sagemaker.serve.model_builder_utils

sagemaker.serve.model_builder_utils#

Utility functions and mixins for ModelBuilder.

This module provides utility functions for: - Session management and initialization - Instance type detection and optimization - Container image auto-detection - HuggingFace and JumpStart model handling - Resource requirement calculation - Framework serialization support - MLflow model integration - General model deployment utilities

Example

Basic usage as a mixin class:

class MyModelBuilder(ModelBuilderUtils):
    def __init__(self):
        self.model = "huggingface-model-id"
        self.instance_type = "ml.g5.xlarge"

    def build(self):
        self._auto_detect_image_uri()
        return self.image_uri

Classes

TritonSerializer(input_serializer, dtype[, ...])

A wrapper of JSONSerializer because Triton expects input to be certain format
class sagemaker.serve.model_builder_utils.TritonSerializer(input_serializer, dtype: str, content_type='application/json')[source]#

Bases: JSONSerializer

A wrapper of JSONSerializer because Triton expects input to be certain format

serialize(data)[source]#

Serialize data into Triton-compatible format.

Contents