do sample

app.py

@@ -58,16 +58,17 @@ def predict(code):
     print(f"Tokenized: {tokenizer.tokenize(prompt)}")
     pipe_out = pipe(
         prompt,
-        max_length=10**6,
+        do_sample=True,
         top_k=100,
         top_p=0.09,
         temperature=1.2,
         repetition_penalty=1.1,
-        return_full_text=False
+        return_full_text=False,
     )
     print(f"Pipe out: {repr(pipe_out)}")
 
-    return pipe_out[0]['generated_text']
+    return pipe_out[0]["generated_text"]
+
 
 demo = gr.Interface(
     fn=predict,