fixed convert

example/42_groupnorm/groupnorm_sigmoid_mul_fp16.cpp

@@ -14,14 +14,14 @@ using ComputeDataType = float;
 
 struct YElementOp
 {
-    template <typename T>
-    __host__ __device__ void operator()(T& y, const T& x) const
+    template <typename Y, typename X>
+    __host__ __device__ void operator()(Y& y, const X& x) const
     {
-        static_assert(ck::is_same<T, float>::value || ck::is_same<T, double>::value ||
-                          ck::is_same<T, ck::half_t>::value,
+        static_assert(ck::is_same<X, float>::value || ck::is_same<X, double>::value ||
+                          ck::is_same<X, ck::half_t>::value,
                       "Data type is not supported by this operation!");
 
-        T a;
+        X a;
 
         ck::tensor_operation::element_wise::Sigmoid{}(a, x);
 

include/ck/tensor_operation/gpu/element/unary_element_wise_operation.hpp

@@ -87,6 +87,12 @@ struct PassThrough
         y = type_convert<int8_t>(x);
     }
 
+    template <>
+    __host__ __device__ void operator()<int8_t, float>(int8_t& y, const float& x) const
+    {
+        y = type_convert<int8_t>(x);
+    }
+
 #ifdef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
     template <>
     __host__ __device__ void operator()<int4_t, int4_t>(int4_t& y, const int4_t& x) const