Adding the model and container¶

You can create an algorithm in one of two ways: either by linking a GitHub repository (see Option 1) or by uploading a container directly (see Option 2). In both cases, you have the flexibility to include your model weights within the container or upload them separately.

⚠️ Note on GitHub Storage Limits.

If you choose Option 1 and include the model weights in your GitHub repository, be aware that GitHub enforces bandwidth and storage limits of 1 GiB per month for repository downloads. Your algorithm will not be updated when this limit is exceeded. For larger models or frequent use, consider uploading the container and model weights separately to avoid these limitations.

If you choose to upload the weights separately, go to the Models page of your algorithm and upload the model files—ideally as a tarball. During inference, the tarball will be extracted and made available to your container at /opt/ml/model/.