[FEATURE] Quickstart example for sockeye about budgetml HOT 3 CLOSED

Hi @michaelhochleitner! I was not really aware of sockeye before your request. But from a brief glance into the repo and docs it seems it should be possible to pass sockeye in the requirements array and then just build a predictor that uses the MXNet model underneath to make a prediction. See the example for details on how to do it with HuggingFace (its the same logic really). You dont need to use sockeye_serving for this.

What would help to give a more concrete example is if you could give me the code to actually run inference on a trained sockeye model. How does it consume the data (What format?). How does it do the forward pass in the Transformer? Could you send over a script where you do that locally and then I can add an example in the repo to use BudgetML to deploy an API for it

from budgetml.

@michaelhochleitner Any update on this? Otherwise id have to close

from budgetml.

Closing due to inactivity

from budgetml.