The logs for the container, which usually go "Pulling -> Pulled -> Created -> Started" now consistently write "Pulling -> Pulling -> Pulled -> Created -> Started" when a GPU resource is requested.
This results in an extra 3-minutes to start a GPU-equipped ACI.
The logs for the container, which usually go "Pulling -> Pulled -> Created -> Started" now consistently write "Pulling -> Pulling -> Pulled -> Created -> Started" when a GPU resource is requested.
This results in an extra 3-minutes to start a GPU-equipped ACI.