return empty still if not bit 4

Signed-off-by: Lu Fang <[email protected]>

Author: luccafong

vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe.py

@@ -1566,14 +1566,9 @@ def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
     def get_fused_moe_quant_config(
         self, layer: torch.nn.Module
     ) -> FusedMoEQuantConfig | None:
-        assert self.num_bits == 4 or self.num_bits == 8
-        config_builder = (
-            int4_w4a16_moe_quant_config
-            if self.num_bits == 4
-            else int8_w8a16_moe_quant_config
-        )
-
-        return config_builder(
+        if self.num_bits != 4:
+            return None
+        return int4_w4a16_moe_quant_config(
             w1_scale=layer.w13_weight_scale,
             w2_scale=layer.w2_weight_scale,
             w1_zp=None,
@@ -1586,7 +1581,6 @@ def select_gemm_impl(
         prepare_finalize: mk.FusedMoEPrepareAndFinalize,
         layer: torch.nn.Module,
     ) -> mk.FusedMoEPermuteExpertsUnpermute:
-
         layer.w13_weight = layer.w13_weight_packed
         layer.w2_weight = layer.w2_weight_packed
         assert all([w is not None for w in [layer.w13_weight, layer.w2_weight]])