Is it possible to register models into MMS with user specific custom python packages installed during model registration.
Example Scenario:
User 1 has model A, with a specific set of python dependencies (ex Tensorflow = 1.8.0) and its already registered in Multi Model server. Now User 2 wants to register model B, which again uses Tensorflow's latest version (2.8.0) into the same Multi Model server.
Question: Now how can user manage different versions of same python package within Multi Model server for different models. Does Multi Model server provide user an option install model specific python package installation to allow seamless model serving.
Hi,
Is it possible to register models into MMS with user specific custom python packages installed during model registration.
Example Scenario: User 1 has model A, with a specific set of python dependencies (ex Tensorflow = 1.8.0) and its already registered in Multi Model server. Now User 2 wants to register model B, which again uses Tensorflow's latest version (2.8.0) into the same Multi Model server.
Question: Now how can user manage different versions of same python package within Multi Model server for different models. Does Multi Model server provide user an option install model specific python package installation to allow seamless model serving.