Fix unit test order to balance the tasks in CI (#1665)

python/sglang/srt/managers/schedule_batch.py

@@ -421,6 +421,7 @@ class ScheduleBatch:
     extend_lens: List[int] = None
     extend_num_tokens: int = None
     running_bs: int = None
+    decoding_reqs: List[Req] = None
 
     # Stream
     has_stream: bool = False

python/sglang/test/few_shot_gsm8k.py

@@ -76,7 +76,9 @@ def run_eval(args):
     def few_shot_gsm8k(s, question):
         s += few_shot_examples + question
         s += sgl.gen(
-            "answer", max_tokens=512, stop=["Question", "Assistant:", "<|separator|>"]
+            "answer",
+            max_tokens=args.max_new_tokens,
+            stop=["Question", "Assistant:", "<|separator|>"],
         )
 
     #####################################
@@ -131,6 +133,7 @@ if __name__ == "__main__":
     parser.add_argument("--num-shots", type=int, default=5)
     parser.add_argument("--data-path", type=str, default="test.jsonl")
     parser.add_argument("--num-questions", type=int, default=200)
+    parser.add_argument("--max-new-tokens", type=int, default=512)
     parser.add_argument("--parallel", type=int, default=128)
     parser.add_argument("--host", type=str, default="http://127.0.0.1")
     parser.add_argument("--port", type=int, default=30000)