mlserver_custom_model_with_decode_args_predict.py

1from mlserver import MLModel
2from mlserver.types import InferenceRequest, InferenceResponse
3from mlserver.codecs import decode_args
4
5class MyModel(MLModel):
6    async def load(self) -> bool:
7        # Perform any heavy loading logic here (e.g. loading weights)
8        self.ready = True
9        return self.ready
10
11    @decode_args
12    async def predict(self, payload: list[str]) -> InferenceResponse:
13        # Custom logic to process the input payload
14        # In this example, we just return the input as is
15        return InferenceResponse(
16            model_name=self.name,
17            model_version=self.version,
18            outputs=[]
19        )