[5/N] MoE Refactor: Update MoE parallelism arguments (#8658)

2025-08-01 01:20:03 -07:00
parent c8d3a402c1
commit 6c88f6c8d9
38 changed files with 342 additions and 299 deletions
--- a/test/srt/test_deepep_large.py
+++ b/test/srt/test_deepep_large.py
@@ -33,7 +33,8 @@ class TestDeepseek(CustomTestCase):
                "--moe-dense-tp-size",
                "1",
                "--enable-dp-lm-head",
-                "--enable-deepep-moe",
+                "--moe-a2a-backend",
+                "deepep",
                "--enable-two-batch-overlap",
                "--ep-num-redundant-experts",
                "32",
@@ -88,7 +89,8 @@ class TestDeepseekMTP(CustomTestCase):
                "--moe-dense-tp-size",
                "1",
                "--enable-dp-lm-head",
-                "--enable-deepep-moe",
+                "--moe-a2a-backend",
+                "deepep",
                "--enable-two-batch-overlap",
                "--ep-num-redundant-experts",
                "32",