Model URI

Supported URIs


Google Cloud Storage


Please make sure your google cloud storage bucket is public.



This path is mapped to /phfs/path/to/the/model path in the PrimeHub Notebook.

MLflow Model


It maps to the MLflow Model Registry. The version could be either a version number or the stage of a model.



This path is mapped to /nfs/to/my/model in the Model Deployment containers

How does the Model URI work?

If a PrimeHub Deployment runs with a Model URI, it will start a model downloader before a deployment gets ready to serve.

The deployment and its model downloader will share the same volume /mnt/models. When the downloader finished its job, the deployment could load and create a model to serve.

Test GCS Model URI by Docker

In model deployment, we use the image to download the model files into the /mnt/models folder.

You can use the following command to test if the download process is work as expected.

docker run -v ${PWD}/models:/mnt/models --rm  \ \
  gs://primehub-models/pytorch/CIFAR10 \

Last updated