Microsoft Cognitive Toolkit
Prix de Microsoft Cognitive Toolkit : gratuit
Version : 2.2
Microsoft Cognitive Toolkit (CNTK) is a library for deep learning, featuring™ excellent scaling, high speed and precision calculations. The library components allow you to process multidimensional data from Python, C ++ and BrainScript. There are different types of neural networks supported in Microsoft Cognitive Toolkit: FFN, CNN, RNN / LSTM, etc.
Main advantages of Microsoft Cognitive Toolkit are following:
- Properly optimized embedded components
- Improved convenience with working on large data sets
- Suitable for commercial solutions using AI
- Efficient use of resources
- Availability of built-in neural network training algorithms
Leaders in data science, as well as large companies such as Skype, Cortana, Bing, Xbox, already using the Microsoft Cognitive Toolkit in their tasks. You can use LeaderGPU® to work with Microsoft Cognitive Toolkit.
1. Install Nvidia® CUDA® 9.0 (all commands on https://developer.nvidia®.com/cuda-90-download-archive).
sudo dpkg -i cuda-repo-ubuntu1604-9-0-local_9.0.176-1_amd64.deb
sudo apt-key add /var/cuda-repo-9-0-local/7fa2af80.pub
sudo apt-get update
sudo apt-get install cuda®
2. Install NVIDIA® cuDNN 7.0 for CUDA® 9.0
3. Install GPU Drivers.
4. Install CUB
wget https://github.com/NVlabs/cub/archive/1.7.4.zip
unzip ./1.7.4.zip
sudo cp -r cub-1.7.4 /usr/local
5. Install OpenMPI
sudo apt-get install openmpi-bin
6. install the GPU version of CNTK:
pip install cntk-gpu
7. Export MKL path to environment variable LD_LIBRARY_PATH, e.g.:
export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
8. Activate CNTK
python
import cntk
9. Check CNTK version:
cntk.__version__
Catégorie
Machine learning
Secteur d’activité
Automotive, Oil & Gas, Electronic, Biotechnology, Finance
Licence
Opensource/Free