Merge pull request #85764 from j-martens/patch-608

Update dsvm-deep-learning-ai-frameworks.md

Author: PRMerger12

articles/machine-learning/data-science-virtual-machine/dsvm-deep-learning-ai-frameworks.md

@@ -274,3 +274,9 @@ The service takes about one minute to start. Start a web browser and navigate to
 |[Theano](https://github.com/Theano/Theano) | No | Yes (Ubuntu) |Theano is installed in Python 2.7 (_root_), as well as Python 3.5 (_py35_) environment.<br/><br/>**To run it**: <br/>* Terminal: Activate the Python version you want (root or py35), run python, then import theano.<br/>* Jupyter: Select the Python 2.7 or 3.5 kernel, then import theano.  <br/>To work around a recent MKL bug, you need to first set the MKL threading layer:<br/><br/>_export MKL_THREADING_LAYER=GNU_|
 |[Chainer](https://chainer.org/) |No | Yes |Chainer is installed in [Python 3.5](dsvm-languages.md#python-linux-and-windows-server-2012-edition). ChainerRL and ChainerCV are also installed. <br/><br/>Sample notebooks are included in JupyterHub.<br/><br/>**To run it**: <br/>* Terminal: Activate the [Python 3.5](dsvm-languages.md#python-linux-and-windows-server-2012-edition) environment, run _python_, then import chainer. <br/> * JupyterHub: [Connect to JupyterHub](dsvm-ubuntu-intro.md#how-to-access-the-data-science-virtual-machine-for-linux), then navigate to the Chainer directory to find sample notebooks.|
 |[NVidia Digits](https://github.com/NVIDIA/DIGITS) | No | Yes (Ubuntu) |Deep learning system from NVIDIA for rapidly training deep learning models. DIGITS is installed in `/dsvm/tools/DIGITS` and is available a service called _digits_.  <br/><br/>**To run it**: <br/>Log in to the VM with X2Go. At a terminal, start the service ```sudo systemctl start digits```. <br/><br/>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.|
+|[CUDA, cuDNN, NVIDIA Driver](https://developer.nvidia.com/cuda-toolkit) |Yes | Yes | |
+|nvidia-smi|Yes | Yes |NVIDIA tool for querying GPU activity. _nvidia-smi_ is available on the system path. <br/><br/>Start a command prompt (on Windows) or a terminal (on Linux), then run _nvidia-smi_.|
+|[TensorFlow Serving](https://www.tensorflow.org/serving/) | No | Yes |A server to inference on a TensorFlow model. _tensorflow_model_server_ is available at the terminal. Samples are available [online](https://www.tensorflow.org/serving/).|
+|[TensorRT](https://developer.nvidia.com/tensorrt) |  No | Yes (Ubuntu) |A deep learning inference server from NVIDIA. TensorRT is installed as an _apt_ package. Samples are available [online](https://docs.nvidia.com/deeplearning/sdk/tensorrt-developer-guide/index.html#samples).|
+|[Microsoft Cognitive Toolkit (CNTK)](https://docs.microsoft.com/cognitive-toolkit/)|Yes | Yes | Installed in Python 3.5 on [Linux and Windows 2012](dsvm-languages.md#python-linux-and-windows-server-2012-edition) and Python 3.6 on [Windows 2016](dsvm-languages.md#python-windows-server-2016-edition). Sample Jupyter notebooks are included on DSVM. <br/><br/>**To run it**: <br/>Terminal: Activate the correct environment and run Python. <br/>Jupyter: Connect to [Jupyter](provision-vm.md#tools-installed-on-the-microsoft-data-science-virtual-machine) or [JupyterHub](dsvm-ubuntu-intro.md#how-to-access-the-data-science-virtual-machine-for-linux), then open the CNTK directory for samples. |
+|Deep Water|No | Yes (Ubuntu) |Deep learning framework for H2O, Deep Water is installed in [Python 3.5](dsvm-languages.md#python-linux-and-windows-server-2012-edition) and is also available in `/dsvm/tools/deep_water`. Sample notebooks are included in JupyterHub. 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:<br/><br/>```export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:${LD_LIBRARY_PATH}```<br/>```export CUDA_ROOT=/usr/local/cuda-8.0```<br/><br/>To use Deep Water:<br/>* Terminal: activate the [Python 3.5](dsvm-languages.md#python-linux-and-windows-server-2012-edition) environment, then run _python_. <br/>* JupyterHub: [connect to JupyterHub](dsvm-ubuntu-intro.md#how-to-access-the-data-science-virtual-machine-for-linux), then navigate to the deep_water directory to find sample notebooks.|