Skip to content

Latest commit

 

History

History
60 lines (43 loc) · 1.7 KB

Functions.md

File metadata and controls

60 lines (43 loc) · 1.7 KB

Functions

This guide contains info on the main functions that the user should interact with.

MLDashboardBackend

createDashboard()

Loads a dashboard in a seperate process. Returns process, updatelist, and return list.

Params:

  • config: Path to config file (default: dashboard.json)
  • waitforstart: Pauses main process until dashboard is ready (default: True)

MLCallbacksBackend

DashboardCallbacks

Inherits from keras.Callbacks.Callback. This contains all communication between model training and the dashboard.

Params:

  • updatelist: List from dashboard creation
  • returnlist: List from dashboard creation
  • model: Tensorflow model
  • x_train: Training set features
  • y_train: Training set output
  • x_test: Test set features
  • y_test: Test set output
  • prediction_labels: Allows images to be labeled with friendly text
  • config: Customize when data is sent

Example: model.fit() with callbacks

from MLDashboard.MLCallbacksBackend import DashboardCallbacks, CallbackConfig
config = CallbackConfig()
labels = ['zero', 'one', 'two', 'three', 'four', 'five', 'six', 'seven', 'eight', 'nine']
callback = DashboardCallbacks(updatelist, returnlist, model, x_train, y_train, x_test, y_test, labels, config)
model.fit()

MLCommunicationBackend

MessageMode

Message mode is an enum with different types of messages that the dashboard should receieve or send.

Message

Message is a class that contains a mode and a data payload. All data going to and from the dashboard is a Message.

Example: Send the dashboard the end message

from MLDashboard.MLCallbacksBackend import Message, MessageMode
updatelist.append(Message(MessageMode.End, {})) #message does not need a payload