Skip to content

Recipe Creators

To add or modify a recipe's ingredients, you need to get a Pantry of ingredients. This can be done easily by building the Pantry:

leip_recipe_designer.Pantry.build staticmethod 

build(pantry_store_dir: Union[str, Path], packages: Optional[Union[str, List[str]]] = None, force_rebuild: bool = False) -> IngredientCache

Builds a pantry of ingredients for each of the specified packages.

Parameters:

  • pantry_store_dir (Union[str, Path]) –

    The path to store the pantry of ingredients.

  • packages (Optional[Union[str, List[str]]], default: None ) –

    The names of the packages to which you want to build pantries for. If packages are not specified, it detects the precense of compatible packages in your environment and builds the pantry for each.

  • force_rebuild (bool, default: False ) –

    If true, it will rebuild even if the pantry already exists.

Returns:

  • pantry ( IngredientCache ) –

    A pantry consisting of ingredients from all the packages specified or found in the environment.

Examples:

from leip_recipe_designer import Pantry
pantry = Pantry.build("./local_pantry")

# build a pantry for the Application Framework only
from leip_recipe_designer import Pantry
pantry = Pantry.build("./local_pantry", packages=["leip_af"])

Load a Saved Recipe

Info

You can optionally pass a pantry to load the recipes in "editable" mode. Not passing a pantry will still allow you to execute tasks on these recipes, but you wont be able to make any modifications.

leip_recipe_designer.create.from_recipe_id 

from_recipe_id(recipe_id: str, pantry: Optional[IngredientCache] = None, allow_upgrade: bool = False, allow_deprecated: bool = False, task: str = 'vision.detection.2d', volume: str = 'xval_det', verbose: int = 0, default: bool = False) -> RecipeNode

Reconstruct a Golden Volume Recipe from its ID.

This function allows reconstructing a recipe from the given ID within a specified Golden Volume. The recipe can optionally be upgraded, use deprecated ingredients, or default to specific task configurations.

Parameters:

  • recipe_id (str) –

    ID of the recipe in any of the Golden Volume dataframes.

  • pantry (Optional[IngredientCache], default: None ) –

    The ingredients storage to use during the recipe creation. If not provided, a limited pantry will be recreated from the serialized information. Note: this limited pantry includes only ingredients that are assigned to the recipe when pickle.dump() was called.

  • allow_upgrade (bool, default: False ) –

    This applies only if the pantry was provided. If True, the removed and updated by major change ingredients are translated to their latest variants.

  • allow_deprecated (bool, default: False ) –

    This applies only if the pantry was provided. If True, the deprecated ingredients are used instead of their latest variants.

  • task (str, default: 'vision.detection.2d' ) –

    The task associated with the recipe. Supported options: - "vision.detection.2d" (default) - "vision.classification.2d"

  • volume (str, default: 'xval_det' ) –

    Specify the golden volume in which to look up the recipe ID. - For detection tasks: Defaults to xval_det. Other options include bdd100k, chemistrylab, wheat, etc. - For classification tasks: Use xval_cls.

  • verbose (int, default: 0 ) –

    An integer specifying the verbosity level. Verbose Levels: 0: No output. 1: Shows the summary of the upgraded ingredients. 2: Additionally shows detailed information about rules selection and translation process.

  • default (bool, default: False ) –

    If True, applies default task-related settings for the recipe.

Returns:

  • recipe ( RecipeNode ) –

    The reconstructed recipe instance.

Examples:

For detection tasks (default volume: xval_det): 

from leip_recipe_designer import Pantry, create
pantry = Pantry.build("./local_pantry")
recipe = create.from_recipe_id(
    recipe_id="59845",
    pantry=pantry,
    allow_upgrade=True
)

For classification tasks: 

from leip_recipe_designer import Pantry, create
pantry = Pantry.build("./local_pantry")
recipe = create.from_recipe_id(
    recipe_id="4163",
    pantry=pantry,
    task="vision.classification.2d",
    volume="xval_cls",
    default=True
)

Notes
  • Detection workflows default to xval_det unless otherwise specified.
  • For classification workflows, explicitly use xval_cls and set task="vision.classification.2d".

leip_recipe_designer.create.from_sppr 

from_sppr(serialized: str, pantry: Optional[IngredientCache] = None, allow_upgrade: bool = False, allow_deprecated: bool = False, verbose: int = 0) -> RecipeNode

Reconstruct a recipe from its serialized representation.

