Support fmod=1 in Mod OP (#921)

* fixed Gemm bug

* re-lint

* fixed typo error

* support fmod=1

tests/onnx/test_auto_scan_mod.py

@@ -42,7 +42,7 @@ class TestModConvert(OPConvertAutoScanTest):
             input_data[abs(input_data) < 1.0] = 1.0
             return input_data
 
-        input_dtype = draw(st.sampled_from(["int32", "int64"]))
+        input_dtype = draw(st.sampled_from(["float32", "int32", "int64"]))
 
         config = {
             "op_names": ["Mod"],

x2paddle/op_mapper/onnx2paddle/opset10.py

@@ -103,15 +103,52 @@ class OpSet10(OpSet9):
         # Support Mod op Since opset version >= 10
         val_x = self.graph.get_input_node(node, idx=0, copy=True)
         val_y = self.graph.get_input_node(node, idx=1, copy=True)
-        input_dtype = str(val_x.dtype)
-        assert "int" in input_dtype, 'Now only support int32 or int64 dtype'
         fmod = node.get_attr('fmod', 0)
-        assert fmod == 0, 'Now only support fmod == 0'
+        x_dtype = str(val_x.dtype)
+        y_dtype = str(val_y.dtype)
+        if "int" in x_dtype and "int" in y_dtype and fmod == 1:
+            fmod = 0
+        if fmod == 0:
             self.paddle_graph.add_layer(
                 'paddle.mod',
                 inputs={"x": val_x.name,
                         "y": val_y.name},
                 outputs=[node.name])
+        else:
+            # Step1:trunc_div(a, b) = sign(a / b) * floor(abs(a / b))
+            self.paddle_graph.add_layer(
+                'paddle.divide',
+                inputs={"x": val_x.name,
+                        "y": val_y.name},
+                outputs=[node.name + "_divide"])
+            self.paddle_graph.add_layer(
+                'paddle.sign',
+                inputs={"x": node.name + "_divide"},
+                outputs=[node.name + "_sign"])
+            self.paddle_graph.add_layer(
+                'paddle.abs',
+                inputs={"x": node.name + "_divide"},
+                outputs=[node.name + "_abs"])
+            self.paddle_graph.add_layer(
+                'paddle.floor',
+                inputs={"x": node.name + "_abs"},
+                outputs=[node.name + "_floor"])
+            self.paddle_graph.add_layer(
+                'paddle.multiply',
+                inputs={"x": node.name + "_sign",
+                        "y": node.name + "_floor"},
+                outputs=[node.name + "_trunc_div"])
+            # Step2:result = a - trunc_div(a, b) * b
+            self.paddle_graph.add_layer(
+                'paddle.multiply',
+                inputs={"x": node.name + "_trunc_div",
+                        "y": val_y.name},
+                outputs=[node.name + "_trunc_div"])
+            self.paddle_graph.add_layer(
+                'paddle.subtract',
+                inputs={"x": val_x.name,
+                        "y": node.name + "_trunc_div"},
+                outputs=[node.name])
 
     @print_mapping_info
     def IsInf(self, node):