Accelerating machine learning algorithms using FPGAs and GPUs.
BUILT FOR EASE-OF-USE
- Integrated with Kestrel Runtime
- Parameterizable
OPTIMIZED FOR PERFORMANCE
- Highly optimized for supported accelerator cards
DESIGNED FOR PORTABILITY
- Support for FPGAs and GPUs
- Retargetable
BUILT-IN INTEGRATION WITH SPARK
- Integration with Spark for MLlib and Caffe for CNN for transparent acceleration