minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[AMD] Enable mamba-extra-buffer for Qwen3.5 on ROCm (#30359)

Browse Source 
Co-authored-by: ntgiang71096 <nguyentruonggiang71096@gmail.com>
This commit is contained in:
Bingxu Chen
2026-07-15 10:18:12 -07:00
committed by GitHub
co-authored by ntgiang71096
parent a8b60433c2
commit d36e96ce23
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/server_args.py
+2 -2
View File
@@ -4683,8 +4683,8 @@ class ServerArgs:
view, model_arch view, model_arch
), f"extra_buffer is not supported for {model_arch}; use no_buffer." ), f"extra_buffer is not supported for {model_arch}; use no_buffer."
assert ( assert (
is_cuda() or is_musa() or is_npu() is_cuda() or is_musa() or is_npu() or is_hip()
), "extra_buffer needs CUDA/MUSA/NPU (FLA)." ), "extra_buffer needs CUDA/MUSA/NPU/ROCm (FLA)."
if view.speculative_num_draft_tokens is not None: if view.speculative_num_draft_tokens is not None:
assert ( assert (
view.mamba_radix_cache_strategy != "extra_buffer_lazy" view.mamba_radix_cache_strategy != "extra_buffer_lazy"