diff --git a/x2paddle/decoder/tf_decoder.py b/x2paddle/decoder/tf_decoder.py
index dc04172ac9e3fbc659cf5ce05ec297a1fc4764ab..f23fc32b2ed812e4a3adf9a5c506a8137f7f4cd0 100644
--- a/x2paddle/decoder/tf_decoder.py
+++ b/x2paddle/decoder/tf_decoder.py
@@ -60,7 +60,7 @@ class TFGraphNode(GraphNode):
 
     @property
     def dtype(self):
-        keys = ['dtype', 'Tidx', 'T', 'DstT']
+        keys = ['dtype', 'T', 'DstT']
         for k in keys:
             dtype = self.layer.attr[k].type
             if dtype > 0:
diff --git a/x2paddle/op_mapper/tf_op_mapper.py b/x2paddle/op_mapper/tf_op_mapper.py
index c15ccb0790e4ad58a46000ca7fb91b71a992dc84..541ee3c845d78df5e0ec20763b6e9b7d92b6791e 100644
--- a/x2paddle/op_mapper/tf_op_mapper.py
+++ b/x2paddle/op_mapper/tf_op_mapper.py
@@ -744,7 +744,7 @@ class TFOpMapper(OpMapper):
 
         input_names = [i.name for i in inputs]
         for i, ipt in enumerate(inputs):
-            if node.dtype == 'bool':
+            if ipt.dtype == 'bool':
                 cast_name = gen_name('concat', 'cast')
                 program.add_layer(
                     kernel="fluid.layers.cast",
@@ -1213,9 +1213,17 @@ class TFOpMapper(OpMapper):
         attr["dim"] = reduce_idx.value.tolist()
         attr["keep_dim"] = node.get_attr("keep_dims")
 
+        input_name = input.name
+        if input.dtype != "bool":
+            input_name = gen_name("all", "cast")
+            program.add_layer(
+                "fluid.layers.cast",
+                inputs={"x": input.name},
+                outputs=[input_name],
+                dtype=string("bool"))
         program.add_layer(
             "fluid.layers.reduce_all",
-            inputs={"input": input.name},
+            inputs={"input": input_name},
             outputs=[node.name],
             **attr)
 
diff --git a/x2paddle/optimizer/tensorflow/batch_norm.py b/x2paddle/optimizer/tensorflow/batch_norm.py
index 315e94968da1c5094f9245173cfb53238a2f8924..3e3d81360598c2b66a00735cd23059998f3bef2d 100644
--- a/x2paddle/optimizer/tensorflow/batch_norm.py
+++ b/x2paddle/optimizer/tensorflow/batch_norm.py
@@ -8,6 +8,7 @@ class BatchNormOpt:
         pass
 
     def run(self, graph):
+        print("Optimize: BatchNormOpt...")
         layers = copy.deepcopy(graph.layers)
         for layer_id, layer in layers.items():
             if layer.kernel != "fluid.layers.elementwise_add":
diff --git a/x2paddle/optimizer/tensorflow/bias.py b/x2paddle/optimizer/tensorflow/bias.py
index ced691a700afb4213351cdd6095cf8822602795f..593095127b5729f3ef6aae52af41341132b9e73b 100644
--- a/x2paddle/optimizer/tensorflow/bias.py
+++ b/x2paddle/optimizer/tensorflow/bias.py
@@ -13,6 +13,7 @@ class BiasOpt:
         ]
 
     def run(self, graph):
+        print("Optimize: BiasOpt...")
         layers = copy.deepcopy(graph.layers)
         for layer_id, layer in layers.items():
             if layer.kernel in self.conv_layers or layer.kernel == "fluid.layers.transpose":
diff --git a/x2paddle/optimizer/tensorflow/transpose.py b/x2paddle/optimizer/tensorflow/transpose.py
index 81bd073444970eaac0e4a72535b68d4a1290dbdf..fbd10ab230d81e3463a74c64a7eda7a9fba6d93d 100644
--- a/x2paddle/optimizer/tensorflow/transpose.py
+++ b/x2paddle/optimizer/tensorflow/transpose.py
@@ -36,6 +36,7 @@ class TransposeOpt:
         return count
 
     def run(self, graph):
+        print("Optimize: TransposeOpt...")
         total_layer_num = len(graph.layers)
         scanned_layers = set()
         optimized_transpose_layers = list()