techytushar / tf_notification_callback

TensorFlow/Keras Callback for receiving notifications

TensorFlow Notification Callback

A Tensorflow/Keras callback which sends information about your model training, on various messaging platforms.

Installation

Using pip:

pip install tf_notification_callback

Usage

Import the required module and add it to the list callbacks while training your model.

Example:

>>> from tf_notification_callback import TelegramCallback
>>> telegram_callback = TelegramCallback('<BotToken>',
                                         '<ChatID>',
	                                 'CNN Model',
	                                 ['loss', 'val_loss'],
	                                 ['accuracy', 'val_accuracy'],
	                                 True)
>>> model.fit(x_train, y_train,
              batch_size=32,
              epochs=10,
              validation_data=(x_test, y_test),
              callbacks=[telegram_callback])

Telegram

1. Create a telegram bot using BotFather
   • Search for @BotFather on telegram.
   • Send /help to get list of all commands.
   • Send /newbot to create a new bot and complete the setup.
   • Copy the bot token after creating the bot.
2. Get the chat ID
   • Search for the bot you created and send it any random message.
   • Go to this URL https://api.telegram.org/bot<BOT_TOKEN>/getUpdates (replace <BOT_TOKEN> with your bot token)
   • Copy the chat id of the user you want to send messages to.
3. Use the TelegramCallback() class.

TelegramCallback(bot_token=None, chat_id=None, modelName='model', loss_metrics=['loss'], acc_metrics=[], getSummary=False):

Arguments:

• bot_token : unique token of Telegram bot {str}
• chat_id : Telegram chat id you want to send message to {str}
• modelName : name of your model {str}
• loss_metrics : loss metrics you want in the loss graph {list of strings}
• acc_metrics : accuracy metrics you want in the accuracy graphs {list of strings}
• getSummary : Do you want message for each epoch (False) or a single message containing information about all epochs (True). {bool}

Slack

1. Create a Slack workspace
2. Create a new channel
3. Search for the Incoming Webhooks in the Apps and install it.
4. Copy the Webhook URL
5. Use the SlackCallback() class.

SlackCallback(bot_token=None, chat_id=None, modelName='model', loss_metrics=['loss'], acc_metrics=[], getSummary=False):

Arguments:

• webhookURL : unique webhook URL of the app {str}
• channel : channel name or username you want to send message to {str}
• modelName : name of your model {str}
• loss_metrics : loss metrics you want in the loss graph {list of strings}
• acc_metrics : accuracy metrics you want in the accuracy graph {list of strings}
• getSummary : Do you want message for each epoch (False) or a single message containing information about all epochs (True). {bool}

Sending images in Slack is not supported currently.

ToDo

• WhatsApp
• E-Mail
• Zulip
• Messages

Motivation

As the Deep Learning models are getting more and more complex and computationally heavy, they take a very long time to train. During my internship, people used to start the model training and left it overnight. They could only check its progress the next day. So I thought it would be great if there was a simple way to get the training info remotely on their devices.