cellmap_flow.finetune.trainer.tensorboard
=========================================

.. py:module:: cellmap_flow.finetune.trainer.tensorboard

.. autoapi-nested-parse::

   A run's TensorBoard event files: the config card, a patch image strip, and
   the per-step and per-epoch scalars.

   File-based, so it works from GPU nodes with no network and needs no
   service; one ``tensorboard --logdir`` over the training/ tree overlays every
   run ever made. Silently off if unavailable or not asked for.



Attributes
----------

.. autoapisummary::

   cellmap_flow.finetune.trainer.tensorboard.logger


Classes
-------

.. autoapisummary::

   cellmap_flow.finetune.trainer.tensorboard.TensorBoard


Module Contents
---------------

.. py:data:: logger

.. py:class:: TensorBoard(log_dir, enabled=True, start_step=0, start_epoch=0)

   The writer, and the step and epoch it is at.

   Monotonic across restarts: the CLI makes a new trainer for every
   iteration, writing to the same tensorboard/ directory, and passes the
   previous one's position on (``start_step``, ``start_epoch``). Starting
   each at 0 drew every iteration's curves on top of each other.


   .. py:attribute:: image_every
      :value: 5



   .. py:attribute:: dir


   .. py:attribute:: step
      :value: 0



   .. py:attribute:: epoch
      :value: 0



   .. py:attribute:: writer
      :value: None



   .. py:property:: on
      :type: bool



   .. py:method:: config_card(rows)

      The run's settings as a table under the "config" text tag.



   .. py:method:: images(raw, target, pred, mask, channel, output_class)

      Mid-Z slice of one sample as one strip: raw (centre-cropped to the
      output) | target | prediction | mask, at the current epoch.

      This is the picture that would have shown augmentation doing nothing
      for five months, and that shows raw and labels moving together once
      it does something. Never lets a display problem stop training.



   .. py:method:: log_step(*, loss, supervised, distillation, other_channels, bce_metrics, interval_terms, lr, step_s, data_wait_s)

      One optimizer step's scalars, at the next step; ``distillation`` and
      ``other_channels`` (the hold on the channels not trained) None when off.



   .. py:method:: log_epoch(*, loss, supervised, best_supervised, bce_metrics, other_channels, data_wait_s, compute_s, peak_memory_gb)

      One epoch's scalars, at the next epoch; ``supervised`` None when nothing was
      supervised, ``other_channels`` (how far the channels not trained moved) None
      without the hold.



   .. py:method:: flush()


   .. py:method:: close()

      Close the writer; the CLI calls this when an iteration is done.