Canvas

lit.sdk.model.canvas

This module provides functionality for managing model canvases in the LIT platform. A model canvas is a no-code representation of a machine learning model, allowing users to define, modify, and manage model configurations.

Canvas

Represents a model canvas, the no-code representation of a machine learning model.

inputs property

Retrieves the names of the inputs to the canvas.

Returns:

Type	Description
list[str]	A list input names.

Examples:

>>> canvas = Asset("contoso", "demo")
>>> canvas.inputs
[['90_1daybar_lp%'],
['RSI14TA_day'],
['vwap2060_day'],
['20sma_day'],
['ATR14_day']]

name = name instance-attribute

The name of the canvas.

outputs property

Retrieves the names of the outputs from the canvas.

Returns:

Type	Description
list[str]	A list input names.

Examples:

>>> canvas = Asset("contoso", "demo")
>>> canvas.outputs
[{'component_name': '',
  'label_index': 4,
  'gaussian_noise': 0.5,
  'dropout': 0.5,
  'class_weighting': 1,
  'activation': 'auto',
  'feature': 'long_updown'}]

path instance-attribute

The path to the JSON representation of the canvas on disk.

team = team instance-attribute

The name of a team.

__init__(team, name=None, path=None)

Initializes a new instance of Canvas.

Parameters:

Name	Type	Description	Default
team	str	The team which owns the Canvas.	required
name	str	The name of the Canvas.	None

duplicate()

Duplicates a canvas.

Raises:

Type	Description
LitModelError	If an error occurred while duplicating the canvas.

Returns:

Type	Description
Canvas	The duplicated canvas.

get_available_components()

List available components that can be added to a design canvas.

Raises:

Type	Description
LitModelError	If an error occurred while listing the available components.

Returns:

Type	Description
list	a list of components.

Examples:

>>> canvas = Canvas('contoso', 'demo')
>>> available_components = canvas.get_available_components()

remove()

Removes a canvas permanently.

Raises:

Type	Description
LitModelError	If an error occurred while deleting the canvas.

train(device, session_name=None, temp_canvas=False)

Starts a new experiment.

Parameters:

Name	Type	Description	Default
device	str	The name of a device; e.g. gpu4	required

Raises:

Type	Description
LitModelError	If an error occurred while starting the experiment.

Returns:

Name	Type	Description
session_name	str	The name of the screen session running the experiment.

transpile()

Compiles a design canvas.

Raises:

Type	Description
LitModelError	If an error occurred while [ACTION].

Returns:

Type	Description
Functional	A Tensorflow Keras model.

Examples:

Compile the canvas named 'demo' for the team named 'contoso' to a Tensorflow Keras model.

>>> canvas = Canvas('contoso', 'demo')
>>> model = canvas.transpile()
>>> model
<keras.src.models.Functional at 0x7f3ed7193850>

create_canvas(team, project, asset, features)

Create a new design canvas.

Parameters:

Name	Type	Description	Default
team	str	The name of the team.	required
project	str	The name of the project.	required
asset	str	Name of production asset.	required
features	list[str]	List of features.	required

Raises:

Type	Description
LitModelError	If an error occurred while [ACTION].

Returns:

Name	Type	Description
Canvas	Canvas	Canvas object

Examples:

>>> create_canvas('contoso', 'my_project')
Canvas(team="contoso", name="my_project")

get_canvases(team)

Get the saved design canvases for a given team.

Parameters:

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

Raises:

Type	Description
LitModelError	If an error occurred while [ACTION].

Returns:

Type	Description
list[Canvas]	A list of canvas objects.

Examples:

Get the path of the first saved definition file for team 'contoso'.

>>> canvases = get_canvases('contoso')
>>> canvases[0].path
/data/contoso/definitions/demo.json