Update app.py

app.py

@@ -42,6 +42,8 @@ base_model = AutoModelForCausalLM.from_pretrained(
 )
 
 model = PeftModel.from_pretrained(base_model, "lakshyaag/llama38binstruct_summarize")
+model.merge_and_unload()
+
 
 # Move model to GPU if available
 if torch.cuda.is_available():
@@ -95,9 +97,7 @@ async def main(message: cl.Message):
     )
 
     # decode output from tokenized output to str output
-    decoded_output = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
-
-    print(decoded_output)
+    decoded_output = tokenizer.batch_decode(generated_ids)
 
     # return only the generated response (not the prompt) as output
     response = decoded_output[0].split("<|end_header_id|>")[-1]