Enterprises collect large amounts of data and analyze it to obtain a competitive advantage. Some are using machine learning techniques to create predictive applications for fraud detection, demand forecasting, click prediction, and other data-intensive analyses. Recent advancements in machine learning make it possible to go even further, bringing deep learning within reach of developers everywhere. Now, computer vision, speech recognition, natural language processing, and audio recognition applications are being developed to give enterprises a competitive advantage.
The challenge: Building an enterprise-grade deep learning environment
Processing large amounts of data for deep learning requires large amounts of computational power. As new tools designed specifically for deep learning become available, developers are using them to build their applications on advanced clusters that take advantage of accelerators such as NVIDIA GPUs or Intel DL Boost.
When organizational advantage is tied to the insights achieved via deep learning, the underlying IT infrastructure needs to be deployed and managed as enterprise-grade, not as a lab experiment. However, building and managing an advanced cluster, installing the software that satisfies all of the library dependencies, and making it all work together can be an enormous challenge.
-Steve Conway, IDC
The Bright Solution for Deep Learning
Bright Cluster Manager for Data Science makes it faster and easier for organizations to gain actionable insights from rich, complex data. To achieve this, Bright offers a comprehensive deep learning solution that includes:
A modern and flexible deep learning environment
Bright provides everything needed to spin up an effective deep learning environment, and manage it effectively. Packages support a rich variety of Python interpreters, instruction sets (AVX/AVX2/FMA/SSE4.2 and AVX-512 VNNI) and accelerator libraries. As a result, administrators can easily set up a working environment suitable for their needs.
Choice of machine learning frameworks
Bright Cluster Manager provides a choice of machine learning frameworks, including Chainer, MXNet, PyTorch, TensorFlow, and Theano, to simplify your deep learning projects.
Choice of machine learning libraries and tools
Bright includes a selection of the most popular machine learning libraries and tools to help you access datasets and build your applications. These include NVIDIA hardware drivers, CUDA Toolkit, CUDA Deep Neural Network library (cuDNN), CUB, TensorRT, Collective Communications Library (NCCL), as well as OpenCV, Protocol Buffers, and XGBoost.
A scalable deep learning environment
Bright machine learning packages support Open Message Passing Interface (MPI) and can be used to run distributed applications. By taking advantage of high-level frameworks such as Horovod and Keras, it is possible to run MXNet, PyTorch and TensorFlow applications at scale.
The world of deep learning is very fast paced and it is our objective to make the set of tools we provide as current as possible. The following is a selection of packages supported in Bright Cluster Manager for Data Science as of December 2019. The latest list of machine learning packages available for the various Bright Cluster Manager versions can be found here.
- Chainer: 6.2.0
- CNTK: 2.7
- DyNet: 2.1
- fast.ai: 1.0.58
- Horovod: 0.17.1
- Keras: 2.3.1
- MXNet: 1.5.0
- PyTorch: 1.3.0
- TensorFlow: 2.0.0
- Theano: 1.0.4
Libraries and Tools
- Bazel: 0.24.1
- CUB: 1.8.0
- cuDNN: 7.5.0
- NCCL: 2.4.8
- OpenCV: 3.4.7
- Protocol Buffers: 3.7.1
- TensorRT: 18.104.22.168
- XGBoost: 0.90
- Matplotlib: 3.1.1
- Numba: 0.45.1
- NumPy: 1.16.4
- PyCUDA: 2019.1.2
- SciPy: 1.2.1
- seaborn: 0.9.0