Dont use standalone convolution on fp by default

src/targets/gpu/fuse_mlir.cpp

@@ -150,8 +150,8 @@ fuse_input_ops_and_gemm_based_op(module_ref mm, instruction_ref gemm_based_op)
 enum class mlir_mode
 {
     all,
-    int8,
     fast,
+    int8,
     none
 };
 
@@ -192,7 +192,9 @@ auto is_mlir_conv(mlir_mode mode)
             return false;
         if(ins->get_shape().type() == shape::int8_type)
             return true;
-        if(mode != mlir_mode::fast)
+        if(mode == mlir_mode::int8)
+            return false;
+        if(mode == mlir_mode::all)
             return true;
         auto w = ins->inputs().at(1)->get_shape();
         if(w.lens().size() != 4)
@@ -416,6 +418,7 @@ void fuse_mlir::apply(module_pass_manager& mpm) const
     match::find_matches(mpm,
                         find_mlir_fused_ops{.conv_mode = get_mode("fused", mlir_mode::fast),
                                             .dot_mode  = get_mode("fused", mode)});
+
     match::find_matches(
         mpm,
         find_mlir_standalone_convolution_op{get_mode("convolution", mlir_mode::int8)},