From 35ef38c61b2de3ce905836f18cfffd80e56c0a00 Mon Sep 17 00:00:00 2001 From: Mohammad Miadh Angkad <176301910+mmangkad@users.noreply.github.com> Date: Sat, 28 Feb 2026 02:30:00 +0800 Subject: [PATCH] Remove gpt-oss hybrid swa gate for trtllm_mha (#19079) --- python/sglang/srt/server_args.py | 10 ---------- 1 file changed, 10 deletions(-) diff --git a/python/sglang/srt/server_args.py b/python/sglang/srt/server_args.py index 27196c549..8b9c0a9ff 100644 --- a/python/sglang/srt/server_args.py +++ b/python/sglang/srt/server_args.py @@ -1356,16 +1356,6 @@ class ServerArgs: f"- Decode: {decode_attn_backend}\n" ) - if ( - prefill_attn_backend == "trtllm_mha" - or decode_attn_backend == "trtllm_mha" - ): - # TODO: support swa kv indices translation for trtllm_mha attention backend - self.disable_hybrid_swa_memory = True - logger.warning( - "Disable hybrid SWA memory for GPT-OSS model with trtllm_mha attention backend." - ) - quant_method = get_quantization_config(hf_config) is_mxfp4_quant_format = quant_method == "mxfp4" if is_blackwell_supported():