Update

models.py

@@ -84,6 +84,12 @@ def generate(
   print(gradio_helpers.get_paths())
   model = llama_cpp.Llama(
       model_path,
+      chat_handler=llama_cpp.llama_chat_format.PaligemmaChatHandler(
+        clip_path
+      ),
+      n_ctx=1024,
+      n_ubatch=512,
+      n_batch=512,
   )
   # return model.tokenizer.to_str(tokens[0])
   return "output"