This should be your first option when looking to work with Machine Learning toolboxes. Rendsolve has specially tailored this image for running heavy Machine Learning workloads without any kind of setup. Just launch and start coding away!
The building system for these images is heavily based on a very well tested recipe from a standard in the field: the Deepo project Deepo Project and includes all the most popular toolkits in the default Python environment.
After launching TensorBox, the following frameworks will be available for you:
This library was one of the main tools responsible for the booming in Machine Learning interest around 2016.
You can enjoy the last stable version of this library from you Tensorbox instance by importing it in your ipython shell/normal script in the standard way:
import tensorflow as tf
Note that it will automatically setup the underlying GPU and get it ready to do the grunt work needed by your models.
This library was key to opening the field of Deep Learning to less experienced developers, and provided a strong foundation for the apparition of many innovative neural architectures.
To start using the included version of keras, you will only need to import it from you code with:
This last contender on the field sports a much more pythonic interface, and a more direct and powerful GPU communication interface.
To start using this library, just import it with:
Additional options and how to import them:
Caffe is one of the pioneer Deep Learning libraries and was the responsible for the initial proliferation of deep learning GPU based architectures. To start using it, you simply write on an ipython console:
Caffe2 is a new effort in building a base Deep Learning library, modular, lightweight and scalable. It is now being merged with a common codebase with Pythorch. To use it, just import it via:
CNTK (Microsoft Cognitive ToolKit)
The Microsoft Cognitive Toolkit Microsofts take in the computational graph based deep learning toolkits space, and is slowly getting more attention in the last months.
Use the following import command to get access to cntk’s functionallity.
Mxnet is an Apache Sponsored Machine Learning toolkit, which has the support from many big names in the industry. In the last times, it has gained support as Keras backend, with very good performance improvements over Tensorflow in some cases.
To use it, just invoke the library with:
Sonnet is a recent higher level offering from Google, to simplify the implementation of complex Machine Learning models.
To start using the library, load it with the following statement:
Additional offerings from TensorBox
Additionally to the mentioned libraries, the Tensorbox images offers the last stable version of OpenCV and Jupyter out of the box.
To import OpenCV’s Python bindings, use:
Jupyter can be called from the command line simply by: