TorchServe for Partially Trained Models


I am reading into torchserve and from my understanding it provides a system for clients to send inference requests to a trained model.

Is it possible to use a model which needs to perform gradient ascent to predict values with torchserve, using a partially trained model?

If not, what alternatives could one use?


Could you elaborate what you mean by gradient ascent on a partially trained model? If your intent is to perform an inference on a pretrained model then that should work as is with torchserve,

If your intent is to retrain a model based on the result of some inference then you can deploy pretrained model, get inference, replace via a new model - see for more information