Honored Contributor
Honored ContributorA few of the popular Machine Learning frameworks (Caffe, Theano, Tensorflow) have experimental OpenCL branches (search google, they are easy to find). Though of course they might not work at all out of the box and they will definitely not work on FPGAs out of the box and you will have to do some modifications to get them to work on an FPGA. Needless to say, the initial performance on an FPGA will be terrible since the kernels are optimized for GPUs, and if you want proper performance, you also have to manually optimize the kernels.
