with tf.train.MonitoredTrainingSession( checkpoint_dir=FLAGS.train_dir, hooks=[ tf.train.StopAtStepHook(last_step=FLAGS.max_steps), tf.train.NanTensorHook(loss), _LoggerHook() ], config=tf.ConfigProto(log_device_placement=FLAGS. log_device_placement)) as mon_sess: print('Hooks attached, starting training') while not mon_sess.should_stop(): mon_sess.run(train_op)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.contrib.framework.get_or_create_global_step() # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() #images, labels = cifar10.inputs() # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference(images) # Calculate loss. loss = cifar10.loss(logits, labels) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) class _LoggerHook(tf.train.SessionRunHook): """Logs loss and runtime.""" def begin(self): self._step = -1 def before_run(self, run_context): self._step += 1 self._start_time = time.time() return tf.train.SessionRunArgs(loss) # Asks for loss value. def after_run(self, run_context, run_values): duration = time.time() - self._start_time loss_value = run_values.results if self._step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print (format_str % (datetime.now(), self._step, loss_value, examples_per_sec, sec_per_batch)) with tf.train.MonitoredTrainingSession( checkpoint_dir=FLAGS.train_dir, hooks=[tf.train.StopAtStepHook(last_step=FLAGS.max_steps), tf.train.NanTensorHook(loss), _LoggerHook()], config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) as mon_sess: while not mon_sess.should_stop(): mon_sess.run(train_op)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) images, labels = cifar10.distorted_inputs() logits = cifar10.inference(images) ## EDIT: Softmax activation softmax = tf.nn.softmax(logits) loss = cifar10.loss(softmax, labels) train_op = cifar10.train(loss, global_step) saver = tf.train.Saver(tf.all_variables()) summary_op = tf.merge_all_summaries() init = tf.initialize_all_variables() sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, sess.graph) for step in xrange(FLAGS.max_steps): start_time = time.time() _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print (format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) if step % 100 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) # Save the model checkpoint periodically. if step % 1000 == 0 or (step + 1) == FLAGS.max_steps: checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() print(labels) # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference(images) # Calculate loss. loss = cifar10.loss(logits, labels) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) # Create a saver. saver = tf.train.Saver(tf.all_variables()) # Build the summary operation based on the TF collection of Summaries. summary_op = tf.merge_all_summaries() # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) # Start the queue runners. tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, graph_def=sess.graph_def) for step in xrange(FLAGS.max_steps): start_time = time.time() _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print (format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) if step % 100 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) # Save the model checkpoint periodically. if step % 10 == 0 or (step + 1) == FLAGS.max_steps: checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') print("module save") saver.save(sess, checkpoint_path, global_step=step)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) #GETTING THE TRAINING IMAGES images, labels = cifar10.distorted_inputs() # DATA FOR GRAPH. logits = cifar10.inference(images) # LOSS FUNCTION loss = cifar10.loss(logits, labels) # CREATING AND RUNNING A TENSORBOARD GRAPH train_op = cifar10.train(loss, global_step) saver = tf.train.Saver(tf.all_variables()) summary_op = tf.merge_all_summaries() init = tf.initialize_all_variables() sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) # START THE QUEUE tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, sess.graph) for step in xrange(FLAGS.max_steps): start_time = time.time() _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print (format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) if step % 100 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) # SAVE CHECKPOINT TO EVALUATE PERIODICALLY if step % 1000 == 0 or (step + 1) == FLAGS.max_steps: checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step)
