Adjust temperature

app.py

@@ -33,7 +33,7 @@ def generate_text(prompt, example):
 
     logging.info(f"Input : {input}")
 
-    output_stream = llm(input, max_tokens=1024, stop=["</s>", "###"], stream=True)
+    output_stream = llm(input, max_tokens=-1, stop=["</s>", "###"], stream=True, temperature=args.temp)
 
     full_text = ""
     for output_chunk in output_stream:
@@ -48,6 +48,7 @@ parser = argparse.ArgumentParser()
 parser.add_argument("--model", help="Path to the model file")
 parser.add_argument("--gpu", help="How many GPU layers to use", default=0, type=int)
 parser.add_argument("--ctx", help="How many context tokens to use", default=512, type=int)
+parser.add_argument("--temp", help="temperature", default=0.1, type=float)
 args = parser.parse_args()
 
 if args.model: