[Quant] [Bugfix] Fix quantization config matching with hf_to_vllm_mapper (#20046)

2025-07-01 06:20:34 -04:00
parent c05596f1a3
commit 9025a9a705
17 changed files with 107 additions and 29 deletions
--- a/vllm/model_executor/layers/quantization/torchao.py
+++ b/vllm/model_executor/layers/quantization/torchao.py
@@ -55,6 +55,7 @@ class TorchAOConfig(QuantizationConfig):
            os.environ["VLLM_DISABLE_COMPILE_CACHE"] = "1"
            logger.info("Using TorchAO: Setting VLLM_DISABLE_COMPILE_CACHE=1")
        """
+        super().__init__()
        self.torchao_config = torchao_config
        self.skip_modules = skip_modules or []