diff --git a/src/core/impl/graph/cg_impl.cpp b/src/core/impl/graph/cg_impl.cpp
index db738e1e810d88200fe65f0fec2716ba824cd715..88b9f23b7178316a312ac7acb101b7ea9ab43efd 100644
--- a/src/core/impl/graph/cg_impl.cpp
+++ b/src/core/impl/graph/cg_impl.cpp
@@ -474,8 +474,7 @@ ComputingGraphImpl::CompileState ComputingGraphImpl::compile_prepare(
     }
 #endif
     gopt::GraphOptimizer optimizer;
-    optimizer.apply_optimize_options(options().graph_opt);
-    options().graph_opt.reset();
+    optimizer.add_passes_for_optimize_options(options().graph_opt, true);
     optimizer.apply_inplace(dest_vars);
 
     const OprNodeArray* opr_seq = nullptr;
diff --git a/src/core/include/megbrain/graph/cg.h b/src/core/include/megbrain/graph/cg.h
index 229d101f1d9ba81d8b79ff462ac1eaef01987d5c..0f5bddfd584b48771558a112376495bc4e66bd65 100644
--- a/src/core/include/megbrain/graph/cg.h
+++ b/src/core/include/megbrain/graph/cg.h
@@ -107,19 +107,17 @@ struct GraphCommonOptimizeOptions {
     };
     LayoutTransform layout_transform = LayoutTransform::DEFAULT;
 
-    void reset() {
-        f16_io_f32_comp = false;
-        f16_io_comp = false;
-        fuse_conv_bias_nonlinearity = false;
-        fuse_conv_bias_with_z = false;
-        layout_transform = LayoutTransform::DEFAULT;
-    }
+#define SET(n)                                  \
+    GraphCommonOptimizeOptions& enable_##n() {  \
+        n = true;                               \
+        return *this;                           \
+    }                                           \
+    GraphCommonOptimizeOptions& disable_##n() { \
+        n = false;                              \
+        return *this;                           \
+    }                                           \
+    bool has_set_##n() { return n == true; }
 
-#define SET(n)                                 \
-    GraphCommonOptimizeOptions& enable_##n() { \
-        n = true;                              \
-        return *this;                          \
-    }
     SET(f16_io_f32_comp);
     SET(f16_io_comp);
     SET(fuse_conv_bias_nonlinearity);
@@ -131,7 +129,11 @@ struct GraphCommonOptimizeOptions {
         layout_transform = LayoutTransform::_trans_capital;         \
         return *this;                                               \
     }                                                               \
-    bool transform_##_trans() const {                               \
+    GraphCommonOptimizeOptions& disable_##_trans() {                \
+        layout_transform = LayoutTransform::DEFAULT;                \
+        return *this;                                               \
+    }                                                               \
+    bool has_set_##_trans() const {                                 \
         return layout_transform == LayoutTransform::_trans_capital; \
     }
 
diff --git a/src/gopt/impl/framework.cpp b/src/gopt/impl/framework.cpp
index aa287d37220e4a4ae036bb147ad5acf09ac6a20a..8552c3be13310294c7ccb2f5f345b12926331ae1 100644
--- a/src/gopt/impl/framework.cpp
+++ b/src/gopt/impl/framework.cpp
@@ -675,7 +675,7 @@ GraphOptimizer& GraphOptimizer::add_preset_passes(
 
     if (inference_opt) {
         add_pass<ParamFusePass>();
-        apply_optimize_options(*inference_opt);
+        add_passes_for_optimize_options(*inference_opt);
     }
 
 
@@ -704,56 +704,56 @@ VarNode* GraphOptimizer::var_replace_lookup(VarNode *var) {
     }
 }
 
