[Feature] Support AWQ MoE W4A16 Quantization (#142)

Signed-off-by: tangshiwen <tangshiwen@baidu.com> Co-authored-by: Li Wei <liwei.109@outlook.com>
2026-01-26 18:56:05 +08:00
parent 2a998286c0
commit 0711c1abfa
7 changed files with 639 additions and 126 deletions
--- a/vllm_kunlun/ops/fused_moe/layer.py
+++ b/vllm_kunlun/ops/fused_moe/layer.py
@@ -162,7 +162,7 @@ class KunlunFusedMoE(FusedMoE):
        if (self.quant_config is None) or (
            should_ignore_layer(
                prefix,
-                ignore=self.quant_config.ignore,
+                ignore=getattr(self.quant_config, "ignore", tuple()),
                fused_mapping=self.quant_config.packed_modules_mapping,
            )
        ):