[5/n] Lora support cuda graph (#21647)

2026-04-04 00:31:46 -07:00
parent a94c3804c2
commit ff8e47edf9
15 changed files with 311 additions and 70 deletions
@@ -4663,10 +4663,11 @@ class ServerArgs:
        parser.add_argument(
            "--experts-shared-outer-loras",
            default=ServerArgs.experts_shared_outer_loras,
-            action="store_true",
+            action=argparse.BooleanOptionalAction,
            help="Force shared outer LoRA mode for MoE models. "
            "When set, w1/w3 lora_A and w2 lora_B are shared across experts "
-            "(expert_dim=1). By default this is auto-detected from adapter weights.",
+            "(expert_dim=1). Use --no-experts-shared-outer-loras to force disable. "
+            "By default this is auto-detected from adapter weights.",
        )

        # Kernel backend