[Int4-AWQ] Fix AWQ Marlin check for ROCm (vllm-project#206)

mht-sharma · Sep 27, 2024 · b79f9f4 · b79f9f4
1 parent 9858710
commit b79f9f4
Showing 1 changed file with 9 additions and 3 deletions.
diff --git a/vllm/config.py b/vllm/config.py
@@ -285,9 +285,15 @@ def _verify_quantization(self) -> None:
                 quantization_override = method.override_quantization_method(
                     quant_cfg, self.quantization)
                 if quantization_override:
-                    quant_method = quantization_override
-                    self.quantization = quantization_override
-                    break
+                    if is_hip():
+                        if quantization_override in rocm_supported_quantization:
+                            quant_method = quantization_override
+                            self.quantization = quantization_override
+                            break
+                    else:
+                        quant_method = quantization_override
+                        self.quantization = quantization_override
+                        break
 
             # Verify quantization configurations.
             if self.quantization is None: