Update app.py

app.py

@@ -117,8 +117,8 @@ def download_and_merge_model(base_model_name, lora_model_name, output_dir, devic
     adapter_tokenizer = AutoTokenizer.from_pretrained(lora_model_name)
     added_tokens_decoder = adapter_tokenizer.added_tokens_decoder
     model.resize_token_embeddings(adapter_tokenizer.vocab_size + len(added_tokens_decoder))
-    model.load_adapter(lora_model_name, low_cpu_mem_usage=True)
-    model = model.merge_and_unload()
+    peft_model = PeftModel.from_pretrained(model, lora_model_name, low_cpu_mem_usage=True)
+    model = peft_model.merge_and_unload()
     model.save_pretrained(output_dir)
     adapter_tokenizer.save_pretrained(output_dir)
     return output_dir