From db452760e5b2378efd06b1ceb9385d2eeb6d217c Mon Sep 17 00:00:00 2001 From: Xiaoyu Zhang <35585791+BBuf@users.noreply.github.com> Date: Mon, 7 Apr 2025 21:15:46 +0800 Subject: [PATCH] [ci] fix llama4 ci error (#5126) --- .../quantization/compressed_tensors/compressed_tensors_moe.py | 2 ++ 1 file changed, 2 insertions(+) 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 7e5b3231f..393d6369c 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 @@ -285,6 +285,7 @@ class CompressedTensorsW8A8Fp8MoEMethod(CompressedTensorsMoEMethod): activation: str = "silu", inplace: bool = True, no_combine: bool = False, + apply_router_weight_on_input: bool = False, ) -> torch.Tensor: from sglang.srt.layers.moe.fused_moe_triton import fused_experts from sglang.srt.layers.moe.topk import select_experts @@ -314,6 +315,7 @@ class CompressedTensorsW8A8Fp8MoEMethod(CompressedTensorsMoEMethod): w2_scale=layer.w2_weight_scale, a1_scale=layer.w13_input_scale, a2_scale=layer.w2_input_scale, + apply_router_weight_on_input=apply_router_weight_on_input, )