sagemaker.core.jumpstart.payload_utils

This module stores inference payload utilities for JumpStart models.

Classes

PayloadSerializer([bucket, region, s3_client])

Utility class for serializing payloads associated with JumpStart models.

class sagemaker.core.jumpstart.payload_utils.PayloadSerializer(bucket: str | None = None, region: str | None = None, s3_client: client | None = None)

Bases: object

Utility class for serializing payloads associated with JumpStart models.

Many JumpStart models embed byte-streams into payloads corresponding to images, sounds, and other content types which require downloading from S3.

embed_s3_references_in_json_payload(payload_body: list | dict | str | int | float) → list | dict | str | int | float

Finds all S3 references in payload and embeds serialized S3 data.

If no S3 references are found, the payload is returned un-modified.

Raises:

ValueError – If the payload has an unrecognized type.

embed_s3_references_in_str_payload(payload: str) → str

Inserts serialized S3 content into string payload.

If no S3 content is embedded in payload, original string is returned.

get_bytes_payload_with_s3_references(payload_str: str) → bytes

Returns bytes object corresponding to referenced S3 object.

Raises:

ValueError – If the raw bytes payload is not formatted correctly.

serialize(payload: JumpStartSerializablePayload) → str | bytes

Returns payload string or bytes that can be inputted to inference endpoint.

Raises:

ValueError – If the payload has an unrecognized type.