From c5e8330997dc3969818c6696a79820bcee44a702 Mon Sep 17 00:00:00 2001 From: Cheng Li Date: Mon, 22 Jul 2024 19:25:05 -0700 Subject: [PATCH] [Bugfix] Fix null `modules_to_not_convert` in FBGEMM Fp8 quantization (#6665) --- vllm/model_executor/layers/quantization/fbgemm_fp8.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm/model_executor/layers/quantization/fbgemm_fp8.py b/vllm/model_executor/layers/quantization/fbgemm_fp8.py index e8456471..6b329231 100644 --- a/vllm/model_executor/layers/quantization/fbgemm_fp8.py +++ b/vllm/model_executor/layers/quantization/fbgemm_fp8.py @@ -31,7 +31,7 @@ class FBGEMMFp8Config(QuantizationConfig): """Config class for FBGEMM Fp8.""" def __init__(self, ignore_list: List[str], input_scale_ub: float): - self.ignore_list = ignore_list + self.ignore_list = ignore_list if ignore_list else [] self.input_scale_ub = input_scale_ub # For GPUs that lack FP8 hardware support, we can leverage the Marlin