From 7df0289782ab500b2713b7521979c28de2b21cac Mon Sep 17 00:00:00 2001 From: Michael Goin Date: Tue, 25 Nov 2025 17:52:31 -0500 Subject: [PATCH] Change warning logs to debug for unimplemented MXFP4 Linear/Attention (#29441) Signed-off-by: Michael Goin Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com> --- vllm/model_executor/layers/quantization/mxfp4.py | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/vllm/model_executor/layers/quantization/mxfp4.py b/vllm/model_executor/layers/quantization/mxfp4.py index 198feb03be3e4..d975131f7cff7 100644 --- a/vllm/model_executor/layers/quantization/mxfp4.py +++ b/vllm/model_executor/layers/quantization/mxfp4.py @@ -196,9 +196,10 @@ class Mxfp4Config(QuantizationConfig): # TODO: Add support for MXFP4 Linear Method. # MXFP4 LinearMethod is available in AMD-Quark, refer to that implementation # if you are interested in enabling MXFP4 here. - logger.warning_once( + logger.debug_once( "MXFP4 linear layer is not implemented - falling back to " - "UnquantizedLinearMethod." + "UnquantizedLinearMethod.", + scope="local", ) return UnquantizedLinearMethod() elif isinstance(layer, FusedMoE): @@ -208,9 +209,10 @@ class Mxfp4Config(QuantizationConfig): return Mxfp4MoEMethod(layer.moe_config) elif isinstance(layer, Attention): # TODO: Add support for MXFP4 Attention. - logger.warning_once( + logger.debug_once( "MXFP4 attention layer is not implemented. " - "Skipping quantization for this layer." + "Skipping quantization for this layer.", + scope="local", ) return None