Function postprocess_decode

src/diffusers/utils/remote_utils.py:94–144

(
    response: requests.Response,
    processor: "VaeImageProcessor" | "VideoProcessor" | None = None,
    output_type: Literal["mp4", "pil", "pt"] = "pil",
    return_type: Literal["mp4", "pil", "pt"] = "pil",
    partial_postprocess: bool = False,
)

def postprocess_decode(
    response: requests.Response,
    processor: "VaeImageProcessor" | "VideoProcessor" | None = None,
    output_type: Literal["mp4", "pil", "pt"] = "pil",
    return_type: Literal["mp4", "pil", "pt"] = "pil",
    partial_postprocess: bool = False,
):
    if output_type == "pt" or (output_type == "pil" and processor is not None):
        output_tensor = response.content
        parameters = response.headers
        shape = json.loads(parameters["shape"])
        dtype = parameters["dtype"]
        torch_dtype = DTYPE_MAP[dtype]
        output_tensor = torch.frombuffer(bytearray(output_tensor), dtype=torch_dtype).reshape(shape)
    if output_type == "pt":
        if partial_postprocess:
            if return_type == "pil":
                output = [Image.fromarray(image.numpy()) for image in output_tensor]
                if len(output) == 1:
                    output = output[0]
            elif return_type == "pt":
                output = output_tensor
        else:
            if processor is None or return_type == "pt":
                output = output_tensor
            else:
                if isinstance(processor, VideoProcessor):
                    output = cast(
                        list[Image.Image],
                        processor.postprocess_video(output_tensor, output_type="pil")[0],
                    )
                else:
                    output = cast(
                        Image.Image,
                        processor.postprocess(output_tensor, output_type="pil")[0],
                    )
    elif output_type == "pil" and return_type == "pil" and processor is None:
        output = Image.open(io.BytesIO(response.content)).convert("RGB")
        detected_format = detect_image_type(response.content)
        output.format = detected_format
    elif output_type == "pil" and processor is not None:
        if return_type == "pil":
            output = [
                Image.fromarray(image)
                for image in (output_tensor.permute(0, 2, 3, 1).float().numpy() * 255).round().astype("uint8")
            ]
        elif return_type == "pt":
            output = output_tensor
    elif output_type == "mp4" and return_type == "mp4":
        output = response.content
    return output


def prepare_decode(