def preprocess()

in torchserve/inf2/llama2/workspace/inf2_handler.py [0:0]

• 14 lines of code
• 5 McCabe index (conditional complexity)

    def preprocess(self, requests):
        input_text = []
        for req in requests:
            data = req.get("data") or req.get("body")
            if isinstance(data, (bytes, bytearray)):
                data = data.decode("utf-8")
            logger.info(f"received req={data}")
            input_text.append(data.strip())

        # Ensure the compiled model can handle the input received
        if len(input_text) > self.handle.micro_batch_size:
            raise ValueError(
                f"Model is compiled for batch size {self.handle.micro_batch_size} but received input of size {len(input_text)}"
            )

        # Pad input to match compiled model batch size
        input_text.extend([""] * (self.handle.micro_batch_size - len(input_text)))

        return self.tokenizer(input_text, return_tensors="pt", padding=True)