0.3.0 🌬️

A custom callback for email (via SMTP) integration is introduced, called EmailCallback:

import tensorflow as tf
from tf_notify import EmailCallback

model = tf.keras.Sequential(name="neural-network")
model.add(tf.keras.layers.Dense(1, input_dim=784))
model.compile(
    optimizer=tf.keras.optimizers.RMSprop(learning_rate=0.1),
    loss="mean_squared_error",
    metrics=["mean_absolute_error"],
)

model.fit(
    x_train,
    y_train,
    batch_size=128,
    epochs=2,
    verbose=0,
    validation_split=0.5,
    callbacks=[
        EmailCallback(
            to="[email protected]",
            from_="[email protected]",
            host="smtp.mail.yahoo.com",
            port=465,
            username="my-cool-username",
            password="my-cool-password",  # one-time app password
            ssl=True,
        )
    ],
)

Documentation:
The public atrributes for all available callbacks are now documented

0.2.0 ☀️

Instead of the very strict tensorflow = "~2.9.1" required dependency, we now support all Tensorflow versions between v2.7.3 and v2.9.1.

Additionally, CI matrix has been expanded:

matrix:
  os: [ubuntu-latest, macos-latest, windows-latest]
  python-version: ['3.7', '3.8', '3.9', '3.10']
  tf-version: ['2.7.3', '2.8.2', '2.9.1']
  exclude:
    - python-version: '3.10'
      tf-version: '2.7.3'

in order to test any new code against the allowed Tensorflow versions as well (in combination with os and python-version). Note that any matrix combination that consists of the sub-combination (python-version, tf-version) = ('3.10', '2.7.3') is excluded from the matrix, as support for python 3.10 has been introduced with tensorflow 2.8.

0.1.0 ⛰️

A custom callback for Telegram integration is introduced, called TelegramCallback:

import tensorflow as tf
from tf_notify import TelegramCallback


# define the tf.keras model to add callbacks to
model = tf.keras.Sequential(name='neural-network')
model.add(tf.keras.layers.Dense(1, input_dim=784))
model.compile(
    optimizer=tf.keras.optimizers.RMSprop(learning_rate=0.1),
    loss="mean_squared_error",
    metrics=["mean_absolute_error"],
)

model.fit(
    x_train,
    y_train,
    batch_size=128,
    epochs=2,
    verbose=0,
    validation_split=0.5,
    callbacks=[
        TelegramCallback(token='XXXX:YYYY', chat_id=-1234)
    ],  # send a Telegram notification when training ends!
)

0.1.0.beta1 🌊

This is the inaugural release of tf-notify! 🎉🎉🎉

It is considered a beta release.

Features:

A custom callback for Slack integration is introduced, called SlackCallback:

import tensorflow as tf
from tf_notify import SlackCallback


# define the tf.keras model to add callbacks to
model = tf.keras.Sequential(name='neural-network')
model.add(tf.keras.layers.Dense(1, input_dim=784))
model.compile(
    optimizer=tf.keras.optimizers.RMSprop(learning_rate=0.1),
    loss="mean_squared_error",
    metrics=["mean_absolute_error"],
)

model.fit(
    x_train,
    y_train,
    batch_size=128,
    epochs=2,
    verbose=0,
    validation_split=0.5,
    callbacks=[
        SlackCallback(webhook_url='https://url.to/webhook')
    ],  # send a Slack notification when training ends!
)

Also, users are able to inspect the package version via:

import tf_notify


tf_notify.__version__
'0.1.0.beta1'

Documentation:

Project documentation is served in a continuous fashion over at https://tf-notify.readthedocs.io.

The following packages are utilized:

• mkdocs: Project documentation with Markdown
• mkdocs-material: A Material Design theme for MkDocs
• mkdocstrings: Automatic documentation from sources, for MkDocs.

To interact with the project's documentation locally, run:

mkdocs serve

inside a virtual environment.

CI/CD:

Continuous Integration and Continuous Delivery for the package is enabled through Github Actions.

For more details, see:

• .github/workflows/ci.yml
• .github/workflows/cd.yml