1 系统环境

硬件环境(Ascend/GPU/CPU): CPU
MindSpore版本: 1.9
执行模式: 静态图
Python版本: 3.7.5
操作系统平台: linux

2 报错信息

2.1 问题描述

有一个模型，在construct里进行了一些计算，计算完毕后想删除某个tensor

2.2 报错信息

Traceback (most recent call last): File "D:/cons.py", line 11, in <module> m(1, 2) File "D:\python3.7\lib\site-packages\mindspore\nn\cell.py", line 622, in __call__ out = self.compile_and_run(*args) File "D:\python3.7\lib\site-packages\mindspore\nn\cell.py", line 1007, in compile_and_run self.compile(*inputs) File "D:\python3.7\lib\site-packages\mindspore\nn\cell.py", line 979, in compile jit_config_dict=self._jit_config_dict) File "D:\python3.7\lib\site-packages\mindspore\common\api.py", line 1147, in compile result = self._graph_executor.compile(obj, args_list, phase, self._use_vm_mode()) RuntimeError: Unsupported statement 'Delete'. More details please refer to syntax support at https://www.mindspore.cn ---------------------------------------------------- - The Traceback of Net Construct Code: ---------------------------------------------------- # In file D:/cons.py:9 del a ^ ---------------------------------------------------- - C++ Call Stack: (For framework developers) ---------------------------------------------------- mindspore\ccsrc\pipeline\jit\parse\parse.cc:772 mindspore::parse::Parser::ParseStatement Process finished with exit code 1

2.3 脚本代码

from mindspore import nn import mindspore as ms ms.set_context(mode=ms.GRAPH_MODE) class Model(nn.Cell): def construct(self, x, y): a = x + 1 y = a + y del a return y m = Model() m(1, 2)

3 根因分析

在Graph模式下，Python代码并不是由Python解释器去执行，而是将代码编译成静态计算图，然后执行静态计算图。
由于语法解析的限制，当前在编译构图时，支持的数据类型、语法以及相关操作并没有完全与Python语法保持一致，部分使用受限。
具体参考：静态图语法支持 | MindSpore 1.8 文档 | 昇思MindSpore社区
出错的代码
https://gitee.com/mindspore/mindspore/blob/master/mindspore/ccsrc/pipeline/jit/parse/parse.cc

790×277 12.9 KB

796×552 36 KB

从说明文档和代码上都能看出del不被支持

4 解决方案

不用del，将变量置None

from mindspore import nn import mindspore as ms ms.set_context(mode=ms.GRAPH_MODE) class Model(nn.Cell): def construct(self, x, y): a = x + 1 y = a + y a = None return y m = Model() m(1, 2)