从Colab将Optuna研究保存到Google Drive时出错

Question

我可以将一个随机文件保存到我的驱动器colab中，如下所示：

with open ("gdrive/My Drive/chapter_classification/output/hello.txt",'w')as f:
f.write('hello')

工作正常，但当我使用Optuna的官方文档方法时，使用代码：

direction = 'minimize'
name = 'opt1'
study = optuna.create_study(sampler=optuna.samplers.TPESampler(),direction=direction,study_name=name, storage=f"gdrive/My Drive/chapter_classification/output/sqlite:///{name}.db",load_if_exists=True)
study.optimize(tune, n_trials=1000)

抛出错误为：

ArgumentError                             Traceback (most recent call last)
<ipython-input-177-f32da2c0f69a> in <module>()
2 direction = 'minimize'
3 name = 'opt1'
----> 4 study = optuna.create_study(sampler=optuna.samplers.TPESampler(),direction=direction,study_name=name, storage="gdrive/My Drive/chapter_classification/output/sqlite:///opt1.db",load_if_exists=True)
5 study.optimize(tune, n_trials=1000)
6 frames
/usr/local/lib/python3.7/dist-packages/optuna/study/study.py in create_study(storage, sampler, pruner, study_name, direction, load_if_exists, directions)
1134     ]
1135 
-> 1136     storage = storages.get_storage(storage)
1137     try:
1138         study_id = storage.create_new_study(study_name)
/usr/local/lib/python3.7/dist-packages/optuna/storages/__init__.py in get_storage(storage)
29             return RedisStorage(storage)
30         else:
---> 31             return _CachedStorage(RDBStorage(storage))
32     elif isinstance(storage, RDBStorage):
33         return _CachedStorage(storage)
/usr/local/lib/python3.7/dist-packages/optuna/storages/_rdb/storage.py in __init__(self, url, engine_kwargs, skip_compatibility_check, heartbeat_interval, grace_period, failed_trial_callback)
173 
174         try:
--> 175             self.engine = create_engine(self.url, **self.engine_kwargs)
176         except ImportError as e:
177             raise ImportError(
<string> in create_engine(url, **kwargs)
/usr/local/lib/python3.7/dist-packages/sqlalchemy/util/deprecations.py in warned(fn, *args, **kwargs)
307                         stacklevel=3,
308                     )
--> 309             return fn(*args, **kwargs)
310 
311         doc = fn.__doc__ is not None and fn.__doc__ or ""
/usr/local/lib/python3.7/dist-packages/sqlalchemy/engine/create.py in create_engine(url, **kwargs)
528 
529     # create url.URL object
--> 530     u = _url.make_url(url)
531 
532     u, plugins, kwargs = u._instantiate_plugins(kwargs)
/usr/local/lib/python3.7/dist-packages/sqlalchemy/engine/url.py in make_url(name_or_url)
713 
714     if isinstance(name_or_url, util.string_types):
--> 715         return _parse_rfc1738_args(name_or_url)
716     else:
717         return name_or_url
/usr/local/lib/python3.7/dist-packages/sqlalchemy/engine/url.py in _parse_rfc1738_args(name)
775     else:
776         raise exc.ArgumentError(
--> 777             "Could not parse rfc1738 URL from string '%s'" % name
778         )
779 
ArgumentError: Could not parse rfc1738 URL from string 'gdrive/My Drive/chapter_classification/output/sqlite:///opt1.db'

Answer 1

因此，根据create_study的官方文档

当传递数据库URL时，Optuna内部使用SQLAlchemy来处理数据库。有关更多详细信息，请参阅SQLAlchemy的文档。如果要为SQLAlchemy引擎指定非默认选项，可以使用所需选项实例化RDBStorage，并将其传递给存储参数，而不是URL。

当您访问SQLAlchemy的文档时，您会发现它使用绝对路径。

所以你所要做的就是改变

storage=f"gdrive/My Drive/chapter_classification/output/sqlite:///{name}.db"

到绝对路径为：

storage = f"sqlite:///gdrive/My Drive/chapter_classification/output{name}.db"