[Bug Fix] Reject pp_max_micro_batch_size=0 to prevent silent deadlock on generate() (#23799)

python/sglang/srt/managers/scheduler.py

@@ -702,7 +702,7 @@ class Scheduler(
             _,
             _,
         ) = self.tp_worker.get_worker_info()
-        if get_global_server_args().pp_max_micro_batch_size is None:
+        if not get_global_server_args().pp_max_micro_batch_size:
             get_global_server_args().pp_max_micro_batch_size = max(
                 self.max_running_requests // self.pp_size, 1
             )

python/sglang/srt/server_args.py

@@ -6619,6 +6619,13 @@ class ServerArgs:
             self.tp_size * self.pp_size
         ) % self.nnodes == 0, "tp_size must be divisible by number of nodes"
 
+        assert (
+            self.pp_max_micro_batch_size is None or self.pp_max_micro_batch_size >= 1
+        ), (
+            "pp_max_micro_batch_size must be a positive integer or None (for auto-compute). "
+            f"Got: {self.pp_max_micro_batch_size}"
+        )
+
         if self.pp_size > 1:
             assert (
                 self.disable_overlap_schedule and self.speculative_algorithm is None