improve the robustness of parsing vlms config in AutoRound (#18894)

Signed-off-by: wenhuach21 <wenhua.cheng@intel.com>
2025-05-30 10:24:47 +08:00
parent 3132290a14
commit 3de3eadf5b
1 changed files with 3 additions and 2 deletions
--- a/vllm/model_executor/layers/quantization/auto_round.py
+++ b/vllm/model_executor/layers/quantization/auto_round.py
@@ -116,8 +116,9 @@ class AutoRoundConfig(QuantizationConfig):

        quantized = True
        if self.block_name_to_quantize:
-            quantized = any(name in layer_name
-                            for name in self.block_name_to_quantize)
+            quantized = any(
+                layer_name.startswith(name)
+                for name in self.block_name_to_quantize)
        elif isinstance(layer, ParallelLMHead):
            quantized = False