I was trying to use the DCNv2 which uses the
THFloatBlas_gemm. Now since that has been removed in Pytorch 1.7, I was wondering where I could find the source code of the same function so that I can define the said function and make the model.
You could (and should) use
torch::addmm instead as described here.
Thanks a lot for replying @ptrblck, will do that