Pre-requisites

Before training your algorithm on Sagemaker platform, it is important to make sure it is set up properly. Please refer to programming_guide.md in this repository for instructions.

SagemakerTemplate

1. Install Requirements (will be most likely taken care of by the dev-container)

   1. Docker
   2. AWS Command-Line Interface (AWS CLI)
   3. VS Code

2. Login into AWS and ECR

   1. lin - anywhere from the terminal in the dev-container
   2. Open up the AWS SSO that you use to sign in >> AWS Account >> Command line or programmatic access >> Option 3
   3. The region should be set to us-east-1
   4. The format should be set to json
   5. Check instruction 2.2 for session token

3. Copy code

   1. copy <absolute path to your code>

4. Build and push

   Note:- If an already built image exists in the ECR, you can speed up this step by running pull <unique identifier for your image> first.

   1. build_and_push <unique identifier for your image>

5. Alternatively, you can build and push your image to ECR in different steps

   Note:- If an already built image exists in the ECR, you can speed up this step by running pull <unique identifier for your image> first.

   1. build <unique identifier for your image>
   2. push <unique identifier for your image used in step 5.1>

Next Steps

1. Clone the SagemakerNotebook repo into your Sagemaker instance
2. Follow the instructions in the README

Troubleshooting

If you run into any issues, feel free to reach out to Ritik.