[dnn/shift] optimizations for NCHW layout
This commit is contained in:
@@ -33,7 +33,7 @@ int main() {
|
||||
triton::dnn::shift shift(B, C, 1, H, W, 1, R, S, F, 1, 1,
|
||||
shift_h.data(), shift_w.data(),
|
||||
numeric_t_str, numeric_t_str,
|
||||
op, false);
|
||||
op, false, triton::dnn::shift::NCHW);
|
||||
// host buffers
|
||||
std::vector<NumericT> ha(B*C*H*W);
|
||||
std::vector<NumericT> hb(C*F);
|
||||
|
Reference in New Issue
Block a user