SageMaker Core

Warnings

• gotcha Prefer the high-level `sagemaker` SDK for most use cases. `sagemaker-core` is a low-level dependency; direct imports are generally only needed for advanced customizations or when explicitly building on its base components.
  Fix: For general SageMaker interaction, `pip install sagemaker` and import from `sagemaker` (e.g., `from sagemaker.session import Session`, `from sagemaker.estimator import Estimator`). Only use `sagemaker_core` when absolutely necessary for specific low-level utilities.
• gotcha The `sagemaker-core` package has its own versioning (e.g., 2.x.x) which does not directly align with the major version of the main `sagemaker` Python SDK (which also has 2.x.x and 3.x.x lines). Manually pinning `sagemaker-core` can lead to unexpected compatibility issues.
  Fix: Always install the main `sagemaker` package and let it manage its `sagemaker-core` dependency. Avoid manually pinning `sagemaker-core` unless you are explicitly building a custom SageMaker extension and have thoroughly tested compatibility.
• gotcha Older versions of `sagemaker-core` (and thus older `sagemaker` SDK versions) have strict `protobuf` version constraints (e.g., `<4.0.0` for `sagemaker-core==2.7.1`). This can cause dependency conflicts when used in environments with other libraries requiring newer `protobuf` versions.
  Fix: If encountering `protobuf` conflicts, try upgrading your `sagemaker` SDK to a very recent version (e.g., `pip install 'sagemaker>=2.257.1,<3'`) which may have relaxed its own `protobuf` requirements. If that fails, consider using a dedicated virtual environment or carefully managing `protobuf` versions globally.

Install

• pip install sagemaker-core Install `sagemaker-core` directly
• pip install sagemaker Recommended: Install `sagemaker` which includes `sagemaker-core`

Imports

• Session

  from sagemaker_core.session import Session

  Initializes a SageMaker session; this is the core session logic.
• Processor

  from sagemaker_core.processor import Processor

  Base class for SageMaker Processing jobs. Users typically use `sagemaker.processing.Processor` which wraps this.

Quickstart

Demonstrates initializing a `Session` directly from `sagemaker_core.session`, retrieving basic AWS configuration details like region, account ID, default S3 bucket, and the execution role. This illustrates direct interaction with core SageMaker components.

import sagemaker_core.session
import os

# Initialize a SageMaker Session directly from sagemaker_core.
# This session can then be used to interact with AWS SageMaker services.
# For full functionality, you would typically use `sagemaker.session.Session`
# which provides more high-level abstractions.

try:
    sagemaker_session = sagemaker_core.session.Session()
    region = sagemaker_session.boto_region_name
    account_id = sagemaker_session.account_id()
    print(f"Successfully initialized sagemaker_core.session.Session.")
    print(f"AWS Region: {region}")
    print(f"AWS Account ID: {account_id}")

    # Example: Get default S3 bucket for SageMaker artifacts
    default_bucket = sagemaker_session.default_bucket()
    print(f"Default S3 bucket: {default_bucket}")

    # Example: Attempt to get the IAM execution role.
    # This often works automatically in SageMaker Studio/Notebooks.
    try:
        role = sagemaker_session.get_execution_role()
        print(f"Execution Role: {role}")
    except Exception as e:
        print(f"Could not get execution role (expected if not in SM Studio/notebook or roles not configured): {e}")

except Exception as e:
    print(f"Error initializing sagemaker_core.session.Session: {e}")
    print("Ensure AWS credentials are configured (e.g., via environment variables, ~/.aws/credentials, or IAM roles).")