Permalink
Fetching contributors…
Cannot retrieve contributors at this time
261 lines (199 sloc) 13.9 KB
title description keywords services documentationcenter author manager ms.assetid ms.service ms.component ms.workload ms.tgt_pltfrm ms.devlang ms.topic ms.date ms.author
Deep Learning and AI frameworks - Azure | Microsoft Docs
Deep Learning and AI frameworks
data science tools, data science virtual machine, tools for data science, linux data science
machine-learning
gopitk
cgronlun
machine-learning
data-science-vm
data-services
na
na
conceptual
09/11/2017
gokuma

Deep Learning and AI frameworks

The Data Science Virtual Machine (DSVM) and the Deep Learning VM supports a number of deep learning frameworks to help build Artificial Intelligence (AI) applications with predictive analytics and cognitive capabilities like image and language understanding.

Here are the details on all the deep learning frameworks available on the DSVM.

Microsoft Cognitive Toolkit

What is it? Deep learning framework
Supported DSVM Editions Windows, Linux
How is it configured / installed on the DSVM? The Microsoft Cognitive Toolkit (CNTK) is installed in Python 3.5 on Linux and Windows 2012 and Python 3.6 on Windows 2016.
Links to Samples Sample Jupyter notebooks are included.
Related Tools on the DSVM Keras
How to use / run it? * At a terminal: activate the correct environment, then run Python.
  • In Jupyter: Connect to Jupyter or JupyterHub, then open the CNTK directory for samples. |

TensorFlow

What is it? Deep learning framework
Supported DSVM Editions Windows, Linux
How is it configured / installed on the DSVM? TensorFlow is installed in Python 3.5 on Linux and Windows 2012 and Python 3.6 on Windows 2016.
Links to Samples Sample Jupyter notebooks are included.
Related Tools on the DSVM Keras
How to use / run it? * At a terminal: activate the correct environment, then run Python.
  • In Jupyter: Connect to Jupyter or JupyterHub, then open the TensorFlow directory for samples. |

Horovod

What is it? Distribued deep learning framework for TensorFlow
Supported DSVM Editions Ubuntu
How is it configured / installed on the DSVM? Horovod is installed in Python 3.5 on Ubuntu.
Links to Samples https://github.com/uber/horovod/tree/master/examples
Related Tools on the DSVM TensorFlow
How to use / run it? At a terminal: activate the correct environment, then run Python.

Keras

What is it? High-level deep learning API
Supported DSVM Editions Windows, Linux
How is it configured / installed on the DSVM? TensorFlow is installed in Python 3.5 on Linux and Windows 2012 and Python 3.6 on Windows 2016.
Links to Samples https://github.com/fchollet/keras/tree/master/examples
Related Tools on the DSVM Microsoft Cognitive Toolkit, TensorFlow, Theano
How to use / run it? * At a terminal: activate the correct environment, then run Python.
  • In Jupyter: Download the samples from the Github location, connect to Jupyter or JupyterHub, then open the sample directory. |

Caffe

What is it? Deep learning framework
Supported DSVM Editions Ubuntu
How is it configured / installed on the DSVM? Caffe is installed in /opt/caffe.
How to switch to Python 2.7 Run source activate root
Links to Samples Samples are included in /opt/caffe/examples.
Related Tools on the DSVM Caffe2

How to use / run it?

Use X2Go to log in to your VM, then start a new terminal and enter

cd /opt/caffe/examples
source activate root
jupyter notebook

A new browser window opens with sample notebooks.

Binaries are installed in /opt/caffe/build/install/bin.

Installed version of Caffe requires Python 2.7 and won't work with Python 3.5 activated by default. Run source activate root to switch Anaconda environment.

Caffe2

What is it? Deep learning framework
Supported DSVM Editions Ubuntu
How is it configured / installed on the DSVM? Caffe2 is installed in the Python 2.7 (root) conda environment. The source is in /opt/caffe2.
Links to Samples Sample notebooks are included in JupyterHub.
Related Tools on the DSVM Caffe
How to use / run it? * At the terminal: activate the root Python environment, start Python, and import caffe2.
* In JupyterHub: connect to JupyterHub, then navigate to the Caffe2 directory to find sample notebooks. Some notebooks require the Caffe2 root to be set in the Python code; enter /opt/caffe2.
Build notes Caffe2 is built from source on Linux and includes CUDA, cuDNN, and Intel MKL. The current commit is 0d9c0d48c6f20143d6404b99cc568efd29d5a4be, which was chosen for stability on all GPUs and samples tested.

Chainer

What is it? Deep learning framework
Supported DSVM Editions Windows, Linux
How is it configured / installed on the DSVM? Chainer is installed in Python 3.5. ChainerRL and ChainerCV are also installed.
Links to Samples Sample notebooks are included in JupyterHub.
Related Tools on the DSVM Caffe
How to use / run it? * At the terminal: activate the Python 3.5 environment, run python, then import chainer.

