Method get_profile_idx

scripts/trt.py:141–196

(self, p, model_name: str, model_type: ModelType)

                )

    def get_profile_idx(self, p, model_name: str, model_type: ModelType) -> (int, int):
        best_hr = None

        if self.is_img2img:
            hr_scale = 1
        else:
            hr_scale = p.hr_scale if p.enable_hr else 1
        (
            valid_models,
            distances,
            idx,
        ) = modelmanager.get_valid_models(
            model_name,
            p.width,
            p.height,
            p.batch_size,
            77,  # model_type
        )  # TODO: max_embedding, just ignore?
        if len(valid_models) == 0:
            gr.Error(
                f"""No valid profile found for ({model_name}) LOWRES. Please go to the TensorRT tab and generate an engine with the necessary profile. 
                If using hires.fix, you need an engine for both the base and upscaled resolutions. Otherwise, use the default (torch) U-Net."""
            )
            return None, None
        best = idx[np.argmin(distances)]
        best_hr = best

        if hr_scale != 1:
            hr_w = int(p.width * p.hr_scale)
            hr_h = int(p.height * p.hr_scale)
            valid_models_hr, distances_hr, idx_hr = modelmanager.get_valid_models(
                model_name,
                hr_w,
                hr_h,
                p.batch_size,
                77,  # model_type
            )  # TODO: max_embedding
            if len(valid_models_hr) == 0:
                gr.Error(
                    f"""No valid profile found for ({model_name}) HIRES. Please go to the TensorRT tab and generate an engine with the necessary profile. 
                    If using hires.fix, you need an engine for both the base and upscaled resolutions. Otherwise, use the default (torch) U-Net."""
                )
            merged_idx = [i for i, id in enumerate(idx) if id in idx_hr]
            if len(merged_idx) == 0:
                gr.Warning(
                    "No model available for both ({}) LOWRES ({}x{}) and HIRES ({}x{}). This will slow-down inference.".format(
                        model_name, p.width, p.height, hr_w, hr_h
                    )
                )
                return None, None
            else:
                _distances = [distances[i] for i in merged_idx]
                best_hr = merged_idx[np.argmin(_distances)]
                best = best_hr

        return best, best_hr

    def get_loras(self, p):