How to use several summary collections in Tensorfl

问题:

I have 2 distinctive groups of summaries. One is collected once per batch another one is collected once per epoch. How can I use merge_all_summaries(key='???') to collect summaries in this two groups separately? Doing it manually is always an option but there seems to be a better way.

Illustration of how i think it should work:

      # once per batch 
      tf.scalar_summary("loss", graph.loss)
      tf.scalar_summary("batch_acc", batch_accuracy)
      # once per epoch
      gradients = tf.gradients(graph.loss, [W, D])
      tf.histogram_summary("embedding/W", W, collections='per_epoch')
      tf.histogram_summary("embedding/D", D, collections='per_epoch')

      tf.merge_all_summaries()                # -> (MergeSummary...) :)
      tf.merge_all_summaries(key='per_epoch') # -> NONE              :(

回答1:

Problem solved. collections parameter of a summary is supposed to be a list. Solution:

  # once per batch 
  tf.scalar_summary("loss", graph.loss)
  tf.scalar_summary("batch_acc", batch_accuracy)
  # once per epoch
  tf.histogram_summary("embedding/W", W, collections=['per_epoch'])
  tf.histogram_summary("embedding/D", D, collections=['per_epoch'])

  tf.merge_all_summaries()                # -> (MergeSummary...) :)
  tf.merge_all_summaries(key='per_epoch') # -> (MergeSummary...) :)

Edit. Syntactical change in TF:

# once per batch 
  tf.summary.scalar("loss", graph.loss)
  tf.summary.scalar("batch_acc", batch_accuracy)
  # once per epoch
  tf.summary.histogram("embedding/W", W, collections=['per_epoch'])
  tf.summary.histogram("embedding/D", D, collections=['per_epoch'])

  tf.summaries.merge_all()                # -> (MergeSummary...) :)
  tf.summaries.merge_all(key='per_epoch') # -> (MergeSummary...) :)