FPGA-Accelerated ML Suite for Distributed Systems
Product Overview
InAccel FPGA-Accelerated ML (AML) suite provides a set of hardware accelerators for Amazon AWS f1.2x, f1.4x and f1.16x instances. It is shipped as a fully integrated AMI/AFI bundle offering an accelerated machine learning library for Apache Spark framework. InAccel's novel FPGA manager handles any available FPGA resources allowing the seamless scalability not only from one to many FPGAs but also to multiple F1 nodes.
This version allows the acceleration of Logistic Regression with Gradient Descent and K-Means ML applications. AML suite provides all the required APIs in Scala, Java and Python for the utilization of FPGA accelerators without requiring any changes in user's original code. InAccel's accelerated hardware modules promise up to 12x speedup compared to the SW-only execution on AWS F1 instances.