objects

Callbacks

🎉 New feature introduced in pyhectiqlab>=2.0.11 🎉

For tensorflow user, you can use the pyhectiqlab.callbacks.KerasCallback object to push automatically your metrics. It supports the training, validattion and prediction metrics.

Usage

You simply need to create a run and use it to initialize the callback.

from pyhectiqlab import Run
from pyhectiqlab.callbacks import KerasCallback
run = Run(name="Push metrics", project="lab/demo")

callback = KerasCallback(run=run)

# Use it
model.fit(
    x_train,
    y_train,
    callbacks=[callback],
)

The callback supports other parameters for additional control over what you push.

Method

pyhectiqlab.callbacks.KerasCallback(run: pyhectiqlab.Run, level: str = batch, exclude_train_metrics: bool = False, exclude_val_metrics: bool = False, exclude_predict_metrics: bool = False, exclude_metrics: List[str] = [])

A custom callback object for keras models.

Parameters

Property	Type	Default	Description
`run`	`pyhectiqlab.Run`	`-`	The target run to host the metrics
`level`	`str`	`batch`	Either 'batch' or 'epoch'. If 'batch', the metrics are pushed on every batch and the stepstamp incremented on every batch. If 'epoch', the metrics are pushed every epoch.
`exclude_train_metrics`	`bool`	`False`	Set to true to exclude the training metrics.
`exclude_val_metrics`	`bool`	`False`	Set to true to exclude the validation metrics.
`exclude_predict_metrics`	`bool`	`False`	Set to true to exclude the prediction metrics.
`exclude_metrics`	`List[str]`	`[]`	List the metrics to remove from tracking. For validation metrics, add 'val_'. For instance, use `exclude_metrics=['mse', 'val_mse']` to exclude the `mse` metrics in training and validation.