From a34d3abb54dd9d97862040cc3a73a77a7fdc787b Mon Sep 17 00:00:00 2001 From: Xiaoyu Zhang <35585791+BBuf@users.noreply.github.com> Date: Fri, 21 Nov 2025 18:15:46 +0800 Subject: [PATCH] [Clean code] Compressed_tensors_moe code clean (#13719) --- .../compressed_tensors/compressed_tensors_moe.py | 5 ----- 1 file changed, 5 deletions(-) diff --git a/python/sglang/srt/layers/quantization/compressed_tensors/compressed_tensors_moe.py b/python/sglang/srt/layers/quantization/compressed_tensors/compressed_tensors_moe.py index 28453602b..5ab499f67 100644 --- a/python/sglang/srt/layers/quantization/compressed_tensors/compressed_tensors_moe.py +++ b/python/sglang/srt/layers/quantization/compressed_tensors/compressed_tensors_moe.py @@ -52,11 +52,6 @@ if _use_aiter: logger = logging.getLogger(__name__) -def _mask_topk_ids_cpu_experts(topk_ids: torch.Tensor, num_gpu_experts: int): - """Mask topk_ids >= num_gpu_experts by setting them to -1.""" - topk_ids[topk_ids >= num_gpu_experts] = -1 - - class GPTQMarlinState(Enum): REPACK = enum.auto() READY = enum.auto()