Fix side effect brought by supporting codellama: sequence_start is …

…always true when calling `model.get_prompt` (#466)
InternLM · Sep 25, 2023 · e980377 · e980377
1 parent 7194500
commit e980377
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/lmdeploy/turbomind/chat.py b/lmdeploy/turbomind/chat.py
@@ -123,7 +123,7 @@ def main(model_path,
             step = 0
             seed = random.getrandbits(64)
         else:
-            prompt = model.get_prompt(prompt, nth_round)
+            prompt = model.get_prompt(prompt, nth_round == 1)
             input_ids = tokenizer.encode(prompt)
             if step + len(input_ids) >= tm_model.session_len:
                 print('WARNING: exceed session max length.'