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

[NPU] Support radix-cache with mamba-extra-buffer for Qwen3.5 (#23891)

Browse Source
This commit is contained in:
silencejade
2026-05-11 09:40:44 +08:00
committed by GitHub
parent e9a15b95da
commit 0fc9598b3e
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
@@ -2465,8 +2465,8 @@ class ServerArgs:
) )
assert ( assert (
is_cuda() or is_musa() is_cuda() or is_musa() or is_npu()
), "Mamba extra_buffer is only supported on CUDA and MUSA devices with FLA backend" ), "Mamba extra_buffer is only supported on CUDA and MUSA and NPU devices with FLA backend"
if self.speculative_num_draft_tokens is not None: if self.speculative_num_draft_tokens is not None:
assert ( assert (
self.mamba_track_interval >= self.speculative_num_draft_tokens self.mamba_track_interval >= self.speculative_num_draft_tokens