From d36e96ce234a502013e254d47f2f5bf6ceb65acb Mon Sep 17 00:00:00 2001 From: Bingxu Chen Date: Thu, 16 Jul 2026 01:18:12 +0800 Subject: [PATCH] [AMD] Enable mamba-extra-buffer for Qwen3.5 on ROCm (#30359) Co-authored-by: ntgiang71096 --- python/sglang/srt/server_args.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/python/sglang/srt/server_args.py b/python/sglang/srt/server_args.py index ea83e2041..187cf992d 100644 --- a/python/sglang/srt/server_args.py +++ b/python/sglang/srt/server_args.py @@ -4683,8 +4683,8 @@ class ServerArgs: view, model_arch ), f"extra_buffer is not supported for {model_arch}; use no_buffer." assert ( - is_cuda() or is_musa() or is_npu() - ), "extra_buffer needs CUDA/MUSA/NPU (FLA)." + is_cuda() or is_musa() or is_npu() or is_hip() + ), "extra_buffer needs CUDA/MUSA/NPU/ROCm (FLA)." if view.speculative_num_draft_tokens is not None: assert ( view.mamba_radix_cache_strategy != "extra_buffer_lazy"