Remove redundant type conversion (#4513)

2025-03-17 05:57:35 -07:00
parent 5f9b2c62ff
commit 82dec1f70b
6 changed files with 16 additions and 10 deletions
--- a/python/sglang/srt/managers/tp_worker_overlap_thread.py
+++ b/python/sglang/srt/managers/tp_worker_overlap_thread.py
@@ -69,7 +69,7 @@ class TpModelWorkerClient:
        self.future_token_ids_ct = 0
        self.future_token_ids_limit = self.max_running_requests * 3
        self.future_token_ids_map = torch.empty(
-            (self.max_running_requests * 5,), dtype=torch.int32, device=self.device
+            (self.max_running_requests * 5,), dtype=torch.int64, device=self.device
        )

        # Launch threads