How do you pass an S3 path as a parameter in an asynchronous inference call to a model in SageMaker?

Yes, it is possible to specify different S3 paths for saving model results when making asynchronous inference calls to a SageMaker endpoint. When invoking an asynchronous endpoint, you can provide the S3 path where you want the results to be stored.

To do this, you use the InvokeEndpointAsync API call. In this call, you can specify an InputLocation parameter, which is the S3 URI of your input data. The response from this API call will include an OutputLocation, which is where the results will be stored.

Here's a general approach to achieve this:

1. For each inference call, prepare your input data and upload it to a specific S3 location.

2. Use the InvokeEndpointAsync API, providing the S3 URI of your input data as the InputLocation.

3. The API call will return a response object that includes the OutputLocation where the results will be stored.

4. You can then retrieve the results from this OutputLocation once the processing is complete.

Here's a simplified example of how you might structure this in Python using boto3:

import boto3
import uuid

sagemaker_runtime = boto3.client("sagemaker-runtime", region_name="your-region")

# Specify the S3 URI of the input
input_location = "s3://your-bucket-name/input/data.csv"

response = sagemaker_runtime.invoke_endpoint_async(
    EndpointName='your-endpoint-name',
    InputLocation=input_location,
    InferenceId=str(uuid.uuid4()),
    ContentType="text/csv"  # Adjust based on your data type
)

output_location = response['OutputLocation']
print(f"Results will be stored at: {output_location}")

By changing the input_location for each call, you can effectively control where the results are stored, as the output location is typically based on the input location.

Remember that the exact implementation may vary depending on your specific use case and the configuration of your SageMaker endpoint. Always ensure that your IAM roles have the necessary permissions to read from and write to the specified S3 locations.
Sources
Asynchronous inference - Amazon SageMaker AI
Check prediction results - Amazon SageMaker AI