def train(): # ops global_step = tf.Variable(0, trainable=False) images, labels = cifar10.distorted_inputs() logits = cifar10.inference(tf.image.resize_images(images, cifar10.IMAGE_SIZE, cifar10.IMAGE_SIZE)) loss = cifar10.loss(logits, labels) train_op = cifar10.train(loss, global_step) summary_op = tf.merge_all_summaries() with tf.Session() as sess: saver = tf.train.Saver(tf.all_variables(), max_to_keep=21) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir) # restore or initialize variables ckpt = tf.train.get_checkpoint_state(FLAGS.train_dir) if ckpt and ckpt.model_checkpoint_path: saver.restore(sess, ckpt.model_checkpoint_path) else: sess.run(tf.initialize_all_variables()) # Start the queue runners. tf.train.start_queue_runners(sess=sess) start = sess.run(global_step) for step in xrange(start, FLAGS.max_steps): start_time = time.time() _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' print '%d: %f (%.3f sec/batch)' % (step, loss_value, duration) if step % 100 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) if step % 500 == 0 or (step + 1) == FLAGS.max_steps: checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step)
def train(): ps_hosts = FLAGS.ps_hosts.split(',') worker_hosts = FLAGS.worker_hosts.split(',') print('PS hosts are: %s' % ps_hosts) print('Worker hosts are: %s' % worker_hosts) server = tf.train.Server({ 'ps': ps_hosts, 'worker': worker_hosts }, job_name=FLAGS.job_name, task_index=FLAGS.task_id) if FLAGS.job_name == 'ps': server.join() is_chief = (FLAGS.task_id == 0) if is_chief: if tf.gfile.Exists(FLAGS.train_dir): tf.gfile.DeleteRecursively(FLAGS.train_dir) tf.gfile.MakeDirs(FLAGS.train_dir) device_setter = tf.train.replica_device_setter(ps_tasks=1) with tf.device('/job:worker/task:%d' % FLAGS.task_id): with tf.device(device_setter): global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference(images) # Calculate loss. loss = cifar10.loss(logits, labels) train_op = cifar10.train(loss, global_step) saver = tf.train.Saver() # We run the summaries in the same thread as the training operations by # passing in None for summary_op to avoid a summary_thread being started. # Running summaries and training operations in parallel could run out of # GPU memory. sv = tf.train.Supervisor(is_chief=is_chief, logdir=FLAGS.train_dir, init_op=tf.initialize_all_variables(), summary_op=tf.merge_all_summaries(), global_step=global_step, saver=saver, save_model_secs=60) tf.logging.info('%s Supervisor' % datetime.now()) sess_config = tf.ConfigProto( allow_soft_placement=True, log_device_placement=FLAGS.log_device_placement) print("Before session init") # Get a session. sess = sv.prepare_or_wait_for_session(server.target, config=sess_config) print("Session init done") # Start the queue runners. queue_runners = tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS) sv.start_queue_runners(sess, queue_runners) print('Started %d queues for processing input data.' % len(queue_runners)) """Train CIFAR-10 for a number of steps.""" for step in xrange(FLAGS.max_steps): start_time = time.time() _, loss_value, gs = sess.run([train_op, loss, global_step]) duration = time.time() - start_time assert not np.isnan( loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ( '%s: step %d (global_step %d), loss = %.2f (%.1f examples/sec; %.3f sec/batch)' ) print(format_str % (datetime.now(), step, gs, loss_value, examples_per_sec, sec_per_batch)) if is_chief: saver.save(sess, os.path.join(FLAGS.train_dir, 'model.ckpt'), global_step=global_step)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference(images) # Calculate loss. loss = cifar10.loss(logits, labels) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) # Create a saver. saver = tf.train.Saver(tf.all_variables()) # Build the summary operation based on the TF collection of Summaries. summary_op = tf.merge_all_summaries() # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) # Start the queue runners. tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir) summary_writer0 = tf.train.SummaryWriter(FLAGS.train_dir0) summary_writer1 = tf.train.SummaryWriter(FLAGS.train_dir1) summary_writer2 = tf.train.SummaryWriter(FLAGS.train_dir2) summary_writer3 = tf.train.SummaryWriter(FLAGS.train_dir3) summary_writer4 = tf.train.SummaryWriter(FLAGS.train_dir4) summary_writer5 = tf.train.SummaryWriter(FLAGS.train_dir5) summary_writer6 = tf.train.SummaryWriter(FLAGS.train_dir6) summary_writer7 = tf.train.SummaryWriter(FLAGS.train_dir7) summary_writer8 = tf.train.SummaryWriter(FLAGS.train_dir8) summary_writer9 = tf.train.SummaryWriter(FLAGS.train_dir9) summary_writer10 = tf.train.SummaryWriter(FLAGS.train_dir10) summary_writer11 = tf.train.SummaryWriter(FLAGS.train_dir11) summary_writer12 = tf.train.SummaryWriter(FLAGS.train_dir12) summary_writer13 = tf.train.SummaryWriter(FLAGS.train_dir13) summary_writer14 = tf.train.SummaryWriter(FLAGS.train_dir14) summary_writer15 = tf.train.SummaryWriter(FLAGS.train_dir15) summary_writer16 = tf.train.SummaryWriter(FLAGS.train_dir16) summary_writer17 = tf.train.SummaryWriter(FLAGS.train_dir17) summary_writer18 = tf.train.SummaryWriter(FLAGS.train_dir18) summary_writer19 = tf.train.SummaryWriter(FLAGS.train_dir19) for step in xrange(FLAGS.max_steps): start_time = time.time() _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ( '%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print(format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) if