Hi there, I'm having a small problem with input type for api

class SpeechRecognizerRunnable(bentoml.Runnable):
    SUPPORTED_RESOURCES = ("<http://nvidia.com/gpu|nvidia.com/gpu>",)
    SUPPORTS_CPU_MULTI_THREADING = True

    def __init__(self) -> None:
        self.__asr_model = ASRModel(model_size="medium")
        self.__vad_model = VADModel()

    @bentoml.Runnable.method(batchable=True, batch_dim=0)
    def predict(self, audio_bytes: bytes) -> str:
        return audio_transcriber(audio_bytes, self.__asr_model, self.__vad_model)
    
asr_runner = bentoml.Runner(SpeechRecognizerRunnable)
svc = bentoml.Service("asr_service", runners= [asr_runner])

@svc.api(input=bytes, output=str) # Error right here
async def transcribe(audio_bytes: bytes) -> str:
    text = await asr_runner.predict.async_run(audio_bytes)
    return text

And I got:

AttributeError: type object 'bytes' has no attribute 'input_type'

You guys have any idea how to define bytes type?

Hello, would you like to tryout our 1.2 API? bytes is support there.

@svc.api(input=bytes, output=str) # Error right here

This is a 1.1 API, which requires you to use IO descriptors instead of python types

Thanks. I'll give a try

Can websocket method batchable?

bentoml doesn't not have websocket support for now