我正在尝试为Python设置sacred
,我正在通过tutorial。我能够使用pip install sacred
设置神圣而没有任何问题。我在运行基本代码时遇到问题:
from sacred import Experiment
ex = Experiment("hello_world")
运行此代码将返回a ValueError
:
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-25-66f549cfb192> in <module>()
1 from sacred import Experiment
2
----> 3 ex = Experiment("hello_world")
/Users/ryandevera/anaconda/lib/python2.7/site-packages/sacred/experiment.pyc in __init__(self, name, ingredients)
42 super(Experiment, self).__init__(path=name,
43 ingredients=ingredients,
---> 44 _caller_globals=caller_globals)
45 self.default_command = ""
46 self.command(print_config, unobserved=True)
/Users/ryandevera/anaconda/lib/python2.7/site-packages/sacred/ingredient.pyc in __init__(self, path, ingredients, _caller_globals)
48 self.doc = _caller_globals.get('__doc__', "")
49 self.sources, self.dependencies = \
---> 50 gather_sources_and_dependencies(_caller_globals)
51
52 # =========================== Decorators ==================================
/Users/ryandevera/anaconda/lib/python2.7/site-packages/sacred/dependencies.pyc in gather_sources_and_dependencies(globs)
204 def gather_sources_and_dependencies(globs):
205 dependencies = set()
--> 206 main = Source.create(globs.get('__file__'))
207 sources = {main}
208 experiment_path = os.path.dirname(main.filename)
/Users/ryandevera/anaconda/lib/python2.7/site-packages/sacred/dependencies.pyc in create(filename)
61 if not filename or not os.path.exists(filename):
62 raise ValueError('invalid filename or file not found "{}"'
---> 63 .format(filename))
64
65 mainfile = get_py_file_if_possible(os.path.abspath(filename))
ValueError: invalid filename or file not found "None"
我不确定为什么会出现这个错误。在运行代码之前,文档没有说明如何设置实验文件。任何帮助将不胜感激!
给出的回溯指示Experiment
的构造函数搜索其命名空间以查找其定义的文件。
因此,要使示例正常工作,请将示例代码放入文件中并直接运行该文件。
如果您正在使用ipython
,那么您可以尝试使用%%python
命令,该命令将在运行之前有效地捕获您将其提供到文件中的代码(在单独的python进程中)。
根据文档,如果您在IPython / Jupyter中,您可以允许实验在不可重现的交互式环境中运行:
ex = Experiment('jupyter_ex', interactive=True)
https://sacred.readthedocs.io/en/latest/experiment.html#run-the-experiment
The docs很好地说(TL; DR:神圣地为你检查这个并且为了警告你失败了)
Warning
默认情况下,如果在REPL或Jupyter Notebook等交互式环境中运行,则Sacred实验将失败。这是一种预期的安全措施,因为在这些环境中无法确保再现性。如果需要,可以通过将
interactive=True
传递给实验来停用此安全措施,如下所示:ex = Experiment('jupyter_ex', interactive=True)