Skip to main content

Notifications for ML libraries (Keras, Chainer) using telegram bot

Project description


Notifier and manager bot for ML libraries: Keras and Chainer


  • Get status of your training: epoch, metrics
  • Temporarily suspend training and then continue (for example, to free up the C(G)PU for another task (not the memory))
  • Interrupt training immediately
  • Get updates per epoch/batch with metrics
  • Set verbose level:
    • 0 - only completion message;
    • 1 - execution status per epoch
    • 2 - execution status per epoch with on-line metrics (like Keras default output)
  • Easy to build your own notifier and callback

How to use

Set TOKEN and PROXY while creating the instance of NotifierTelegramMenu or NotifierTelegram


from notifyker import NotifierTelegramMenu
from notifyker.chainer import ExtensionNotifierReport

TOKEN = 'xxxx:yyy'
    'proxy_url': 'socks5h://ip:port', 
    'urllib3_proxy_kwargs': {'username': 'username', 'password': 'password'}}

# use NotifierTelegram in order to work without keyboard menu
nfk = NotifierTelegramMenu(TOKEN=TOKEN, PROXY=PROXY)
chainer_not = ExtensionNotifierReport(notifier=nfk)



from notifyker import NotifierTelegramMenu
from notifyker.keras import CallbackSimple

TOKEN = 'xxxx:yyy'
    'proxy_url': 'socks5h://ip:port', 
    'urllib3_proxy_kwargs': {'username': 'username', 'password': 'password'}}

# use NotifierTelegram in order to work without keyboard menu
nfk = NotifierTelegramMenu(TOKEN=TOKEN, PROXY=PROXY)
callback = CallbackSimple(notifier=nfk)

Set PROXY = None if not required


Enter /start command to your telegram bot. Now it started and you receive update messages.


Enter /menu to open keyboard menu with following options:

  • Status - get the status of the last completed epoch

  • Verbose - change verbose level or leave unchanged

Level 0:

Level 2:

  • Pause - suspend training process (with option block on_batch_end callback method)

  • Continue - continue suspended training process

  • Interrupt - stop training and stop bot updates


Also, you can use these commands directly by typing:

  • /status
  • /verbose - only get the current level of verbose (TODO: opportunity to change)
  • /pause
  • /continue
  • /interrupt
  • /help - get the description of commands|options

Your own notifier

import base class (NotifierBase or NotifierTelegram)

from notifyker import NotifierTelegram

Then create your own class with inheritance and implement/reimplement methods. Note: if you add some new command - add it to handler

from telegram.ext import CommandHandler  # if required

class MyOwnNotifyker(NotifierTelegram):
	def my_own_hello(self, bot, update):
		Send hello in reply

    def handlers(self):
        self.updater.dispatcher.add_handler(CommandHandler('hello', self.my_own_hello))

That's it!

Your own callback

Also, you can use your own callback in the same way. Let's say you want only to know when the training ends.

import time

from notifyker import CallbackBase

class MyOwnCallback(CallbackBase):
    def on_training_end(self):
        end_time = time.ctime(int(time.time()))

        self.notifier.message('Training completed {} in {}'.format(tag, end_time))

Register your own telegram bot

In order to create your own telegram bot - talk to - this is official bot to... create bots.

Also, you'll get a TOKEN of your bot, that will be used in the Notifier latter. Click on the name of your bot and the chat will open with your bot.

Also, BotFather allows to manage bot profile image, description, privacy options, etc.

Official introduction is here:

Notifier for TF

It is possible to use this for tensorflow, but you should add necessary functions in your tf code. Just create:

nfk = NotifierTelegramMenu(TOKEN=TOKEN, PROXY=PROXY)

For example, in order to implement pause, after your epoch add:

if 'p' in nfk.flags_batch:
    while 'c' not in nfk.flags_batch and 's' not in nfk.flags_batch:

You can look at callbackNK/ to find more information about implementation.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

notifyker-2.0.3.tar.gz (9.7 kB view hashes)

Uploaded source

Built Distribution

notifyker-2.0.3-py3-none-any.whl (12.5 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page