Get your neural network to talk to you over Telegram, work in progress.
ANNxious is not like the other bots, one day he discovered he could talk to Keras' Artificial Neural Networks and since then it's the only thing he talks about. Add him on Telegram at ANNxiousBot.
- Install the custom Keras' Callback:
pip3 install annxious-callback
- Talk to ANNxious to get your user id.
- Pass the callback to your model like this:
from annxious_callback import ANNxiousRemoteMonitor
log = model.fit(
train_set_X,
train_set_y,
epochs=a_lot,
callbacks=[
ANNxiousRemoteMonitor(
user_id=132562286, # Some number provided by ANNxious
network_id='MyNet', # Whatever you want to name this model
url='http://im_not_hosting_yet.com'
)
]
)
- That's it.
As data scientists we care about data. The only data coming
out of your machine are the metrics you can see in each epoch
line yielded by model.fit
, which are completely meaningless
without context. ANNxious does not have any kind of access to
any part of your dataset.
ANNxious is written in Python 3 and uses PostgreSQL as its
backend to store the users and the metrics of the models,
although another backend can be used without major
modifications.
Other than setting up the database you will need to
create a config module to store the credentials, install the
requirements.txt
and you are good to go, you will need to
register the bot with Telegram's BotFather
to get your own Telegram's API token.
Feel free to submit any bug reports or feature requests in the issues section.