MaidrCallback

keras.MaidrCallback(self, path=None, *, every=1, metrics=None)

Keep an accessible page of a model’s training curves while it trains.

Pass it to model.fit(callbacks=[...]). At the end of every every epochs it writes path again with the charts :func:plot_history draws, so a reader can open the page during a long run and reload it to hear how far the model has come. When training ends, the page is written a last time; with no path, the charts are shown in the notebook instead, if there is one.

Parameters

Name	Type	Description	Default
path	str or os.PathLike	The HTML file to keep up to date. The page is replaced in one step, so a reload never meets a half-written file. A page that cannot be written, or drawn, is warned about, and training goes on. In multi-worker training, give each worker its own path, or the callback to one.	`None`
every	int	Write the page every this many epochs.	`1`
metrics	iterable of str	Only these metrics, in this order, as for :func:`plot_history`.	`None`

Attributes

Name	Type	Description
history	dict of str to list of tuple	Each metric’s values so far, as `(epoch, value)` pairs, epochs counted from 1. Reset when a new `model.fit` begins.

Raises

Name	Type	Description
	ImportError	If Keras is not installed.
	ValueError	If `every` is below 1.

Examples

>>> from maidr.keras import MaidrCallback
>>> model.fit(x, y, validation_split=0.2, epochs=50,
...           callbacks=[MaidrCallback("training.html", every=5)])

Methods

Name	Description
figure	Draw the curves recorded so far.
on_epoch_end	Record the epoch’s metrics, and write the page every `every` epochs.
on_train_begin	Start a new history, as each `model.fit` does.
on_train_end	Write the page a last time, or show the charts in a notebook.

figure

keras.MaidrCallback.figure()

Draw the curves recorded so far.

Returns

Name	Type	Description
	`matplotlib`.`figure`.`Figure`	The charts :func:`plot_history` would draw for them.

on_epoch_end

keras.MaidrCallback.on_epoch_end(epoch, logs=None)

Record the epoch’s metrics, and write the page every every epochs.

Parameters

Name	Type	Description	Default
epoch	int	The epoch that ended, counted from 0 as Keras passes it.	required
logs	`mapping`	Each metric’s value at this epoch. A value that is not a single number is not recorded, with a warning the first time.	`None`

on_train_begin

keras.MaidrCallback.on_train_begin(logs=None)

Start a new history, as each model.fit does.

Parameters

Name	Type	Description	Default
logs	`mapping`	Passed by Keras; not read.	`None`

on_train_end

keras.MaidrCallback.on_train_end(logs=None)

Write the page a last time, or show the charts in a notebook.

Parameters

Name	Type	Description	Default
logs	`mapping`	Passed by Keras; not read.	`None`