step % 100 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) summary_writer0.add_summary(summary_str, step) summary_writer1.add_summary(summary_str, step) summary_writer2.add_summary(summary_str, step) summary_writer3.add_summary(summary_str, step) summary_writer4.add_summary(summary_str, step) summary_writer5.add_summary(summary_str, step) summary_writer6.add_summary(summary_str, step) summary_writer7.add_summary(summary_str, step) summary_writer8.add_summary(summary_str, step) summary_writer9.add_summary(summary_str, step) summary_writer10.add_summary(summary_str, step) summary_writer11.add_summary(summary_str, step) summary_writer12.add_summary(summary_str, step) summary_writer13.add_summary(summary_str, step) summary_writer14.add_summary(summary_str, step) summary_writer15.add_summary(summary_str, step) summary_writer16.add_summary(summary_str, step) summary_writer17.add_summary(summary_str, step) summary_writer18.add_summary(summary_str, step) summary_writer19.add_summary(summary_str, step) # Save the model checkpoint periodically. # if step % 1000 == 0 or (step + 1) == FLAGS.max_steps: # checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') # saver.save(sess, checkpoint_path, global_step=step/100) # hard cord here!!! if step == 100: checkpoint_path = os.path.join(FLAGS.train_dir0, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 200: checkpoint_path = os.path.join(FLAGS.train_dir1, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 300: checkpoint_path = os.path.join(FLAGS.train_dir2, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 400: checkpoint_path = os.path.join(FLAGS.train_dir3, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 500: checkpoint_path = os.path.join(FLAGS.train_dir4, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 600: checkpoint_path = os.path.join(FLAGS.train_dir5, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 700: checkpoint_path = os.path.join(FLAGS.train_dir6, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 800: checkpoint_path = os.path.join(FLAGS.train_dir7, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 900: checkpoint_path = os.path.join(FLAGS.train_dir8, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 1000: checkpoint_path = os.path.join(FLAGS.train_dir9, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 1100: checkpoint_path = os.path.join(FLAGS.train_dir10, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 1200: checkpoint_path = os.path.join(FLAGS.train_dir11, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 1300: checkpoint_path = os.path.join(FLAGS.train_dir12, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 1400: checkpoint_path = os.path.join(FLAGS.train_dir13, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 1500: checkpoint_path = os.path.join(FLAGS.train_dir14, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 1600: checkpoint_path = os.path.join(FLAGS.train_dir15, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 1700: checkpoint_path = os.path.join(FLAGS.train_dir16, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 1800: checkpoint_path = os.path.join(FLAGS.train_dir17, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 1900: checkpoint_path = os.path.join(FLAGS.train_dir18, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step == 2000: checkpoint_path = os.path.join(FLAGS.train_dir19, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference(images) # Calculate loss. loss = cifar10.loss(logits, labels) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) # Create a saver. saver = tf.train.Saver(tf.all_variables()) # Build the summary operation based on the TF collection of Summaries. summary_op = tf.merge_all_summaries() # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) # Start the queue runners. tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir) summary_writer0 = tf.train.SummaryWriter(FLAGS.train_dir0) summary_writer1= tf.train.SummaryWriter(FLAGS.train_dir1) summary_writer2 = tf.train.SummaryWriter(FLAGS.train_dir2) summary_writer3 = tf.train.SummaryWriter(FLAGS.train_dir3) summary_writer4 = tf.train.SummaryWriter(FLAGS.train_dir4) summary_writer5 = tf.train.SummaryWriter(FLAGS.train_dir5) summary_writer6 = tf.train.SummaryWriter(FLAGS.train_dir6) summary_writer7 = tf.train.SummaryWriter(FLAGS.train_dir7) summary_writer8 = tf.train.SummaryWriter(FLAGS.train_dir8) summary_writer9 = tf.train.SummaryWriter(FLAGS.train_dir9) summary_writer10 = tf.train.SummaryWriter(FLAGS.train_dir10) summary_writer11 = tf.train.SummaryWriter(FLAGS.train_dir11) summary_writer12 = tf.train.SummaryWriter(FLAGS.train_dir12) summary_writer13 = tf.train.SummaryWriter(FLAGS.train_dir13) summary_writer14 = tf.train.SummaryWriter(FLAGS.train_dir14) summary_writer15 = tf.train.SummaryWriter(FLAGS.train_dir15) summary_writer16 = tf.train.SummaryWriter(FLAGS.train_dir16) summary_writer17 = tf.train.SummaryWriter(FLAGS.train_dir17) summary_writer18 = tf.train.SummaryWriter(FLAGS.train_dir18) summary_writer19 = tf.train.SummaryWriter(FLAGS.train_dir19) for step in xrange(FLAGS.max_steps): start_time = time.time() _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print (format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) if step % 100 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) summary_writer0.add_summary(summary_str, step) summary_writer1.add_summary(summary_str, step) summary_writer2.add_summary(summary_str, step) summary_writer3.add_summary(summary_str, step) summary_writer4.add_summary(summary_str, step) summary_writer5.add_summary(summary_str, step) summary_writer6.add_summary(summary_str, step) summary_writer7.add_summary(summary_str, step) summary_writer8.add_summary(summary_str, step) summary_writer9.add_summary(summary_str, step) summary_writer10.add_summary(summary_str, step) summary_writer11.add_summary(summary_str, step) summary_writer12.add_summary(summary_str, step) summary_writer13.add_summary(summary_str, step) summary_writer14.add_summary(summary_str, step) summary_writer15.add_summary(summary_str, step) summary_writer16.add_summary(summary_str, step) summary_writer17.add_summary(summary_str, step) summary_writer18.add_summary(summary_str, step) summary_writer19.add_summary(summary_str, step) # Save the model checkpoint periodically. # if step % 1000 == 0 or (step + 1) == FLAGS.max_steps: # checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') # saver.save(sess, checkpoint_path, global_step=step/100) # hard cord here!!! if step==100: checkpoint_path = os.path.join(FLAGS.train_dir0, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==200: checkpoint_path = os.path.join(FLAGS.train_dir1, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==300: checkpoint_path = os.path.join(FLAGS.train_dir2, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==400: checkpoint_path = os.path.join(FLAGS.train_dir3, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==500: checkpoint_path = os.path.join(FLAGS.train_dir4, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==600: checkpoint_path = os.path.join(FLAGS.train_dir5, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==700: checkpoint_path = os.path.join(FLAGS.train_dir6, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==800: checkpoint_path = os.path.join(FLAGS.train_dir7, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==900: checkpoint_path = os.path.join(FLAGS.train_dir8, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==1000: checkpoint_path = os.path.join(FLAGS.train_dir9, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==1100: checkpoint_path = os.path.join(FLAGS.train_dir10, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==1200: checkpoint_path = os.path.join(FLAGS.train_dir11, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==1300: checkpoint_path = os.path.join(FLAGS.train_dir12, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==1400: checkpoint_path = os.path.join(FLAGS.train_dir13, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==1500: checkpoint_path = os.path.join(FLAGS.train_dir14, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==1600: checkpoint_path = os.path.join(FLAGS.train_dir15, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==1700: checkpoint_path = os.path.join(FLAGS.train_dir16, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==1800: checkpoint_path = os.path.join(FLAGS.train_dir17, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==1900: checkpoint_path = os.path.join(FLAGS.train_dir18, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) if step==2000: checkpoint_path = os.path.join(FLAGS.train_dir19, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step)
def train(): ps_hosts = FLAGS.ps_hosts.split(',') worker_hosts = FLAGS.worker_hosts.split(',') print ('PS hosts are: %s' % ps_hosts) print ('Worker hosts are: %s' % worker_hosts) server = tf.train.Server( {'ps': ps_hosts, 'worker': worker_hosts}, job_name = FLAGS.job_name, task_index=FLAGS.task_id) if FLAGS.job_name == 'ps': server.join() is_chief = (FLAGS.task_id == 0) if is_chief: if tf.gfile.Exists(FLAGS.train_dir): tf.gfile.DeleteRecursively(FLAGS.train_dir) tf.gfile.MakeDirs(FLAGS.train_dir) device_setter = tf.train.replica_device_setter(ps_tasks=1) with tf.device('/job:worker/task:%d' % FLAGS.task_id): with tf.device(device_setter): global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference(images) # Calculate loss. loss = cifar10.loss(logits, labels) train_op = cifar10.train(loss, global_step) saver = tf.train.Saver() # We run the summaries in the same thread as the training operations by # passing in None for summary_op to avoid a summary_thread being started. # Running summaries and training operations in parallel could run out of # GPU memory. sv = tf.train.Supervisor(is_chief=is_chief, logdir=FLAGS.train_dir, init_op=tf.initialize_all_variables(), summary_op=tf.merge_all_summaries(), global_step=global_step, saver=saver, save_model_secs=60) tf.logging.info('%s Supervisor' % datetime.now()) sess_config = tf.ConfigProto(allow_soft_placement=True, log_device_placement=FLAGS.log_device_placement) print ("Before session init") # Get a session. sess = sv.prepare_or_wait_for_session(server.target, config=sess_config) print ("Session init done") # Start the queue runners. queue_runners = tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS) sv.start_queue_runners(sess, queue_runners) print ('Started %d queues for processing input data.' % len(queue_runners)) """Train CIFAR-10 for a number of steps.""" for step in xrange(FLAGS.max_steps): start_time = time.time() _, loss_value, gs = sess.run([train_op, loss, global_step]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d (global_step %d), loss = %.2f (%.1f examples/sec; %.3f sec/batch)') print (format_str % (datetime.now(), step, gs, loss_value, examples_per_sec, sec_per_batch)) if is_chief: saver.save(sess, os.path.join(FLAGS.train_dir, 'model.ckpt'), global_step=global_step)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. #with tf.device('/gpu:%d' % FLAGS.gpu_number): images, labels = cifar10.distorted_inputs() # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference(images) loss_per_batch = cifar10.loss_per_batch(logits, labels) # Calculate loss. loss = cifar10.loss(logits, labels) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step, FLAGS.gpu_number) # Create a saver. saver = tf.train.Saver(tf.all_variables(), max_to_keep=None) # Build the summary operation based on the TF collection of Summaries. summary_op = tf.merge_all_summaries() # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. config = tf.ConfigProto() config.gpu_options.allow_growth=True config.allow_soft_placement=True config.log_device_placement=FLAGS.log_device_placement sess = tf.Session(config=config) tf.train.write_graph(sess.graph_def, FLAGS.train_dir, "cifar10_train.pb", False) sess.run(init) # Start the queue runners. tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, sess.graph) train_start_time = time.time() for step in xrange(FLAGS.max_steps): start_time = time.time() _, loss_value, logits_value, loss_per_batch_value, labels_value = sess.run([train_op, loss, logits, loss_per_batch, labels]) duration = time.time() - start_time #logits_str = print_logits(logits_value, labels_value, loss_per_batch_value) #with open(os.path.join(FLAGS.train_dir, 'logits_%d.log' % step),'w') as f: # f.write("%s" % logits_str) assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') log_str = (format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) print(log_str) with open(os.path.join(FLAGS.train_dir, 'train.log'),'a+') as f: f.write("%s\n" % log_str) if step % 500 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') save_path = saver.save(sess, checkpoint_path, global_step=step) train_duration = time.time() - train_start_time log_str = ("Finishing. Training %d batches of %d images took %fs\n" % (FLAGS.max_steps, FLAGS.batch_size, float(train_duration))) print(log_str) with open(os.path.join(FLAGS.train_dir, 'train.log'),'a+') as f: f.write("%s" % log_str)
def run_training(): """Train MNIST for a number of steps.""" # Tell TensorFlow that the model will be built into the default Graph. with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) # Input images and labels. images, labels = inputs(train=True, batch_size=BATCH_SIZE, num_epochs=FLAGS.num_epochs) print('images', images) logits = calc_inference(images) print('logits', logits) print('labels', labels) # Calculate loss. loss = calc_loss(logits, labels) print(loss) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) # Create a saver. saver = tf.train.Saver(tf.all_variables()) # Build the summary operation based on the TF collection of Summaries. summary_op = tf.merge_all_summaries() # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) # Start the queue runners. tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, sess.graph) for step in xrange(FLAGS.max_steps): start_time = time.time() print(step) print(train_op) _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = BATCH_SIZE examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ( '%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print(format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) if step % 100 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) # Save the model checkpoint periodically. if step % 1000 == 0 or (step + 1) == FLAGS.max_steps: checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() # #Adding dropout # keep_drop_prob = tf.placeholder(tf.float32) # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference(images) # ###########Changes for visualization ############### with tf.variable_scope('conv1') as scope_conv: tf.get_variable_scope().reuse_variables() weights = tf.get_variable('weights') grid_x = grid_y = 8 # to get a square grid for 64 conv1 features grid = put_kernels_on_grid(weights, (grid_y, grid_x)) tf.image_summary('conv1/features', grid, max_images=1) # #################################################### # Calculate loss. loss = cifar10.loss(logits, labels) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) # Create a saver. saver = tf.train.Saver(tf.all_variables()) # Build the summary operation based on the TF collection of Summaries. summary_op = tf.merge_all_summaries() # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) # Start the queue runners. tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, sess.graph) for step in range(FLAGS.max_steps): start_time = time.time() _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print(format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) if step % 100 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) # Save the model checkpoint periodically. if step % 1000 == 0 or (step + 1) == FLAGS.max_steps: checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): with tf.variable_scope("model") as scope: global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() images_eval, labels_eval = cifar10.inputs(eval_data=True) # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference(images) scope.reuse_variables() logits_eval = cifar10.inference(images_eval) # Calculate loss. loss = cifar10.loss(logits, labels) # For evaluation top_k = tf.nn.in_top_k(logits, labels, 1) top_k_eval = tf.nn.in_top_k(logits_eval, labels_eval, 1) # Add precision summary summary_train_prec = tf.placeholder(tf.float32) summary_eval_prec = tf.placeholder(tf.float32) tf.scalar_summary('precision/train', summary_train_prec) tf.scalar_summary('precision/eval', summary_eval_prec) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) # Create a saver. saver = tf.train.Saver(tf.all_variables()) # Build the summary operation based on the TF collection of Summaries. summary_op = tf.merge_all_summaries() # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) # Start the queue runners. tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, graph_def=sess.graph_def) for step in xrange(FLAGS.max_steps): start_time = time.time() _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan( loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ( '%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print(format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) EVAL_STEP = 10 EVAL_NUM_EXAMPLES = 1024 if step % EVAL_STEP == 0: prec_train = evaluate_set(sess, top_k, EVAL_NUM_EXAMPLES) prec_eval = evaluate_set(sess, top_k_eval, EVAL_NUM_EXAMPLES) print('%s: precision train = %.3f' % (datetime.now(), prec_train)) print('%s: precision eval = %.3f' % (datetime.now(), prec_eval)) if step % 100 == 0: summary_str = sess.run(summary_op, feed_dict={ summary_train_prec: prec_train, summary_eval_prec: prec_eval }) summary_writer.add_summary(summary_str, step) # Save the model checkpoint periodically. if step % 1000 == 0 or (step + 1) == FLAGS.max_steps: checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference(images) # Calculate loss. loss = cifar10.loss(logits, labels) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) # Create a saver. saver = tf.train.Saver(tf.all_variables()) # Build the summary operation based on the TF collection of Summaries. summary_op = tf.merge_all_summaries() # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) if FLAGS.checkpoint_dir is not None: ckpt = tf.train.get_checkpoint_state(FLAGS.checkpoint_dir) print("checkpoint path is %s" % ckpt.model_checkpoint_path) tf.train.Saver().restore(sess, ckpt.model_checkpoint_path) # Start the queue runners. print("FLAGS.checkpoint_dir is %s" % FLAGS.checkpoint_dir) tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, sess.graph) cur_step = sess.run(global_step) print("current step is %s" % cur_step) interrupt_check_duration = 0.0 elapsed_time = time.time() flag = 0 for step in xrange(cur_step, FLAGS.max_steps): start_time = time.time() _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time interrupt_check_duration += duration if float(interrupt_check_duration) > 5.0: print("checking for interruption: %s", interrupt_check_duration) if decision_for_migration(): print("have to migrate") checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') print("checkpoint path is %s" % checkpoint_path) saver.save(sess, checkpoint_path, global_step=step) random_id = generate_random_prefix() start_new_instance(checkpoint_path, step, random_id) upload_checkpoint_to_s3(checkpoint_path, step, "mj-bucket-1", random_id) break else: print("not interrupted") interrupt_check_duration = 0.0 assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print(format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) if step % 100 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) # Save the model checkpoint periodically. if step % 1000 == 0 or (step + 1) == FLAGS.max_steps: checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step) elapsed = (int(time.time() - elapsed_time)) if elapsed % 300 == 0 and flag == 0: print("uploading current status") uploading_current_status_to_rds(step) flag = 1 elif elapsed % 300 != 0 and flag == 1: flag = 0
# Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() test_images, test_labels = cifar10.inputs(eval_data='test') # Build a Graph that computes the logits predictions from the # inference model. logits = test_model.predict(images) logit_test = test_model.predict(test_images) # Calculate loss. loss = cifar10.loss(logits, labels) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) top_k_op = tf.nn.in_top_k(logit_test, test_labels, 1) # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. #sess = tf.Session(config=tf.ConfigProto( # log_device_placement=FLAGS.log_device_placement)) with tf.Session(config=tf.ConfigProto( log_device_placement=False)) as sess: sess.run(init)
def train(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() eval_data = FLAGS.eval_data == 'test' #timages, tlabels = cifar10.inputs(eval_data=eval_data) # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference(images) #tlogits = cifar10.inference(timages) # Calculate loss. top_k_op = tf.nn.in_top_k(logits, labels, 1) loss = cifar10.loss(logits, labels) #precision = tf.Variable(0.8, name='precision') #tf.scalar_summary('accuracy', precision) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) # Create a saver. saver = tf.train.Saver(tf.all_variables()) # Build the summary operation based on the TF collection of Summaries. summary_op = tf.merge_all_summaries() # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) # Start the queue runners. tf.train.start_queue_runners(sess=sess) sess.graph.finalize() summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, sess.graph) for step in xrange(FLAGS.max_steps): start_time = time.time() _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 100 == 0: # Build a Graph that computes the logits predictions from the # inference model. # Calculate predictions. num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print (format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) num_iter = int(math.ceil(FLAGS.num_examples / FLAGS.batch_size)) true_count = 0 # Counts the number of correct predictions. total_sample_count = num_iter * FLAGS.batch_size i_step = 0 while i_step < num_iter: predictions = sess.run([top_k_op]) true_count += np.sum(predictions) i_step += 1 #Compute precision @ 1. #sess.run(precision.assign(true_count / total_sample_count)) prec = true_count / total_sample_count print(prec) summary = tf.Summary() summary.ParseFromString(sess.run(summary_op)) summary.value.add(tag='accuracy', simple_value=prec) summary_writer.add_summary(summary, step) #summary_str = sess.run(summary_op) #summary_writer.add_summary(summary_str, step) #summary_writer.flush() # Save the model checkpoint periodically. if step % 100 == 0 or (step + 1) == FLAGS.max_steps: checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step)
def SGDBead(self, bead, thresh, maxindex): finalerror = 0. #thresh = .05 # Parameters learning_rate = 0.001 training_epochs = 15 batch_size = 100 display_step = 1 curWeights, curBiases = self.AllBeads[bead] #test_model = multilayer_perceptron(w=curWeights, b=curBiases) test_model = convnet(w=curWeights, b=curBiases) with test_model.g.as_default(): global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() test_images, test_labels = cifar10.inputs(eval_data='test') # Build a Graph that computes the logits predictions from the # inference model. logits = test_model.predict(images) logit_test = test_model.predict(test_images) # Calculate loss. loss = cifar10.loss(logits, labels) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) top_k_op = tf.nn.in_top_k(logit_test, test_labels, 1) # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. #sess = tf.Session(config=tf.ConfigProto( # log_device_placement=FLAGS.log_device_placement)) with tf.Session(config=tf.ConfigProto( log_device_placement=False)) as sess: sess.run(init) tf.train.start_queue_runners(sess=sess) step = 0 stopcond = True while step < max_steps and stopcond: start_time = time.time() _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print (format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) if step % 100 == 0: num_iter = int(math.ceil(num_examples / batch_size)) true_count = 0 # Counts the number of correct predictions. total_sample_count = num_iter * batch_size stepp = 0 while stepp < num_iter: predictions = sess.run([top_k_op]) true_count += np.sum(predictions) stepp += 1 # Compute precision @ 1. precision = true_count / total_sample_count print('%s: precision @ 1 = %.3f' % (datetime.now(), precision)) if precision > 1 - thresh: stopcond = False test_model.params = sess.run(test_model.weightslist), sess.run(test_model.biaseslist) self.AllBeads[bead]=test_model.params finalerror = 1 - precision print ("Final bead error: ",str(finalerror)) step += 1 return finalerror
