Adapt MaxPool3DGrad

mindspore/ops/_op_impl/tbe/max_pool3d_grad.py

@@ -26,7 +26,8 @@ max_pool3d_grad_op_info = TBERegOp("MaxPool3DGrad") \
     .partial_flag(True) \
     .attr("kernel_size", "required", "listInt", "all") \
     .attr("strides", "required", "listInt", "all") \
-    .attr("pad_list", "required", "listInt", "all") \
+    .attr("pad_mode", "optional", "str", "all") \
+    .attr("pad_list", "required", "listInt", "all", "0,0,0") \
     .attr("format", "optional", "str", "all") \
     .input(0, "orig_x", False, "required", "all") \
     .input(1, "orig_y", False, "required", "all") \

mindspore/ops/operations/_grad_ops.py

@@ -977,10 +977,6 @@ class MaxPool3DGrad(PrimitiveWithInfer):
 
     def infer_shape(self, x_shape, y_shape, grad_shape):
         validator.check_equal_int(len(x_shape), 5, "x rank", self.name)
-        pad_list = _get_max_pool3d_grad_pads_by_pad_mode(x_shape, self.kernel_size, self.strides, self.pad_mode)
-        for pad in pad_list:
-            validator.check_non_negative_int(pad, 'element of pad_list', self.name)
-        self.add_prim_attr("pad_list", pad_list)
         return x_shape
 
     def infer_dtype(self, x_dtype, y_dtype, grad_dtype):