def configure()

in threestudio/models/guidance/controlnet_guidance.py [0:0]

• 74 lines of code
• 16 McCabe index (conditional complexity)

    def configure(self) -> None:
        threestudio.info(f"Loading ControlNet ...")

        controlnet_name_or_path: str
        if self.cfg.control_type in ("normal", "input_normal"):
            controlnet_name_or_path = "lllyasviel/control_v11p_sd15_normalbae"
        elif self.cfg.control_type == "canny":
            controlnet_name_or_path = "lllyasviel/control_v11p_sd15_canny"

        self.weights_dtype = (
            torch.float16 if self.cfg.half_precision_weights else torch.float32
        )

        pipe_kwargs = {
            "safety_checker": None,
            "feature_extractor": None,
            "requires_safety_checker": False,
            "torch_dtype": self.weights_dtype,
            "cache_dir": self.cfg.cache_dir,
        }

        controlnet = ControlNetModel.from_pretrained(
            controlnet_name_or_path,
            torch_dtype=self.weights_dtype,
            cache_dir=self.cfg.cache_dir,
        )
        self.pipe = StableDiffusionControlNetPipeline.from_pretrained(
            self.cfg.pretrained_model_name_or_path, controlnet=controlnet, **pipe_kwargs
        ).to(self.device)
        self.scheduler = DDIMScheduler.from_pretrained(
            self.cfg.ddim_scheduler_name_or_path,
            subfolder="scheduler",
            torch_dtype=self.weights_dtype,
            cache_dir=self.cfg.cache_dir,
        )
        self.scheduler.set_timesteps(self.cfg.diffusion_steps)

        if self.cfg.enable_memory_efficient_attention:
            if parse_version(torch.__version__) >= parse_version("2"):
                threestudio.info(
                    "PyTorch2.0 uses memory efficient attention by default."
                )
            elif not is_xformers_available():
                threestudio.warn(
                    "xformers is not available, memory efficient attention is not enabled."
                )
            else:
                self.pipe.enable_xformers_memory_efficient_attention()

        if self.cfg.enable_sequential_cpu_offload:
            self.pipe.enable_sequential_cpu_offload()

        if self.cfg.enable_attention_slicing:
            self.pipe.enable_attention_slicing(1)

        if self.cfg.enable_channels_last_format:
            self.pipe.unet.to(memory_format=torch.channels_last)

        # Create model
        self.vae = self.pipe.vae.eval()
        self.unet = self.pipe.unet.eval()
        self.controlnet = self.pipe.controlnet.eval()

        if self.cfg.control_type == "normal":
            self.preprocessor = NormalBaeDetector.from_pretrained(
                "lllyasviel/Annotators"
            )
            self.preprocessor.model.to(self.device)
        elif self.cfg.control_type == "canny":
            self.preprocessor = CannyDetector()

        for p in self.vae.parameters():
            p.requires_grad_(False)
        for p in self.unet.parameters():
            p.requires_grad_(False)

        self.num_train_timesteps = self.scheduler.config.num_train_timesteps
        self.set_min_max_steps()  # set to default value

        self.alphas: Float[Tensor, "..."] = self.scheduler.alphas_cumprod.to(
            self.device
        )

        self.grad_clip_val: Optional[float] = None

        if self.cfg.use_du:
            if self.cfg.cache_du:
                self.edit_frames = {}
            self.perceptual_loss = PerceptualLoss().eval().to(self.device)

        threestudio.info(f"Loaded ControlNet!")