[RL] use cpu group to prepare_mlp_sync_batch_raw when the server is offloaded (#10152)

2025-10-18 14:29:35 +08:00
parent 31b9f19e54
commit e68a2b5b2f
2 changed files with 4 additions and 1 deletions
--- a/python/sglang/bench_one_batch.py
+++ b/python/sglang/bench_one_batch.py
@@ -320,6 +320,7 @@ def _maybe_prepare_mlp_sync_batch(batch: ScheduleBatch, model_runner):
            speculative_num_draft_tokens=None,
            require_mlp_tp_gather=require_mlp_tp_gather(model_runner.server_args),
            disable_overlap_schedule=model_runner.server_args.disable_overlap_schedule,
+            offload_tags=set(),
        )