[Quantization] Enable compressed-tensors AWQ for Turing GPU (#29732)

Isotr0py · web-flow · commit e1464c3a0861 · 2025-11-30T06:04:28.000Z
Signed-off-by: Isotr0py &lt;mozf@mail2.sysu.edu.cn&gt;
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
@@ -79,8 +79,8 @@ def __init__(
 
     @classmethod
     def get_min_capability(cls) -> int:
-        # ampere and up
-        return 80
+        # Turing and up
+        return 75
 
     def create_weights(
         self,
