Extend available elementwise operations with conv examples (#995)

* Extend available elementwise operations with conv examples

* Fixes

* Remove not needed convert

* Update CMakeFile and dir name
This commit is contained in:
Bartłomiej Kocot
2023-10-19 17:23:19 +02:00
committed by GitHub
parent deef92d5d0
commit 82f3a835d5
17 changed files with 758 additions and 50 deletions

View File

@@ -128,11 +128,9 @@ struct ReferenceConvFwd : public device::BaseOperator
}
}
float v_out;
arg.out_element_op_(v_out, v_acc);
arg.output_(g, n, k, wo) = ck::type_convert<OutDataType>(v_out);
OutDataType v_out;
arg.out_element_op_(v_out, ck::type_convert<OutDataType>(v_acc));
arg.output_(g, n, k, wo) = v_out;
};
make_ParallelTensorFunctor(func,
@@ -184,11 +182,9 @@ struct ReferenceConvFwd : public device::BaseOperator
}
}
float v_out;
arg.out_element_op_(v_out, v_acc);
arg.output_(g, n, k, ho, wo) = ck::type_convert<OutDataType>(v_out);
OutDataType v_out;
arg.out_element_op_(v_out, ck::type_convert<OutDataType>(v_acc));
arg.output_(g, n, k, ho, wo) = v_out;
};
make_ParallelTensorFunctor(func,
@@ -253,11 +249,9 @@ struct ReferenceConvFwd : public device::BaseOperator
}
}
float v_out;
arg.out_element_op_(v_out, v_acc);
arg.output_(g, n, k, d_o, ho, wo) = ck::type_convert<OutDataType>(v_out);
OutDataType v_out;
arg.out_element_op_(v_out, ck::type_convert<OutDataType>(v_acc));
arg.output_(g, n, k, d_o, ho, wo) = v_out;
};
make_ParallelTensorFunctor(func,