Model Weights not mounted at runtime.

Despite uploading a valid algorithmmodel.tar.gz file as instructed, the containerized algorithm does not find the model weights under /opt/ml/model during runtime. An exhaustive inspection of the container filesystem shows that only the input file is present, and no model files are mounted.

Hi Dainele,

This seems to be docker image issue. Are you sure you are correctly mounting model weights in your docker image and tests work locally?

“You can upload your algorithm's model separately from the container as a tarball (.tar.gz). If provided, tarball will be extracted to /opt/ml/model/ at runtime”

From what I read here it seems like they will be automatically mounted at runtime once you provide the .tar.gz in the models section.