https://github.com/neuralyzer/kerasvis

Live visualization of the training of Keras models
https://github.com/neuralyzer/kerasvis

Last synced: 2 months ago
JSON representation

Live visualization of the training of Keras models

• Host: GitHub
• URL: https://github.com/neuralyzer/kerasvis
• Owner: neuralyzer
• License: gpl-3.0
• Created: 2016-04-06T15:16:24.000Z (about 9 years ago)
• Default Branch: master
• Last Pushed: 2017-02-08T13:23:33.000Z (over 8 years ago)
• Last Synced: 2025-03-21T08:48:28.556Z (3 months ago)
• Language: Python
• Size: 199 KB
• Stars: 22
• Watchers: 2
• Forks: 3
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
Keras live visualization

========================

This package visualizes ongoing Keras optimizations live in your browser.

Optimization runs are logged in a database. To use this visualization you have to

add `DBLogger` as callback to your optimization. 

Installation

------------

Kerasvis is developed for Python 3 only.

It does not work with Python 2.

There are no plans to add Python 2 support.

Install via pip the [kerasvis](https://pypi.python.org/pypi/kerasvis) package

    pip install kerasvis

or install directly from github:

    pip install git+git://github.com/neuralyzer/kerasvis.git 

Quickstart example

------------------

    from kerasvis import DBLogger

    from keras.models import Sequential

    from keras.layers import Dense

    import scipy as sp

    X = sp.rand(200, 20)

    y = (sp.rand(200, 1) < .5).astype(int)

    model = Sequential()

    model.add(Dense(200, input_dim=20, activation="sigmoid"))

    model.add(Dense(1, activation="sigmoid"))

    model.compile("sgd", "binary_crossentropy")

    logger = DBLogger(comment="An example run")

    history = model.fit(X, y, nb_epoch=100, batch_size=64, verbose=0,

                        validation_split=0.2, callbacks=[logger]

Check the docstring of the DBLogger class for more information.

Start the keras visualization server with

    export FLASK_APP=kerasvis.runserver

    flask run

    

You'll see a web output similar to the one below

![Loss Accuracy screenshot](loss_accuracy_screenshot.png)

Delete runs from the database

-----------------------------

Runs can be deleted with the `DBLogger`. To delete the run with id 3 you do

    db_logger = DBLogger(id=3)

    db_logger.delete()

    

you will then be asked if you really want to delete and have to confirm with "yes".

Runs can also be delted via the web interface. To disable deletion via the webinterface

set the environment variable

    KERASVIS_ALLOW_DELETE=False