[Dy2St] Optimize while cond break error messages

[gouzil]

PR Category

Execute Infrastructure

PR Types

Improvements

Description

改进控制流 break 报错信息

新的报错信息

Traceback (most recent call last):
  File "/Users/gouzi/Documents/git/Paddle/test.py", line 19, in <module>
    print(paddle.jit.to_static(test_optim_break_in_for, full_graph=True)(np.random.random([10, 16]).astype('float32')))
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gouzi/Documents/git/Paddle/build/python/paddle/jit/dy2static/program_translator.py", line 535, in __call__
    return self._perform_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gouzi/Documents/git/Paddle/build/python/paddle/jit/dy2static/program_translator.py", line 855, in _perform_call
    error_data.raise_new_exception()
  File "/Users/gouzi/Documents/git/Paddle/build/python/paddle/jit/dy2static/error.py", line 454, in raise_new_exception
    raise new_exception from None
paddle.jit.dy2static.utils.Dygraph2StaticException: In transformed code:


    File "/Users/gouzi/.cache/paddle/to_static_tmp/37003/test_optim_break_in_for53qdai3x.py", line 51, in test_optim_break_in_for
	_jst.While(_jst.Ld(while_condition_0), _jst.Ld(while_body_0), _jst.Ld(get_args_0), _jst.Ld(set_args_0), return_name_ids=('_', '__break_0'), push_pop_names=None)
    File "/Users/gouzi/Documents/git/Paddle/build/python/paddle/jit/dy2static/convert_operators.py", line 190, in convert_while_loop
	_run_py_while(cond, body, getter, setter)
    File "/Users/gouzi/Documents/git/Paddle/build/python/paddle/jit/dy2static/convert_operators.py", line 274, in _run_py_while
	raise Dygraph2StaticException(textwrap.dedent(err_msg))

    paddle.jit.dy2static.utils.Dygraph2StaticException: You need to use break_flag in the control flow, Python will mistakenly convert the iteration variable from bool to Tensor,
And Tensor has no known value on static graph, for example:
    >>> x = paddle.to_tensor([5, 10])
    >>> for _ in range(10):
    ...     if x.sum() > 5:
    ...         break

There are one common workarounds available:
Explicitly rewrite to the form of a control flow, for example:
    >>> x = paddle.to_tensor([5, 10])
    >>> for _ in range(paddle.to_tensor(10)):
    ...     if x.sum() > 5:
    ...         break

[paddle-bot]

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[paddle-ci-bot]

Sorry to inform you that 328f3b0's CIs have passed for more than 7 days. To prevent PR conflicts, you need to re-run all CIs manually.

[SigureMo]

我咋看不懂呢？

• 首先说明问题
• 然后给出修改方法
• 最后给出常见修复示例

[paddle-ci-bot]

Sorry to inform you that 7885ce9's CIs have passed for more than 7 days. To prevent PR conflicts, you need to re-run all CIs manually.