When using the model sidecar, the model version should be exposed. This is basic functionality that should be enabled to support prediction tracability to that you can record which model was used to generate a prediction.
Hi Devon!
You can access the model version through the provided model adapter:
@transform(
data=Input(...)
model=ModelInput(..., use_sidecar=True)
def compute(
data: LightweightInput,
model_input: pm.ModelAdapter,
):
# print the model version rid
log.info(model_input.model_version_rid)