[Fix] Fix the case where prompt_len = 0 (#1593)

python/sglang/srt/managers/scheduler.py

@@ -626,9 +626,11 @@ class Scheduler:
             else:
                 logits_output = None
                 if self.tokenizer is not None:
-                    next_token_ids = [self.tokenizer.eos_token_id] * len(batch.reqs)
+                    next_token_ids = torch.full(
+                        (batch.batch_size(),), self.tokenizer.eos_token_id
+                    )
                 else:
-                    next_token_ids = [0] * len(batch.reqs)
+                    next_token_ids = torch.full((batch.batch_size(),), 0)
             return logits_output, next_token_ids
         else:  # embedding or reward model
             assert batch.extend_num_tokens != 0

python/sglang/srt/managers/tokenizer_manager.py

@@ -526,7 +526,7 @@ class TokenizerManager:
             async with self.model_update_lock:
                 # wait for the previous generation requests to finish
                 while len(self.rid_to_state) > 0:
-                    await asyncio.sleep(0)
+                    await asyncio.sleep(0.001)
                 self.send_to_scheduler.send_pyobj(obj)
                 self.model_update_result = asyncio.Future()
                 result = await self.model_update_result

python/sglang/srt/utils.py

@@ -624,6 +624,6 @@ def broadcast_pyobj(
         tensor_data = torch.empty(size, dtype=torch.uint8)
         dist.broadcast(tensor_data, src=0, group=dist_group)
 
-        serialized_data = bytes(tensor_data.tolist())
+        serialized_data = bytes(tensor_data.cpu().numpy())
         data = pickle.loads(serialized_data)
         return data