Hualos - Keras Total Visualization project
For now, this is a simple demo where a Flask server exposes an API to publish and consume events in the form of JSON objects. The Keras callback RemoteMonitor can publish events to the server, and the Hualos landing page listens to the server and displays incoming data on a c3.js graph.
Example:
- start the server:
python api.py - load the landing page:
http://localhost:9000/ - launch a Keras experiment with the
RemoteMonitorcallback:
from keras import callbacks
remote = callbacks.RemoteMonitor(root='http://localhost:9000')
model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, validation_data=(X_test, Y_test), callbacks=[remote])
- the available metrics of the experiment will start being graphed in real time on the landing page.
Dependencies:
-
Python:
- Flask
- gevent
-
JS (included in the repo):
- d3.js
- c3.js