From ca179d0f64743537a430631e7fc79405ec2887cb Mon Sep 17 00:00:00 2001 From: Tmn07 Date: Fri, 2 Jan 2026 13:13:22 +0800 Subject: [PATCH] [Bugfix] Fix activation quantization for compressed-tensors W4A16 (#31572) Signed-off-by: Tmn07 --- .../compressed_tensors/schemes/compressed_tensors_wNa16.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_wNa16.py b/vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_wNa16.py index 3f1b4d883..f8b29041e 100644 --- a/vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_wNa16.py +++ b/vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_wNa16.py @@ -114,7 +114,7 @@ class CompressedTensorsWNA16(CompressedTensorsScheme): logger.info("Using %s for CompressedTensorsWNA16", kernel_type.__name__) self._kernel_backends_being_used.add(kernel_type.__name__) - if isinstance(kernel_type, MarlinLinearKernel): + if kernel_type is MarlinLinearKernel: input_dtype = get_marlin_input_dtype(self.layer_name) if input_dtype is not None: mp_linear_kernel_config.act_type = input_dtype