We’re excited to announce that the newest version of our Bitfusion Ubuntu 14 Tensorflow AMI is now available in the AWS Marketplace. We’ve also made some exciting new changes to our pricing model to help save you money.
Tensorflow is a popular open source framework for numerical computation using data flow graphs. Originally developed by engineers on the Google Brain Team for machine learning and deep neural nets, the nodes represent mathematical equations, and the graph edges connecting them are represented as multi-dimensional arrays (tensors).
The latest AMI includes Tensorflow 0.11, with new features such as:
This release marks the first AMI with our new pricing scheme. Going forward, if an AMI is deployed on a t2.nano, t2.micro, or t2.small, there will be no Bitfusion software fee on top of the AWS charges. We want to make it even simpler to use Bitfusion to test out the latest release of your favorite deep learning framework or to do some initial model prototyping. Also, for large instances, we’ve capped our Bitfusion software fee to $0.297/hr. For a p2.16xlarge, that is almost a 5x in software savings!
So, bottom line:
Whether you prefer Python 2 or Python 3, we’ve added new and updated tools, libraries, and frameworks to support and handle both.
We’re adding the following libraries to provide a more comprehensive Tensorflow toolkit for data science and deep learning.
Matplotlib provides an extensive library of graphs and charts that can be implemented with just a few lines of Python.
NumPy is a Python library that provides a variety of fundamental scientific computing functions, including convenient and fast N-dimensional array manipulation.
SciPy builds on top of NumPy with modules for statistics, optimization, integration, linear algebra, Fourier transforms, signal and image processing, ODE solvers, and more.
Pandas is library of data structures and data analysis tools that extend Python’s data munging and prep strengths with powerful data analysis and modeling capabilities.
SymPy is a lightweight, easy-to-use Python library for symbolic mathematics that aims to provide a full-featured computer algebra system (CAS).
Hyperas is a wrapper around hyperopt, allowing you to use the power of hyperopt without having to learn confusing syntax. The result? Faster, easier prototyping with keras models.
PyCUDA is rapidly becoming the preferred way to interact with Python’s NVIDIA CUDA because it effectively handles memory management conveniently and comprehensively. By bundling this capability with the ability to compile CUDA Kernels directly from Jupyter notebooks, you have the ability to interact with the CUDA parallel computation API at a very deep level while also retaining the simplicity and ease of a notebook user interface.
AWS recently announced their next generation GPU P2 instances. This new generation provides up to 16 NVIDIA K80 GPUs, 64 vCPUs and 732 GiB of host memory. This is a welcome upgrade, given that the previous generation GPU G2 instances were starting to show their age. In the previous release of our Bitfusion Tensorflow AMI, we included updated NVIDIA drivers, the CUDA toolkit, and CUDNN support, allowing you to tap into these new powerful instances.
All of our AMIs include a 5 day free trial. Please try the Bitfusion Ubuntu 14 Tensorflow AMI available in the AWS Marketplace, today.