Linear code in aten/src for cuda?

1

I can find the Linear routines to run the forward calculations and calculate gradients, e.g.

grep -r ‘(Linear’ pytorch/aten | grep API
pytorch/aten/src/THNN/generic/THNN.h:TH_API void THNN_(Linear_updateOutput)(
pytorch/aten/src/THNN/generic/THNN.h:TH_API void THNN_(Linear_updateGradInput)(
pytorch/aten/src/THNN/generic/THNN.h:TH_API void THNN_(Linear_accGradParameters)(

but can’t locate the equivalent routines in THCUNN
?

thanks

2

Actually that was legacy code path. The nn.Linear uses matmul and a plus https://github.com/pytorch/pytorch/blob/master/torch/nn/functional.py#L959