fixed tokenizer

app.py

@@ -240,7 +240,7 @@ def inference(img, text, is_translate):
     # encode text
     if text is None:
       return "No text provided"
-    text = clip_model.tokenize([text]).to(device)
+    text = clip.tokenize([text]).to(device)
     with torch.no_grad():
       prefix = clip_model.encode_text(text).to(device, dtype=torch.float32)