Method encode_image

eval_clip_sim.py:19–23

(self, image)

        self.processor = CLIPProcessor.from_pretrained(model_name)

    def encode_image(self, image):
        pixel_values = self.processor(images=image, return_tensors="pt").pixel_values.to(self.device)
        image_features = self.clip_model.get_image_features(pixel_values=pixel_values)
        image_features = image_features / image_features.norm(dim=-1, keepdim=True)  # normalize features
        return image_features

    def encode_text(self, text):
        inputs = self.processor(text=[text], padding=True, return_tensors="pt").to(self.device)