edit

Environments

Below is the list of Deep Learning environments supported by Floyd. Any of these can be specified in the floyd run command using the --env option.

If no --env is provided, it uses the keras image, which comes with a Python 3 image containing Keras 2.0.4 and Tensorflow 1.1.0 pre-installed.

Framework Env name (--env parameter) Description Docker Image
Tensorflow 1.1 tensorflow Tensorflow 1.1.0 + Keras 2.0.4 on Python3. floydhub/tensorflow:1.1.0-py3
tensorflow:py2 Tensorflow 1.1.0 + Keras 2.0.4 on Python2. floydhub/tensorflow:1.1.0-py2
Tensorflow 1.0 tensorflow-1.0 Tensorflow 1.0.0 + Keras 1.2.2 on Python3. floydhub/tensorflow:1.0.0-py3
tensorflow-1.0:py2 Tensorflow 1.0.0 + Keras 1.2.2 on Python2. floydhub/tensorflow:1.0.0-py2
Tensorflow 0.12 tensorflow-0.12 Tensorflow 0.12.1 + Keras 1.2.2 on Python3. floydhub/tensorflow:0.12.1-py3
tensorflow-0.12:py2 Tensorflow 0.12.1 + Keras 1.2.2 on Python2. floydhub/tensorflow:0.12.1-py2
Theano theano Theano rel-0.8.2 + Keras 1.2.2 on Python3. floydhub/theano:latest-py3
theano:py2 Theano rel-0.8.2 + Keras 1.2.2 on Python2. floydhub/theano:latest-py2
Keras - Use tensorflow or theano for the appropriate Keras backend
Caffe caffe Caffe rc4 on Python3. floydhub/caffe:latest-py3
caffe:py2 Caffe rc4 on Python2. floydhub/caffe:latest-py2
Torch torch Torch 7 with Python 3 env. floydhub/torch:latest-py3
torch:py2 Torch 7 with Python 2 env. floydhub/torch:latest-py2
PyTorch pytorch PyTorch 0.1.9 on Python 3. floydhub/pytorch:latest-py3
torch:py2 PyTorch 0.1.9 on Python 2. floydhub/pytorch:latest-py2
Chainer (beta) chainer Chainer 1.21.0 on Python 3. floydhub/chainer:latest-py3
chainer:py2 Chainer 1.21.0 on Python 2. floydhub/chainer:latest-py2
MxNet (beta) mxnet:py2 MxNet 0.9.3a on Python 2. floydhub/mxnet:latest-py2
Kur kur Kur 0.3.0 on Python 3. floydhub/kur:latest-py3

All environments are available for both CPU and GPU execution. For example,

To run a Tensorflow job on CPU

$ floyd run --env tensorflow:py2 "python mnist_cnn.py" 

To run a Tensorflow job on GPU (CUDA, cuDNN, etc. installed)

$ floyd run --env tensorflow:py2 --gpu "python mnist_cnn.py" 

The following software packages (in addition to many other common libraries) are available in all the environments:

h5py, iPython, Jupyter, matplotlib, numpy, OpenCV, Pandas, Pillow, scikit-learn, scipy, sklearn

Help make this document better

This guide, as well as the rest of our docs, are open-source and available on GitHub. We welcome your contributions.