Parameters:

  • serialized (str) –

    A serialized representation of the recipe. For more details about serialization process, refer to _serialize_portable.

  • pantry (Optional[IngredientCache], default: None ) –

    The ingredients storage to use during the recipe creation. If not provided, the limited pantry will be recreated from the serialized information. Note: this limited pantry includes only ingredients that are assigned to the recipe when to_sppr() was called.

  • allow_upgrade (bool, default: False ) –

    This applies only if the pantry was provided. If True, the removed and updated by major change ingredients are translated to their latest variants.

  • allow_deprecated (bool, default: False ) –

    This applies only if the pantry was provided. If True, the deprecated ingredients are used instead of their latest variants.

  • verbose (int, default: 0 ) –

    An integer specifying the verbosity level. Verbose Levels: 0: No output. 1: Shows the summary of the upgraded ingredients. 2: Additionally shows detailed information about rules selection and translation process.

Returns:

  • recipe ( RecipeNode ) –

    The reconstructed recipe instance.

Examples:

This is how a user would save and load a recipe using this method:

with open("my_recipe.sppr", "w") as file:
    file.write(recipe.sppr())

with open("my_recipe.sppr", "r") as file:
    recovered_recipe = from_sppr(file.read())

leip_recipe_designer.create.from_pickle 

from_pickle(path: Union[str, Path], pantry: Optional[IngredientCache] = None, allow_upgrade: bool = False, allow_deprecated: bool = False, verbose: int = 0) -> RecipeNode

Reconstruct a recipe from its a pickle representation.

Parameters:

  • path (Union[str, Path]) –

    A path to a pickled recipe object.

  • pantry (Optional[IngredientCache], default: None ) –

    The ingredients storage to use during the recipe creation. If not provided, the limited pantry will be recreated from the serialized information. Note: this limited pantry includes only ingredients that are assigned to the recipe when pickle.dump() was called.

  • allow_upgrade (bool, default: False ) –

    This applies only if the pantry was provided. If True, the removed and updated by major change ingredients are translated to their latest variants.

  • allow_deprecated (bool, default: False ) –

    This applies only if the pantry was provided. If True, the deprecated ingredients are used instead of their latest variants.

  • verbose (int, default: 0 ) –

    An integer specifying the verbosity level. Verbose Levels: 0: No output. 1: Shows the summary of the upgraded ingredients. 2: Additionally shows detailed information about rules selection and translation process.

Returns:

  • recipe ( RecipeNode ) –

    The reconstructed recipe instance.

Examples:

recipe.to_pickle("recipe.sppr")
recovered_recipe = rd.create.from_pickle("recipe.sppr")

Initialize an Empty Recipe

leip_recipe_designer.create.empty_detection_recipe 

empty_detection_recipe(pantry: IngredientCache) -> RecipeNode

Instantiate an empty recipe for a detection task.

Parameters:

  • pantry (IngredientCache) –

    The ingredients storage to associate with the new recipe.

Returns:

  • recipe ( RecipeNode ) –

    A blank recipe to populate with detection ingredients.

leip_recipe_designer.create.empty_classification_recipe 

empty_classification_recipe(pantry: IngredientCache) -> RecipeNode

Instantiate an empty recipe for a classification task.

Parameters:

  • pantry (IngredientCache) –

    The ingredients storage to associate with the new recipe.

Returns:

  • recipe ( RecipeNode ) –

    A blank recipe to populate with classification ingredients.

leip_recipe_designer.create.empty_segmentation_recipe 

empty_segmentation_recipe(pantry: IngredientCache) -> RecipeNode

Instantiate an empty recipe for a segmentation task.

Parameters:

  • pantry (IngredientCache) –

    The ingredients storage to associate with the new recipe.

Returns:

  • recipe ( RecipeNode ) –

    A blank recipe to populate with segmentation ingredients.

Starting at LEIP 4.0 users will be able to configure and execute compilation tasks from Recipe Designer. Under the hood this is calling LEIP Optimize

leip_recipe_designer.create.empty_compiler_recipe 

empty_compiler_recipe(pantry: IngredientCache) -> RecipeNode