-const GraphOptimizer& GraphOptimizer::apply_optimize_options(
+
+const GraphOptimizer& GraphOptimizer::add_passes_for_optimize_options(
         const cg::GraphCommonOptimizeOptions& options) {
+    return add_passes_for_optimize_options(
+            const_cast<cg::GraphCommonOptimizeOptions&>(options));
+}
+
+const GraphOptimizer& GraphOptimizer::add_passes_for_optimize_options(
+        cg::GraphCommonOptimizeOptions& options, bool reset) {
     bool need_param_fuse = false;
-    if (options.f16_io_comp) {
-        add_pass(ConvertF32ToF16Pass::make(false));
-        need_param_fuse = true;
-    }
-    if (options.f16_io_f32_comp) {
-        add_pass(ConvertF32ToF16Pass::make(true));
-        need_param_fuse = true;
+
+#define cb(_option, _passes)             \
+    if (options.has_set_##_option()) {   \
+        _passes need_param_fuse = true;  \
+        if (reset) {                     \
+            options.disable_##_option(); \
+        }                                \
     }
-    if (options.transform_nhwcd4()) {
+    cb(f16_io_comp, { add_pass(ConvertF32ToF16Pass::make(false)); });
+    cb(f16_io_f32_comp, { add_pass(ConvertF32ToF16Pass::make(true)); });
+
+    cb(nhwcd4, {
         add_pass<FuseConvBiasNonlinPass>();
         add_pass(ConvertFormatPass::make_nhwcd4_converter());
-        need_param_fuse = true;
-    }
-    if (options.transform_nchw88()) {
-        add_pass(EnableNchwxxPass::make_nchwxx_converter(8));
-        need_param_fuse = true;
-    }
-    if (options.transform_nchw44()) {
-        add_pass(EnableNchwxxPass::make_nchwxx_converter(4));
-        need_param_fuse = true;
-    }
-    if (options.transform_nchw32()) {
+    });
+    cb(nchw88, { add_pass(EnableNchwxxPass::make_nchwxx_converter(8)); });
+    cb(nchw44, { add_pass(EnableNchwxxPass::make_nchwxx_converter(4)); });
+    cb(nchw32, {
         add_pass<FuseConvBiasNonlinPass>();
         add_pass<FuseConvBiasZPass>();
         add_pass(EnableTensorCorePass::make_tensorcore_converter());
         add_pass<ShuffleShuffleRemovePass>();
         add_pass<RemoveRedundantTypeCvtPass>();
-        need_param_fuse = true;
-    }
-    if (options.transform_chwn4()) {
+    });
+    cb(chwn4, {
         add_pass<FuseConvBiasNonlinPass>();
         add_pass<FuseConvBiasZPass>();
         add_pass(EnableCHWN4Pass::make_chwn4_converter());
         add_pass<ShuffleShuffleRemovePass>();
         add_pass<RemoveRedundantTypeCvtPass>();
-        need_param_fuse = true;
-    }
+    });
 
-    if (options.fuse_conv_bias_nonlinearity) {
-        add_pass<FuseConvBiasNonlinPass>();
-        need_param_fuse = true;
-    }
-    if (options.fuse_conv_bias_with_z) {
+    cb(fuse_conv_bias_nonlinearity, { add_pass<FuseConvBiasNonlinPass>(); });
+    cb(fuse_conv_bias_with_z, {
         add_pass<FuseConvBiasNonlinPass>();
         add_pass<FuseConvBiasZPass>();
-        need_param_fuse = true;
-    }
+    });
+
+#undef cb
+
     if (need_param_fuse) {
         add_pass<ParamFusePass>();
     }
diff --git a/src/gopt/include/megbrain/gopt/framework.h b/src/gopt/include/megbrain/gopt/framework.h
index f5f41cdb31c2e3f8a2b789037e8e15518ee61045..56d79f4c62ab6e70059d30e6bc3cdad35d4d58d0 100644
--- a/src/gopt/include/megbrain/gopt/framework.h
+++ b/src/gopt/include/megbrain/gopt/framework.h
@@ -468,9 +468,16 @@ namespace gopt {
             static VarNode* var_replace_lookup(VarNode *var);
 
             /**
-             * \brief apply optimize options
+             * \brief add pass indicated by optimize options.
+             *
+             * \param options common options
+             * \param reset if set true, it will reset options when add passes.
              */
-            const GraphOptimizer& apply_optimize_options(
+            const GraphOptimizer& add_passes_for_optimize_options(
+                    cg::GraphCommonOptimizeOptions& options,
+                    bool reset = false);
+
+            const GraphOptimizer& add_passes_for_optimize_options(
                     const cg::GraphCommonOptimizeOptions& options);
     };