Question

我正在尝试从TensorFlow会话中可视化生成的摘要数据。我已经通过TensorBoard检查功能确认了摘要确实存储了：

tensorboard --logdir=C:\ML\tensorflow_logs --port 6006 --inspect

======================================================================
Processing event files... (this can take a few minutes)
======================================================================

Found event files in:
C:\ML\tensorflow_logs

These tags are in C:\ML\tensorflow_logs:
audio -
histograms -
images -
scalars
   LossValue
   accuracy_1
tensor -
======================================================================

Event statistics for C:\ML\tensorflow_logs:
audio -
graph
   first_step           0
   last_step            0
   max_step             0
   min_step             0
   num_steps            1
   outoforder_steps     []
histograms -
images -
scalars
   first_step           0
   last_step            100
   max_step             100
   min_step             0
   num_steps            101
   outoforder_steps     []
sessionlog:checkpoint -
sessionlog:start -
sessionlog:stop -
tensor -
======================================================================

但是，如果我启动TensorBoard（没有--inspect参数）并在浏览器中打开网站（在这种情况下是Chrome），我只能看到Graph而不是Scalars。对于Scalars，它只是说：

未找到标量数据。

我在Windows上使用Anaconda，使用最新版本的TensorFlow和TensorBoard（0.1.8）。

我使用的与摘要生成相关的代码如下所示：

with graph.as_default():
  .....
  .....
  tf.summary.scalar("LossValue", loss)
  tf.summary.scalar("Accuracy", accuracy_measure)

with tf.Session(graph=graph) as session:
  merged = tf.summary.merge_all()
  writer = tf.summary.FileWriter('C:/ML/tensorflow_logs', session.graph)    
  tf.global_variables_initializer().run()    

  for step in range(train_steps):
    offset = (step * batch_size) % (train_labels.shape[0] - batch_size)
    batch_data = train_dataset[offset:(offset + batch_size), :, :, :]
    batch_labels = train_labels[offset:(offset + batch_size), :]
    feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels, keep_prob : dropout_keep_prob}
    _, l, predictions, summary = session.run([optimizer, loss, train_prediction, merged], feed_dict=feed_dict)    
    writer.add_summary(summary, step)

writer.close()

Answer 1

尝试运行pip install tensorflow-tensorboard

TensorBoard未显示标量摘要，但显示图表

未找到标量数据。

1 个答案: