pyinstaller打包错误：OSError: Can‘t get source for ＜function sigmoid_focal_loss at 0x00000174BE600F70＞.

前言

使用pyinstaller将detectron2项目打包成exe，打包成功后执行exe文件就出现了错误
System：Windows10
Pytorch：1.8.1
Detectron2：0.6
CUDA：11.1
pyinstaller：5.6.2

错误信息

Traceback (most recent call last):
  File "torch\_utils_internal.py", line 55, in get_source_lines_and_file
  File "inspect.py", line 979, in getsourcelines
  File "inspect.py", line 798, in findsource
OSError: could not get source code

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "main.py", line 35, in <module>
  File "<frozen importlib._bootstrap>", line 991, in _find_and_load
  File "<frozen importlib._bootstrap>", line 975, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 671, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 499, in exec_module
  File "my01\utils\my01_get_train_record.py", line 5, in <module>
  File "<frozen importlib._bootstrap>", line 991, in _find_and_load
  File "<frozen importlib._bootstrap>", line 975, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 671, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 499, in exec_module
  File "utils\util.py", line 16, in <module>
  File "<frozen importlib._bootstrap>", line 991, in _find_and_load
  File "<frozen importlib._bootstrap>", line 975, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 671, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 499, in exec_module
  File "mm01_p\mm01_p_main.py", line 17, in <module>
  File "<frozen importlib._bootstrap>", line 991, in _find_and_load
  File "<frozen importlib._bootstrap>", line 975, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 671, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 499, in exec_module
  File "mm01_p\tools\train_mm_p.py", line 8, in <module>
  File "<frozen importlib._bootstrap>", line 991, in _find_and_load
  File "<frozen importlib._bootstrap>", line 975, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 671, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 499, in exec_module
  File "detectron2\engine\__init__.py", line 11, in <module>
  File "<frozen importlib._bootstrap>", line 991, in _find_and_load
  File "<frozen importlib._bootstrap>", line 975, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 671, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 499, in exec_module
  File "detectron2\engine\hooks.py", line 19, in <module>
  File "<frozen importlib._bootstrap>", line 991, in _find_and_load
  File "<frozen importlib._bootstrap>", line 975, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 671, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 499, in exec_module
  File "fvcore\nn\__init__.py", line 4, in <module>
  File "<frozen importlib._bootstrap>", line 991, in _find_and_load
  File "<frozen importlib._bootstrap>", line 975, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 671, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 499, in exec_module
  File "fvcore\nn\focal_loss.py", line 52, in <module>
  File "torch\jit\_script.py", line 986, in script
  File "torch\jit\frontend.py", line 240, in get_jit_def
  File "torch\_utils_internal.py", line 62, in get_source_lines_and_file
OSError: Can't get source for <function sigmoid_focal_loss at 0x00000174BE600F70>. TorchScript requires source access in order to carry out compilation, make sure original .py files are available.

错误截图

解决办法

第一种方法

打包的时main.py，再import torch前加入如下代码，再重新打包

def script_method(fn, _rcb=None):
    return fn


def script(obj, optimize=True, _frames_up=0, _rcb=None):
    return obj


import torch.jit

script_method1 = torch.jit.script_method
script1 = torch.jit.script
torch.jit.script_method = script_method
torch.jit.script = script  # 设置之后导致mm01和mm01_p转TorchScript失败  TODO
import torch

第二种方法

在spec文件中添加如下代码

import torch  #add 
import inspect  #add 
block_cipher = None
from PyInstaller.utils.hooks import collect_data_files  #add 

a = Analysis(
    ['main.py'],
    pathex=[],
    binaries=[],
    datas=[*collect_data_files("torch", include_py_files=True), *collect_data_files("fvcore", include_py_files=True), *collect_data_files("detectron2", include_py_files=True)],  #modify
    hiddenimports=['sklearn.utils._typedefs', 'sklearn.utils._heap', 'sklearn.utils._sorting', 'sklearn.utils._vector_sentinel'],
    hookspath=[],
    hooksconfig={},
    runtime_hooks=[],
    excludes=[],
    win_no_prefer_redirects=False,
    win_private_assemblies=False,
    cipher=block_cipher,
    noarchive=False,
)

重新打包

pyinstaller mian.spec