Method get_image_embedding

demo/inference.py:302–306

(self, image_file)

    @torch.no_grad()
    def get_image_embedding(self, image_file):
        pixel_values = load_image(image_file)
        pixel_values = pixel_values.unsqueeze(0).to(self.device, dtype=self.dtype)
        language_model_inputs = self.model.extract_feature(pixel_values)
        return language_model_inputs

    @torch.no_grad()
    def get_video_embedding(self, video_file):