Regarding building an executable file with PyInstaller that includes the ‘torch’ and ‘ultralytics’ libraries

[hehehe1234567894]

I am working on a GUI project and have imported the following libraries:

import tkinter as tk
from PIL import Image, ImageTk, ImageDraw
from ultralytics import YOLO
import numpy as np
import torch
import os
import subprocess
import json
from tkinter import filedialog, messagebox
import wenxin_AI_zh
import word_creater

Among them, 'wenxin_AI_zh' and 'word_creater' are my own Python code.
They reference these libraries.

import re
import os
import qianfan
import time

and

from docx import Document
from docx.shared import Inches
from PIL import Image
import os

The spec file looks like this:

# -*- mode: python ; coding: utf-8 -*-

a = Analysis(
    ['main.py'],
    pathex=[],
    binaries=[],
    datas=[('settings.json', '.'), ('PDF', 'PDF'), ('picword', 'picword'), ('GHzhg.pt', '.')],
    hiddenimports=['wenxin_AI_zh',
                   'word_creater',
                   'tensorboard',
                   'torch.utils.tensorboard',
                   'numpy',
                   'onnxscript',
                   'qianfan',
                   'time',
                   'docx'
                   're'
                   'docx.shared',
                   'os'],
    hookspath=[],
    hooksconfig={},
    runtime_hooks=[],
    excludes=[],
    noarchive=False,
    optimize=0,
)

pyz = PYZ(a.pure)

exe = EXE(
    pyz,
    a.scripts,
    [],
    exclude_binaries=True,
    name='GoneHealthSystem',
    debug=True,
    bootloader_ignore_signals=False,
    strip=False,
    upx=True,
    console=True,  # 修改为 False 以生成 GUI 应用程序
    disable_windowed_traceback=False,
    argv_emulation=False,
    target_arch=None,
    codesign_identity=None,
    entitlements_file=None,
)

coll = COLLECT(
    exe,
    a.binaries,
    a.datas,
    strip=False,
    upx=True,
    upx_exclude=[],
    name='GoneHealthSystem',
)

However, the executable file exits on its own after starting in the Windows console, and no GUI appears.
There are the following errors during the build:

13192 WARNING: Conda distribution 'numpy', dependency of 'numpy', was not found. If you installed this distribution with pip then you may ignore this warning.
1232109 WARNING: Ignoring /usr/lib64/libgomp.so.1 imported from D:\Users\QIQI\anaconda3\envs\pyai\lib\site-packages\torch\_inductor\codecache.py - only basenames are supported with ctypes imports!
675232 WARNING: Hidden import "scipy.special._cdflib" not found! #I do not have this explicit import option
658851 ERROR: Hidden import 'torch.utilsnumpy' not found#I do not have this explicit import option
202181 WARNING: Failed to collect submodules for 'torch._inductor.kernel' because importing 'torch._inductor.kernel' raised: ImportError: cannot import name '_is_static_problem' from partially initialized module 'torch._inductor.kernel.mm' (most likely due to a circular import) (D:\Users\QIQI\anaconda3\envs\pyai\lib\site-packages\torch\_inductor\kernel\mm.py)

The PyInstaller log mainly shows errors related to torch.
With ‘debug=True’ and ‘console=True’, there are no error messages output. The main program exits mysteriously when run, and the situation is unclear.

[rokm]

With ‘debug=True’ and ‘console=True’, there are no error messages output. The main program exits mysteriously when run, and the situation is unclear.

Then you'll need to either instrument your code with print() statements to see where it stops being executed, or build with equivalent of --debug imports to see what the last module import was before the execution stops. I.e., replace

exe = EXE(
    pyz,
    a.scripts,
    [],
    ...

with

exe = EXE(
    pyz,
    a.scripts,
    [('v', None, 'OPTION')],
    ...

The build-time errors you've shown are mostly due to modules being renamed between different package versions, and are unlikely to be related to your run-time problem.