def multilevel_train_1ord(): """Train CIFAR-10 for a number of steps.""" with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) # Get images and labels for CIFAR-10. images, labels = cifar10.distorted_inputs() # Build a Graph that computes the logits predictions from the # inference model. logits = cifar10.inference(images) # Calculate loss. loss = cifar10.loss(logits, labels) #Accurarcy top_k_op = tf.nn.in_top_k(logits, labels, 1) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) # Create a saver. saver = tf.train.Saver(tf.all_variables()) # Build the summary operation based on the TF collection of Summaries. summary_op = tf.merge_all_summaries() # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) # Start the queue runners. tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, sess.graph) for step in xrange(FLAGS.max_steps): start_time = time.time() _, loss_value = sess.run([train_op, loss]) accurarcy = sess.run(top_k_op) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' output_list = [] # Do something with intermediate data (intermediate) # Save data on iterations of 0, 1000, 2000, 3000 if step % 1000 == 0 or (step + 1) == FLAGS.max_steps: for v in tf.all_variables(): if "conv1/weights:" in v.name: print(v.name) output_list.append( tf.get_default_graph().get_tensor_by_name(v.name)) break if (step == 0): conv1_data_0 = sess.run(output_list) if (step == 1000): conv1_data_1000 = sess.run(output_list) if (step == 2000): conv1_data_2000 = sess.run(output_list) if (step == 3000): conv1_data_3000 = sess.run(output_list) (A, B, C, D, E) = np.array(conv1_data_3000).shape # do something. # do experiments if step == 3000 or (step + 1) == FLAGS.max_steps: print("************\n Chen process executing") _, new_data = process.exp_2_commMax(conv1_data_0, conv1_data_1000, conv1_data_2000, conv1_data_3000) for v in tf.all_variables(): if "conv1/weights:" in v.name: print("start assign: ") sess.run( tf.assign( tf.get_default_graph().get_tensor_by_name( v.name), new_data[0])) break value = sess.run(loss) pred = process.Count(accurarcy) print("new loss value is: " + str(value) + " accurarcy :" + str(pred)) if step % 10 == 0: num_examples_per_step = FLAGS.batch_size examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) predict = process.Count(accurarcy) format_str = ( '%s: step %d, loss = %.2f, accu = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print(format_str % (datetime.now(), step, loss_value, predict, examples_per_sec, sec_per_batch)) if step % 100 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) # Save the model checkpoint periodically. if step % 1000 == 0 or (step + 1) == FLAGS.max_steps: checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step)
def run_training(): """Train MNIST for a number of steps.""" # Tell TensorFlow that the model will be built into the default Graph. with tf.Graph().as_default(): global_step = tf.Variable(0, trainable=False) # Input images and labels. images, labels = inputs(train=True, batch_size=BATCH_SIZE, num_epochs=FLAGS.num_epochs) print('images', images) logits = calc_inference(images) print('logits', logits) print('labels', labels) # Calculate loss. loss = calc_loss(logits, labels) print(loss) # Build a Graph that trains the model with one batch of examples and # updates the model parameters. train_op = cifar10.train(loss, global_step) # Create a saver. saver = tf.train.Saver(tf.all_variables()) # Build the summary operation based on the TF collection of Summaries. summary_op = tf.merge_all_summaries() # Build an initialization operation to run below. init = tf.initialize_all_variables() # Start running operations on the Graph. sess = tf.Session(config=tf.ConfigProto( log_device_placement=FLAGS.log_device_placement)) sess.run(init) # Start the queue runners. tf.train.start_queue_runners(sess=sess) summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, sess.graph) for step in xrange(FLAGS.max_steps): start_time = time.time() print(step) print(train_op) _, loss_value = sess.run([train_op, loss]) duration = time.time() - start_time assert not np.isnan(loss_value), 'Model diverged with loss = NaN' if step % 10 == 0: num_examples_per_step = BATCH_SIZE examples_per_sec = num_examples_per_step / duration sec_per_batch = float(duration) format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f ' 'sec/batch)') print (format_str % (datetime.now(), step, loss_value, examples_per_sec, sec_per_batch)) if step % 100 == 0: summary_str = sess.run(summary_op) summary_writer.add_summary(summary_str, step) # Save the model checkpoint periodically. if step % 1000 == 0 or (step + 1) == FLAGS.max_steps: checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt') saver.save(sess, checkpoint_path, global_step=step)