Deep Water

What is it? Deep learning framework for H2O
Supported DSVM Editions Ubuntu
How is it configured / installed on the DSVM? Deep Water is installed in Python 3.5 and is also available in /dsvm/tools/deep_water.
Links to Samples Sample notebooks are included in JupyterHub.
Related Tools on the DSVM H2O, Sparkling Water

How to use / run it?

Deep Water requires CUDA 8 with cuDNN 5.1. This is not in the library path by default, as other deep learning frameworks use CUDA 9 and cuDNN 7. To use CUDA 8 + cuDNN 5.1 for Deep Water:

export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:${LD_LIBRARY_PATH}
export CUDA_ROOT=/usr/local/cuda-8.0

To use Deep Water:

  • At the terminal: activate the Python 3.5 environment, then run python.
  • In JupyterHub: connect to JupyterHub, then navigate to the deep_water directory to find sample notebooks.

MXNet

What is it? Deep learning framework
Supported DSVM Editions Windows, Linux
How is it configured / installed on the DSVM? MXNet is installed in C:\dsvm\tools\mxnet on Windows and /dsvm/tools/mxnet on Linux. Python bindings are installed in Python 3.5 on Linux and Windows 2012 and Python 3.6 on Windows 2016. R bindings are also installed on Ubuntu.
Links to Samples Sample Jupyter notebooks are included.
Related Tools on the DSVM Keras
How to use / run it? * At a terminal: activate the correct environment, then run Python.
  • In Jupyter: Connect to Jupyter or JupyterHub, then open the mxnet directory for samples. | | Build notes | MXNet is built from source on Linux. This build includes CUDA, cuDNN, NCCL, and MKL. |

NVIDIA DIGITS

What is it? Deep learning system from NVIDIA for rapidly training deep learning models
Supported DSVM Editions Ubuntu
How is it configured / installed on the DSVM? DIGITS is installed in /dsvm/tools/DIGITS and is available a service called digits.

How to use / run it?

Log in to the VM with X2Go. At a terminal, start the service:

sudo systemctl start digits

The service takes about one minute to start. Start a web browser and navigate to http://localhost:5000. Note that DIGITS does not provide a secure login and should not be exposed outside the VM.

nvidia-smi

What is it? NVIDIA tool for querying GPU activity
Supported DSVM Editions Windows, Linux
How is it configured / installed on the DSVM? nvidia-smi is available on the system path.
How to use / run it? Start a command prompt (on Windows) or a terminal (on Linux), then run nvidia-smi.

Theano

What is it? Deep learning framework
Supported DSVM Editions Ubuntu
How is it configured / installed on the DSVM? Theano is installed in Python 2.7 (root), as well as Python 3.5 (py35) environment.
Related Tools on the DSVM Keras
How to use / run it? * At a terminal, activate the Python version you want (root or py35), run python, then import theano.
  • In Jupyter, select the Python 2.7 or 3.5 kernel, then import theano.

To work around a recent MKL bug, you need to first set the MKL threading layer:

_export MKL_THREADING_LAYER=GNU_ |

Torch

What is it? Deep learning framework
Supported DSVM Editions Ubuntu
How is it configured / installed on the DSVM? Torch is installed in /dsvm/tools/torch. PyTorch is installed in Python 2.7 (root), as well as Python 3.5 (py35) environment.
Links to Samples Torch samples are located at /dsvm/samples/torch. PyTorch samples are located at /dsvm/samples/pytorch.

PyTorch

What is it? Deep learning framework
Supported DSVM Editions Linux
How is it configured / installed on the DSVM? PyTorch is installed in Python 3.5.
Links to Samples Sample Jupyter notebooks are included, and samples can also be found in /dsvm/samples/pytorch.
Related Tools on the DSVM Torch
How to use / run it
  • At a terminal: activate the correct environment, then run Python.
  • In Jupyter: Connect to JupyterHub, then open the PyTorch directory for samples. |

MXNet Model Server

What is it? A server to create HTTP endpoints for MXNet and ONNX models
Supported DSVM Editions Linux
How is it configured / installed on the DSVM? mxnet-model-server is available on at the terminal.
Links to Samples Look for up-to-date samples on the MXNet Model Server page.
Related Tools on the DSVM MXNet

TensorFlow Serving

What is it? A server to run inferencing on a TensorFlow model
Supported DSVM Editions Linux
How is it configured / installed on the DSVM? tensorflow_model_server is available at the terminal.
Links to Samples Samples are available online.
Related Tools on the DSVM TensorFlow

TensorRT

What is it? A deep learning inference server from NVIDIA.
Supported DSVM Editions Ubuntu
How is it configured / installed on the DSVM? TensorRT is installed as an apt package.
Links to Samples Samples are available online.
Related Tools on the DSVM TensorFlow Serving, MXNet Model Server