From 9fac6aa30b669de75d8718164cd99676d3530e7d Mon Sep 17 00:00:00 2001 From: Lucas Wilkinson Date: Thu, 18 Sep 2025 17:26:28 -0400 Subject: [PATCH] [BugFix] Fix DeepGEMM warmup, no m.weight_scale_inv (#25206) Signed-off-by: Lucas Wilkinson --- vllm/model_executor/warmup/deep_gemm_warmup.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm/model_executor/warmup/deep_gemm_warmup.py b/vllm/model_executor/warmup/deep_gemm_warmup.py index a636a714145cf..4d1829cd228cd 100644 --- a/vllm/model_executor/warmup/deep_gemm_warmup.py +++ b/vllm/model_executor/warmup/deep_gemm_warmup.py @@ -36,7 +36,7 @@ def _extract_data_from_linear_base_module( assert m.quant_method.quant_config is not None w = m.weight - ws = m.weight_scale_inv + ws = m.weight_scale quant_block_size = m.quant_method.quant_config.weight_block_size assert isinstance(w, torch.Tensor)