Vault

lit.sdk.vault

This module provides functionality for interacting with the Lit vault, including managing and retrieving items stored in the vault.

VaultItem

Represents a single item in the Lit vault.

canvas property

Retrieves design canvas for the vaulted model.

Returns:

Type	Description
dict | None	The design canvas.

details property

Retrieves detailed information about this vault item from the Lit vault.

Returns:

Type	Description
dict | None	A dictionary containing detailed information about this vault item, or None if it does not exist in the vault.

Examples:

Getting the details of a vault item

>>> item = VaultItem("contoso", "my_model1")
>>> item.details
{'host': 'bogdan', 'project': 'my_project', 'runid': 3, 'epoch': 1, 'model': 'model.1.keras', 'name': 'my_model1', 'value': None, 'notes': None, 'color': None, 'dateAdded': '2024-07-12 19:40:19.783233+00:00', 'workers': []}

Getting a non-existent vault item

>>> item = VaultItem("contoso", "my_model0")
>>> item.details
None

name = name instance-attribute

The name of this item.

team = team instance-attribute

The name of the team that this item belongs to.

__init__(team, name)

Initializes a new instance of VaultItem.

Parameters:

Name	Type	Description	Default
team	str	The team which owns the VaultItem.	required
name	str	The name of the VaultItem.	required

get_data_from_path(path, start=None, stop=None, step=1)

Retrieve input data from a given asset path and format it for model prediction.

Parameters:

Name	Type	Description	Default
path	Path	The path of the asset	required
start	number	Start index [optional]	None
stop	number	Stop index [optional]	None

Returns:

Type	Description
[ndarray]	List of numpy arrays filled with data retrieved from the asset.

Examples:

Load the model, prepare input from a random index into the first datafile of the first asset used to train the model, make a prediction.

>>> model = vault_item.load_model()
>>> asset = vault_item.get_assets()[0]
>>> index = random.randint(0, len(asset.files[0]) - 1)
>>> x, y = vault_item.get_data_from_path(asset.files[0].path, index)
>>> model.predict(x)
array([[0.48219436]], dtype=float32)

load_model()

Loads the model.

Returns:

Type	Description
Model	A model grouping layers into an object with training/inference features.

Examples:

Load a model.

>>> model = vault_item.load_model()
>>> model.layers
[<InputLayer name=input_1, built=True>,
<Lambda name=lambda, built=True>,
<Lambda name=lambda_1, built=True>,
<Flatten name=flatten, built=True>,
<Flatten name=flatten_1, built=True>,
<Concatenate name=dense_40c93684-299b-41df-b233-912a47954a84_concatenate, built=True>,
<Dropout name=dropout, built=True>,
<Dense name=dense_40c93684-299b-41df-b233-912a47954a84_0, built=True>,
<GaussianNoise name=gaussian_noise, built=True>,
<Dropout name=dropout_1, built=True>,
<Dense name=end_d1657694-e01d-409b-aec8-15b33a9aefbc, built=True>]

rename(new_name)

Renames this vault item to the specified name.

Parameters:

Name	Type	Description	Default
new_name	str	The desired new name for this vault item.	required

Returns:

Type	Description
dict	A dictionary containing the status and message of the rename operation.

Examples:

Renaming a vault item

>>> item = VaultItem("contoso", "my_model1")
>>> item.rename("my_model2")
{'status': 'SUCCESS', 'message': 'Vault item my_model1 has been renamed my_model2.'}
>>> item.name
'my_model2'

Renaming a non-existent vault item

>>> item = VaultItem("contoso", "my_model0")
>>> item.rename("new_name")
{'status: 'ERROR', 'message': 'item my_model0 not found'}
>>> item.name
'my_model0'

Renaming a vault item to the name of another vault item

>>> item = VaultItem("contoso", "my_model1")
>>> item.rename("my_model1")
{'status: 'ERROR', 'message': 'name already exists'}
>>> item.name
'my_model1'

zeros()

Creates a single batch of initialized inputs for the underlying model. This can be passed directly into predict.

Returns:

Type	Description
[ndarray]	List of numpy arrays of zero-values that match the expected shape of inputs to the neural network.

Examples:

Get model, get zeros, make a prediction.

>>> model = vault_item.load_model()
>>> input_data = vault_item.zeros()
>>> model.predict(input_data)
array([[0.49490067]], dtype=float32)

get_vault_items(team)

Retrieves a list of vault items for the specified team from the Lit vault.

Parameters:

Name	Type	Description	Default
team	str	The name of the team to retrieve vault items for.	required

Returns:

Type	Description
list[VaultItem]	list[VaultItem]: A list of VaultItem objects representing the vault items for the specified team.

Examples:

>>> get_vault_items("contoso")
[VaultItem(team='contoso', name='spy_e1')]

import_model(team, path, features=[], name=None)

Import a Tensorflow model into the Lit vault.

Parameters:

Name	Type	Description	Default
team	str	The name of the team to retrieve vault items for.	required
path	str	The path to the saved model.	required
name	str	The name of the new model (optional).	None

Returns:

Name	Type	Description
VaultItem	VaultItem	A VaultItem objects representing the vault item for the imported model.

Examples:

>>> import_model("contoso", "/data/raw/import/model.h5")
VaultItem(team='contoso', name='new_model')