This PR partially addresses issue #11 and adds a Ray Serve Notebook to the repo. It also includes some changes to how the model is saved in the Pet_Classifier notebook to make model serving a bit simpler.
One outstanding issue to take care of in a future PR, this implementation uses CPU for inference and does not take advantage of the available GPU's. GPU usage was causing issues on the cluster and needs to be investigated further.
This PR partially addresses issue #11 and adds a Ray Serve Notebook to the repo. It also includes some changes to how the model is saved in the Pet_Classifier notebook to make model serving a bit simpler.
One outstanding issue to take care of in a future PR, this implementation uses CPU for inference and does not take advantage of the available GPU's. GPU usage was causing issues on the cluster and needs to be investigated further.