1 | initial version |
if you have a decent GPU, you can try the opencl target, but there is no support for cuda.
and would the workaround this would be to get all the weights and write a cuda version of forward pass myself??
good luck with that idea, it means rewriting the whole dnn module on your own.
2 | No.2 Revision |
if you have a decent GPU, you can try the opencl target, target, but there is no support for cuda.cuda at all here.
and would the workaround this would be to get all the weights and write a cuda version of forward pass myself??
good luck with that idea, it means rewriting the whole dnn module on your own.