Automate experiments and explore your data.
autodora is a framework to help you:
- setup experiments
- running them for multiple parameters
- storing the results
- exploring the results
The aim of this package is to make these steps as easy and integrated as possible.
pip install autodora
Experiments can be tracked using observers. Specialized observers may require optional packages to function that are not included by default (because you might not need them).
pip install autodora[telegram]
In order to use the observer you have to set the environment variables
export TELEGRAM_BOT_TOKEN="<your-bot-token>" export TELEGRAM_CHAT_ID="<your-chat-id>" pytest
Consider the problem of computing the product of two numbers given in a string like this:
"0.1 x 0.3".
The heavy-lifting of this computation is performed by a function
def multiply(x, y): return x * y
We start off by describing the experiment in a file called
class ProductExperiment(Experiment): input = Parameter(str, "0.0x0.0", "The input values to be multiplied (e.g. 0.2x10)") product = Result(float, description="Computed product") @derived(cache=True) def derived_x(self): return float(self.get(self.input).split("x")) @derived(cache=True) def derived_y(self): return float(self.get(self.input).split("x")) def run_internal(self): x, y = self.get("x"), self.get("y") result = multiply(x, y) self["product"] = result ProductExperiment.enable_cli()
... input = Parameter(str, "0.0x0.0", "The input values to be multiplied (e.g. 0.2x10)") ...
The first step is to describe the parameters of the experiment, the name is taken from the variable you assign them to, other than that have to specify the type and optionally a default value and description of the parameter.
While this is a powerful and easy way to set up parameters, you can also add them in the constructor:
class ProductExperiment(Experiment): def __init__(self, group, storage=None, identifier=None): super().__init__(group, storage=None, identifier=None) self.parameters.add_parameter("complicated.name", datetime, None, "Description")
... product = Result(float, description="Computed product") ...
Similar to the parameters, we specify expected results. The Result class is identical to the Parameter class in all but name, it only serves to indicate that you are trying to assign a result.
Computing derived features
... @derived(cache=True) def derived_x(self): return float(self.get(self.input).split("x")) @derived(cache=True) def derived_y(self): return float(self.get(self.input).split("x")) ...
Derived features are computed from other values (or complex computation chains) and can be marked for caching to avoid computing them over and over again: when the experiment is saved to storage, those features will be saved with the experiment.
You can build derived features using the derived decorator, which internally builds a
Derived object and saves it in
experiment.derived_callbacks (Dict[str, Derived]) dictionary. Again, you can do this in the constructor, too.
If the decorated function is called
derived_<name>, it will be shortened to just
Derived features can be accessed by calling
experiment[""derived.name"] to disambiguate
if there are other parameters or results with the same name.
Running the experiment
... def run_internal(self): x, y = self.get("x"), self.get("y") result = multiply(x, y) self["product"] = result ...
experiment.run() is called, it internally calls the
run_internal method, which is responsible for running the
actual experiment. In this case, it fetches the (derived) parameters, computes the result and stores it.
Enabling the command line interface
enable_cli class method, not surprisingly, enables the current file to be run from command line.
This enables several key features:
- Making the experiment executable by command line (for internal and external use)
- Allowing you to manage (plot, list, ...) experiments of this type from command line
Release history Release notifications
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size autodora-0.3.29-py2.py3-none-any.whl (21.3 kB)||File type Wheel||Python version py2.py3||Upload date||Hashes View hashes|
|Filename, size autodora-0.3.29.tar.gz (19.8 kB)||File type Source||Python version None||Upload date||Hashes View hashes|
Hashes for autodora-0.3.29-py2.py3-none-any.whl