Enhance Check mechanism and Support single tuple/list of fetch_list in Executor

[Aurelius84]

PR types

Bug fixes

PR changes

APIs

Describe

Enhance Check mechanism of fetch_list in Executor

在下面的case中，如果传入的fetch_list值为[(var1, var2, var3)]，且var1.shape[0]=-1时，执行会进入无限循环中。

此PR修复了此问题，对fetch_list接受的参数类型兼容性扩展支持：

• 支持fetch_list=[tuple]或者fetch_list=[list]，但tuple和list必须是单层的
• 在_split_optimize_ops_in_fetch_list添加必要的类型检查，避免进入死循环

复现代码：

import paddle
import numpy as np
from inspect import isfunction

paddle.enable_static()
paddle.set_device('cpu')

main_program = paddle.static.Program()
start_program = paddle.static.Program()
main_program.random_seed = 222
start_program.random_seed = 222

feed = {
    "x":  np.array([[0], [0], [1], [0]], dtype='float32')
}


with paddle.utils.unique_name.guard():
    with paddle.static.program_guard(main_program=main_program, startup_program=start_program):
        x = paddle.static.data(name='x', shape=[4, 1], dtype='float32')
        output = paddle.unique_consecutive(x, return_inverse=True, return_counts=True, axis=0)
        exe = paddle.static.Executor(paddle.CPUPlace())
        exe.run(start_program)
        res = exe.run(
            main_program,
            feed=feed,
            fetch_list=[output],
            return_numpy=True
        )
        print(res)

[paddle-bot-old]

Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.