Correctly handle int values for assign_value_op

python/paddle/v2/fluid/layers/tensor.py

@@ -148,10 +148,15 @@ def assign(input, output):
         dtype = convert_np_dtype_to_dtype_(input.dtype)
         if dtype == DataType.FP32:
             value_name = "fp32_values"
+            values = [float(v) for v in input.flat]
         elif dtype == DataType.INT32:
             value_name = "int32_values"
+            values = [int(v) for v in input.flat]
         else:
             raise ValueError("Unsupported dtype %s", input.dtype)
+        if input.size > 1024 * 1024:
+            raise ValueError("The size of input is too big. Please consider "
+                             "saving it to file and 'load_op' to load it")
 
         helper.append_op(
             type='assign_value',
@@ -159,7 +164,7 @@ def assign(input, output):
             attrs={
                 'dtype': dtype,
                 'shape': list(input.shape),
-                value_name: [float(v) for v in input.flat]
+                value_name: values
             })
     else:
         raise ValueError("Wrong type for assign input: %s" % type(input))

python/paddle/v2/fluid/tests/test_assign_value_op.py

@@ -22,16 +22,18 @@ class TestAssignValueOp(op_test.OpTest):
         self.check_output()
 
     def test_assign(self):
-        val = numpy.random.random(size=(2, 5)).astype(numpy.float32)
+        val = (
+            -100 + 200 * numpy.random.random(size=(2, 5))).astype(numpy.int32)
         x = layers.create_tensor(dtype="float32")
         layers.assign(input=val, output=x)
         exe = fluid.Executor(fluid.CPUPlace())
         fetched_x = exe.run(fluid.default_main_program(),
                             feed={},
-                            fetch_list=[x])
+                            fetch_list=[x])[0]
         self.assertTrue(
-            numpy.allclose(fetched_x, val),
+            numpy.array_equal(fetched_x, val),
             "fetch_x=%s val=%s" % (fetched_x, val))
+        self.assertEqual(fetched_x.dtype, val.dtype)
 
 
 if __name__ == '__main__':