Packaging your model (SDK) - Amazon Lookout for Vision

End of support notice: On October 31, 2025, AWS will discontinue support for Amazon Lookout for Vision. After October 31, 2025, you will no longer be able to access the Lookout for Vision console or Lookout for Vision resources. For more information, visit this blog post.

Packaging your model (SDK)

You package a model as a model component by creating a model packaging job. To create a model packaging job you call the StartModelPackagingJob API. The job might take a while to complete. To find out the current status, call DescribeModelPackagingJob and check the Status field in the response.

For information about package settings, see Package settings.

The following procedure shows you how to start a packaging job by using the AWS CLI. You can package the model for a target platform or a target device. For example Java code, see StartModelPackagingJob.

To package your model (SDK)

  1. If you haven't already done so, install and configure the AWS CLI and the AWS SDKs. For more information, see Step 4: Set up the AWS CLI and AWS SDKs.

  2. Make sure that you have the correct permissions to start a model packaging job. For more information, see StartModelPackagingJob.

  3. Use the following CLI commands to package your model for either a target device or a target platform.

    Target platform

    The following CLI command shows how to package a model for a target platform with an NVIDIA accelerator.

    Change the following values:

    • project_name to the name of the project that contains the model that you want to package.

    • model_version to the version of the model that you want to package.

    • (Optional) description to a description for your model packaging job.

    • architecture to the architecture (ARM64 or X86_64) of the AWS IoT Greengrass Version 2 core device where you run the model component.

    • gpu_code to the gpu code of the core device where you run the model component.

    • trt_ver to the TensorRT version you have installed on your core device.

    • cuda_ver to the CUDA version you have installed on your core device.

    • component_name to a name for the model component that you want to create on AWS IoT Greengrass V2.

    • (Optional) component_version to a version for the model component that the packaging job creates. Use the format major.minor.patch. For example, 1.0.0 represents the first major release for a component.

    • bucket to the Amazon S3 bucket where the packaging job stores the model component artifacts.

    • prefix to the location within the Amazon S3 bucket where the packaging job stores the model component artifacts.

    • (Optional) component_description to a description for the model component.

    • (Optional) tag_key1 and tag_key2 to the keys for tags that are attached to the model component.

    • (Optional) tag_value1 and tag_value2 to the key values for the tags that are attached to the model component.

    aws lookoutvision start-model-packaging-job \
  --project-name project_name \
  --model-version model_version \
  --description="description" \
  --configuration "Greengrass={TargetPlatform={Os='LINUX',Arch='architecture',Accelerator='NVIDIA'},CompilerOptions='{\"gpu-code\": \"gpu_code\", \"trt-ver\": \"trt_ver\", \"cuda-ver\": \"cuda_ver\"}',S3OutputLocation={Bucket='bucket',Prefix='prefix'},ComponentName='Component_name',ComponentVersion='component_version',ComponentDescription='component_description',Tags=[{Key='tag_key1',Value='tag_value1'}, {Key='tag_key2',Value='tag_value2'}]}" \
  --profile lookoutvision-access

    For example:

    aws lookoutvision start-model-packaging-job \
  --project-name test-project-01 \
  --model-version 1 \
  --description="Model Packaging Job for G4 Instance using TargetPlatform Option" \
  --configuration "Greengrass={TargetPlatform={Os='LINUX',Arch='X86_64',Accelerator='NVIDIA'},CompilerOptions='{\"gpu-code\": \"sm_75\", \"trt-ver\": \"7.1.3\", \"cuda-ver\": \"10.2\"}',S3OutputLocation={Bucket='bucket',Prefix='test-project-01/folder'},ComponentName='SampleComponentNameX86TargetPlatform',ComponentVersion='0.1.0',ComponentDescription='This is my component',Tags=[{Key='modelKey0',Value='modelValue'}, {Key='modelKey1',Value='modelValue'}]}" \
  --profile lookoutvision-access
    Target Device

    Use the following CLI commands to package a model for a target device.

    Change the following values:

    • project_name to the name of the project that contains the model that you want to package.

    • model_version to the version of the model that you want to package.

    • (Optional) description to a description for your model packaging job.

    • component_name to a name for the model component that you want to create on AWS IoT Greengrass V2.

    • (Optional) component_version to a version for the model component that the packaging job creates. Use the format major.minor.patch. For example, 1.0.0 represents the first major release for a component.

    • bucket to the Amazon S3 bucket where the packaging job stores the model component artifacts.

    • prefix to the location within the Amazon S3 bucket where the packaging job stores the model component artifacts.

    • (Optional) component_description to a description for the model component.

    • (Optional) tag_key1 and tag_key2 to the keys for tags that are attached to the model component.

    • (Optional) tag_value1 and tag_value2 to the key values for the tags that are attached to the model component.

    aws lookoutvision start-model-packaging-job \
  --project-name project_name \
  --model-version model_version \
  --description="description" \
  --configuration "Greengrass={TargetDevice='jetson_xavier',S3OutputLocation={Bucket='bucket',Prefix='prefix'},ComponentName='component_name',ComponentVersion='component_version',ComponentDescription='component_description',Tags=[{Key='tag_key1',Value='tag_value1'}, {Key='tag_key2',Value='tag_value2'}]}" \
  --profile lookoutvision-access

    For example:

    aws lookoutvision start-model-packaging-job \
  --project-name project_01 \
  --model-version 1 \
  --description="description" \
  --configuration "Greengrass={TargetDevice='jetson_xavier',S3OutputLocation={Bucket='bucket',Prefix='component_folder'},ComponentName='jetson_component',ComponentVersion='2.0.0',ComponentDescription='jetson model component',Tags=[{Key='tag_key1',Value='tag_value1'}, {Key='tag_key2',Value='tag_value2'}]}" \
  --profile lookoutvision-access

  4. Note the value of JobName in the response. You need it in the next step. For example: 

    {
    "JobName": "6bcfd0ff-90c3-4463-9a89-6b4be3daf972"
}

  5. Use DescribeModelPackagingJob to get the current status of the job. Change the following:

    • project_name to the name of the project that you are using.

    • job_name to the name of the job that you noted in the previous step.

    aws lookoutvision describe-model-packaging-job \
    --project-name project_name \
    --job-name job_name \
    --profile lookoutvision-access

    The model packaging job is complete if the value of Status is SUCCEEDED. If the value is different, wait a minute and try again.

  6. Continue deployment using AWS IoT Greengrass V2. For more information, see Deploying your components to a device.