ModelConfig

Summary

A Custom Model in Tecton

Example

from tecton import ModelConfig

model_config = ModelConfig(
    name=”mymodel_text_embedder”,
    model_type="pytorch",
    model_file="model_file.py",
    environments=["custom-ml-env-01"],
    artifact_files=["model.safetensors", "tokenizer_config.json" …],
    input_schema=[Field(”my_text_col”, String), …],
    output_schema=Field(”text_embedding”)
)

Attributes

Name	Data Type	Description
name	str	Unique name of model
model_type	str	Type of Model (pytorch or text embedding)
description	Optional[str]	Description of Model
tags	Union[Dict[str, str], NoneType]	Tags associated with this Tecton Object (key-value pairs of arbitrary metadata).
model_file	str	Path of File containing model relative to where the Tecton object is defined.
artifact_files	Optional[List[str]]	Path of other files needed to load and run model relative to where the Tecton object is defined.
environments	List[str]	All the environments allowed for this custom model to run in.
input_schema	List[Field]	Input Schema for model. Each field in this schema corresponds to the input_columns specified in the Inference and is mapped by order.
output_schema	Field	Output Schema of model.

Methods

Name	Description
__init__(...)	Initialize ModelConfig
register()	Register a Model Locally so that it can be used in Local Feature Views.
run(...)	Test Run a Model.

run(...)

Test Run a Model.

Parameters

• mock_inputs (Dict[str,List[Any]]) - Mock Input for each columns defined in input schema

Returns

DataFrame: pandas.Dataframe with inputs and output columns.

Raises

• TectonValidationError: if the input non-parameters are invalid.