From 10f6ab01a931cee9cc3ec4a141d47be7ee7f111c Mon Sep 17 00:00:00 2001 From: jackLei <42642542+jackLei0901@users.noreply.github.com> Date: Fri, 2 Oct 2026 11:37:17 -0700 Subject: [PATCH] [Perf] Use value-only reduction for native per-token FP8 quantization (#59800) Signed-off-by: jackLei0901 <841747318@qq.com> Co-authored-by: Misha Goin --- vllm/model_executor/layers/quantization/input_quant_fp8.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm/model_executor/layers/quantization/input_quant_fp8.py b/vllm/model_executor/layers/quantization/input_quant_fp8.py index 161baf446d..52cfb249dd 100644 --- a/vllm/model_executor/layers/quantization/input_quant_fp8.py +++ b/vllm/model_executor/layers/quantization/input_quant_fp8.py @@ -201,7 +201,7 @@ class QuantFP8(CustomOp): if scale is None: if self.group_shape == GroupShape.PER_TOKEN: - x_max, _ = x.abs().max(dim=-1) + x_max = x.abs().amax(dim=-1) x_max = x_max.unsqueeze(-1).to(torch.float32) if scale_ub is not None: x_max = x_max.clamp(max=scale_ub)