[PP] Add pipeline parallelism (#5724)

2025-04-30 18:18:07 -07:00
parent e97e57e699
commit 11383cec3c
25 changed files with 1150 additions and 308 deletions
--- a/test/srt/test_vlm_accuracy.py
+++ b/test/srt/test_vlm_accuracy.py
@@ -147,6 +147,8 @@ class VisionLLMLogitsBase(unittest.IsolatedAsyncioTestCase):
            gpu_id=0,
            tp_rank=0,
            tp_size=1,
+            pp_rank=0,
+            pp_size=1,
            nccl_port=12435,
            server_args=ServerArgs(
                model_path=self.model_path,