Zebra – Faster Neural Network Inference for AWS F1
Mipsology offers FPGA-based class-leading acceleration for Deep Learning, with no FPGA knowledge required. We leverage more than 20 years designing high-performance FPGA-based systems for Linux to offer our users the best solutions for Deep Learning.
Mipsology's first product, Zebra, offers users FPGA-based class leading acceleration for Neural network inference. There is zero FPGA knowledge required nor a single line of code to write to use Zebra. Zebra runs user defined neural network just as it would on GPU or CPU, switching takes minutes.
High Performance
FPGA are hardware programmable. They offer best-in-class computation throughput for neural network. Zebra executes neural network fast on one or multiple FPGA.
Supports All Neural Networks
User-defined neural networks work on Zebra as they would on CPU and GPU, without changing them. No new engineering effort is required to migrate to EC2 F1.
No FPGA Knowledge Required
No FPGA details are exposed to our users. Zebra users focus on solving their AI problems, Zebra makes it run on FPGA.
Already Integrated into Your Environment
Zebra users don’t have to learn new languages, new infrastructures or new tools, it works with Caffe, MXNET, TensorFlow, etc.
Fast Set-up
Select the AMI on the AWS Marketplace, launch it, link your application to the Zebra library, run it!
Zebra supports 16-bit fixed point, 8-bit fixed point, and more.
User-defined neural networks work on Zebra as they would on CPU and GPU, without changing them. No new engineering effort is required to migrate to EC2 F1.
Same Silicon, Better Performance
Mipsology R&D works to continuously improve performance with the same FPGA. You will get further acceleration with new AMI versions, no need to buy the newest GPU.
Reach Your Goals Sooner
Zebra does not require you compile any FPGA, which can take hours. F1 has all the hardware you need. Zebra runs immediately at high speed, you iterate faster and reach your goals sooner.
Scales with your Needs
Run one Zebra or hundreds, Zebra on AWS scales with your needs. Zebra runs on instances with 1 FPGA or instances with 8 FPGA.
Versatile
No matter whether your neural network is used for classification or recognition, data or content. Just plug it in and Zebra makes it faster.
Low Cost
No up-front cost for special hardware, pay only the time used to accelerate your neural network on ES2 F1. Scale up and down easily.
Looking for More Solutions?
If you need support for a new neural network, help to run even faster, some in-house Zebra solutions or inquire about our technology, just contact us at zebra@mipsology.com.
Mipsology Zebra 1 FPGA for Caffe
(Education & Open Source Community)
- Free of SW charges (you only pay AWS for the F1 instance)
- Ready to run image-classification demonstration
- Pre-trained AlexNet and GoogLeNet
- Pre-installed Caffe infrastructure
- Try your own weights and applications
Mipsology Zebra 1 FPGA for Caffe
(F1C500)
- 30 days of Free Trial
- Supports your CNN similar to AlexNet and GoogLeNet
- Pre-installed Caffe infrastructure
- Supports int16 precision
- Pre-trained CNN models
Look for the Zebra on AWS Marketplace, it’s Free*!
(*) AWS F1 charges apply
