Heroku部署失败：libtorch_hip.so及sndfile库缺失问题求助

问题概述 #

Django应用在Heroku平台构建成功，但部署失败。为控制slug体积（避免超过Heroku 500MB限制），使用torch==1.13.1+cpu而非完整GPU版本。

依赖配置（requirements.txt） #

-f https://download.pytorch.org/whl/torch_stable.html
Django==4.1.4
djangorestframework==3.14.0
requests==2.28.1
whisper==1.0
django-cors-headers==3.13.0
psycopg2==2.9.5
gunicorn==20.1.0
dj-database-url>=1.0,<2.0
whitenoise>=6.0,<7.0
torch==1.13.1+cpu
speechbrain
pyannote.audio
pyannote.core

初始部署错误栈追踪 #

核心错误提示 #

警告：无法导入soundfile，'soundfile'后端不可用
报错：OSError: libtorch_hip.so: 无法打开共享对象文件: 没有那个文件或目录

完整栈追踪代码 #

/app/.heroku/python/lib/python3.8/site-packages/torchaudio/_internal/module_utils.py:99: UserWarning: Failed to import soundfile. 'soundfile' backend is not available.
  warnings.warn("Failed to import soundfile. 'soundfile' backend is not available.")
Traceback (most recent call last):
  File "manage.py", line 22, in <module>
    main()
  File "manage.py", line 18, in main
    execute_from_command_line(sys.argv)
  File "/app/.heroku/python/lib/python3.8/site-packages/django/core/management/__init__.py", line 446, in execute_from_command_line
    utility.execute()
  File "/app/.heroku/python/lib/python3.8/site-packages/django/core/management/__init__.py", line 440, in execute
    self.fetch_command(subcommand).run_from_argv(self.argv)
  File "/app/.heroku/python/lib/python3.8/site-packages/django/core/management/base.py", line 402, in run_from_argv
    self.execute(*args, **cmd_options)
  File "/app/.heroku/python/lib/python3.8/site-packages/django/core/management/base.py", line 448, in execute
    output = self.handle(*args, **options)
  File "/app/.heroku/python/lib/python3.8/site-packages/django/core/management/base.py", line 96, in wrapped
    res = handle_func(*args, **kwargs)
  File "/app/.heroku/python/lib/python3.8/site-packages/django/core/management/commands/migrate.py", line 97, in handle
    self.check(databases=[database])
  File "/app/.heroku/python/lib/python3.8/site-packages/django/core/management/base.py", line 475, in check
    all_issues = checks.run_checks(
  File "/app/.heroku/python/lib/python3.8/site-packages/django/core/checks/registry.py", line 88, in run_checks
    new_errors = check(app_configs=app_configs, databases=databases)
  File "/app/.heroku/python/lib/python3.8/site-packages/django/core/checks/urls.py", line 14, in check_url_config
    return check_resolver(resolver)
  File "/app/.heroku/python/lib/python3.8/site-packages/django/core/checks/urls.py", line 24, in check_resolver
    return check_method()
  File "/app/.heroku/python/lib/python3.8/site-packages/django/urls/resolvers.py", line 494, in check
    for pattern in self.url_patterns:
  File "/app/.heroku/python/lib/python3.8/site-packages/django/utils/functional.py", line 57, in __get__
    res = instance.__dict__[self.name] = self.func(instance)
  File "/app/.heroku/python/lib/python3.8/site-packages/django/urls/resolvers.py", line 715, in url_patterns
    patterns = getattr(self.urlconf_module, "urlpatterns", self.urlconf_module)
  File "/app/.heroku/python/lib/python3.8/site-packages/django/utils/functional.py", line 57, in __get__
    res = instance.__dict__[self.name] = self.func(instance)
  File "/app/.heroku/python/lib/python3.8/site-packages/django/urls/resolvers.py", line 708, in urlconf_module
    return import_module(self.urlconf_name)
  File "/app/.heroku/python/lib/python3.8/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 1014, in _gcd_import
  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 "<frozen importlib._bootstrap_external>", line 843, in exec_module
  File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed
  File "/app/podcast_project_django/urls.py", line 4, in <module>
    from . import views
  File "/app/podcast_project_django/views.py", line 12, in <module>
    from pyannote.audio.pipelines.speaker_verification import PretrainedSpeakerEmbedding
  File "/app/.heroku/python/lib/python3.8/site-packages/pyannote/audio/__init__.py", line 29, in <module>
    from .core.inference import Inference
  File "/app/.heroku/python/lib/python3.8/site-packages/pyannote/audio/core/inference.py", line 34, in <module>
    from pyannote.audio.core.io import AudioFile
  File "/app/.heroku/python/lib/python3.8/site-packages/pyannote/audio/core/io.py", line 38, in <module>
    import torchaudio
  File "/app/.heroku/python/lib/python3.8/site-packages/torchaudio/__init__.py", line 1, in <module>
    from torchaudio import (  # noqa: F401
  File "/app/.heroku/python/lib/python3.8/site-packages/torchaudio/_extension.py", line 135, in <module>
    _init_extension()
  File "/app/.heroku/python/lib/python3.8/site-packages/torchaudio/_extension.py", line 105, in _init_extension
    _load_lib("libtorchaudio")
  File "/app/.heroku/python/lib/python3.8/site-packages/torchaudio/_extension.py", line 52, in _load_lib
    torch.ops.load_library(path)
  File "/app/.heroku/python/lib/python3.8/site-packages/torch/_ops.py", line 573, in load_library
    ctypes.CDLL(path)
  File "/app/.heroku/python/lib/python3.8/ctypes/__init__.py", line 373, in __init__
    self._handle = _dlopen(self._name, mode)
OSError: libtorch_hip.so: cannot open shared object file: No such file or directory

