cascade.repos#

Repo is a container for Lines.

class cascade.repos.LineFactory[source]#

classmethod create(path: str, *args: Any, line_type: Literal['model', 'data', None] = None, **kwargs: Any) → Line[source]#

classmethod read(path: str, *args: Any, **kwargs: Any) → Optional[Line][source]#

class cascade.repos.Repo(folder: str, *args: Any, overwrite: bool = False, meta_fmt: Literal['.json', '.yml', '.yaml'] = '.json', **kwargs: Any)[source]#

An interface to manage series of experiments called Lines. When created, initializes an empty folder constituting a repository of lines.

Example

>>> from cascade.repos import Repo
>>> from cascade.utils.baselines import ConstantBaseline
>>> repo = Repo("repo")
>>> repo.describe("This is a repo with one line for the example.")
>>> line = repo.add_line("const", model_cls=ConstantBaseline)
>>> model = line.create_model(constant=1)
>>> model.fit([0], [1])
>>> line.save(model)

__getitem__(key: Union[str, int])[source]#

Returns:: line – existing line of the name passed in key
Return type:: Line

__init__(folder: str, *args: Any, overwrite: bool = False, meta_fmt: Literal['.json', '.yml', '.yaml'] = '.json', **kwargs: Any) → None[source]#

Parameters:

folder – Path to a folder where Repo needs to be created or already was created if folder does not exist, creates it
overwrite (bool) – if True will remove folder that is passed in first argument and start a new repo in that place
meta_fmt (Literal['.json', '.yml', '.yaml']) – extension of repo’s metadata files and that will be assigned to the lines by default .json and .yml or .yaml are supported

__len__() → int[source]#

Returns:: num – a number of lines
Return type:: int

add_line(name: Optional[str] = None, line_type: Literal['data', 'model', None] = 'model', *args: Any, meta_fmt: Literal['.json', '.yml', '.yaml', None] = None, **kwargs: Any) → Line[source]#

Adds new line to repo if it doesn’t exist and returns it. If line exists, defines it in repo with parameters provided.

Supports all the parameters of Line using args and kwargs.

Parameters:

name (Optional[str], optional) – str, optional
meta_fmt (Literal[".json", ".yml", ".yaml", None], by default None) – str, optional
name – Name of the line. It is used to name a folder of line. Repo prepends it with self._root before creating. Optional argument. If omitted - names new line automatically using f’{len(self):0>5d}’, by default None
line_type (Literal["data", "model"]], by default model) – The type of model line to create, by default model
meta_fmt – Format of meta files. Supported values are the same as for repo. If omitted, inherits format from repo., by default None

Returns:

Created or re-created model line

Return type:

Line

Raises:

AssertionError – If name is None and line with the computed name already exists
FileNotFoundError – If line was not found even though Repo recorded it at creation time
ValueError – If provided arguments do not match the saved model’s configuration on disk

get_line_names() → List[str][source]#: Returns list of line names.

load_obj_meta(obj: str) → List[Dict[Any, Any]][source]#

Loads metadata of an object inside repo from disk

Parameters:: obj (str) – obj slug e.g. fair_squid_of_bliss
Returns:: Obj metadata
Return type:: Meta
Raises:: FileNotFoundError – Raises if failed to find the obj with slug specified

reload() → None[source]#: Updates internal state

class cascade.repos.SingleLineRepo(line: Line, *args: Any, **kwargs: Any)[source]#

Virtual Repo to represent a single line with Repo interface in code

__init__(line: Line, *args: Any, **kwargs: Any) → None[source]#

Parameters:

description – String description of an object
tags (Iterable[str], optional) – The list of tags to be added