feat: normalize cosine similarity

lib/utils/model.py

@@ -1,6 +1,7 @@
 import streamlit as st 
 import yaml
 import torch
+import torch.nn.functional as F
 
 from lib.IRRA.tokenizer import tokenize, SimpleTokenizer
 from lib.IRRA.image import prepare_images
@@ -27,4 +28,7 @@ def get_similarities(text: str, images: list[str], model: IRRA) -> torch.Tensor:
     image_feats = model.encode_image(imgs)
     text_feats = model.encode_text(txt.unsqueeze(0))
 
+    image_feats = F.normalize(image_feats, p=2, dim=1)
+    text_feats = F.normalize(text_feats, p=2, dim=1)
+    
     return text_feats @ image_feats.t()