疑似关联代码 #

model = whisper.load_model("small.en")
embedding_model = PretrainedSpeakerEmbedding(
    "speechbrain/spkrec-ecapa-voxceleb",
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
)

添加torchaudio==0.13.1+cpu后的新错误 #

核心错误提示 #

OSError: sndfile库未找到

完整错误栈代码 #

File "/app/.heroku/python/lib/python3.8/site-packages/soundfile.py", line 142, in <module>
    raise OSError('sndfile library not found')
OSError: sndfile library not found

问题原因分析 #

1. 初始torchaudio加载错误：使用CPU版PyTorch时，torchaudio默认尝试加载GPU相关依赖库libtorch_hip.so，但Heroku环境无该库，导致加载失败。
2. sndfile库缺失错误：soundfile依赖系统级的sndfile库，Heroku默认Python buildpack未预装该库，因此触发报错。

解决建议 #

修复torchaudio加载错误 #

在requirements.txt中明确指定与PyTorch版本严格匹配的CPU版torchaudio，配置如下：

torchaudio==0.13.1+cpu

确保该配置放在torch==1.13.1+cpu之后，并保留PyTorch的索引链接-f https://download.pytorch.org/whl/torch_stable.html。

修复sndfile库缺失错误 #

使用Heroku的apt buildpack安装系统级依赖：

1. 在项目根目录创建Aptfile文件，内容为：

   libsndfile1
   

2. 执行命令添加apt buildpack到Heroku应用：

   heroku buildpacks:add --index 1 heroku-community/apt
   

3. 重新部署应用。

额外优化建议 #

• 检查slug体积：部署前用heroku slug:size查看体积，若接近500MB限制，可尝试：
  • 移除未使用的依赖包；
  • 替换为更小的Whisper模型（如base.en）；
  • 在requirements.txt中添加--no-cache-dir参数减少缓存体积（需通过pip install命令传递，可在Heroku的runtime.txt或Procfile中配置）。

内容的提问来源于stack exchange，提问作者Jonathan Stern