Fix EPLB + FP4 Quantization Compatibility Issue (#13715)

Co-authored-by: Shu Wang <shuw@nvidia.com>

python/sglang/srt/models/longcat_flash.py

@@ -63,6 +63,7 @@ from sglang.srt.layers.moe.ep_moe.kernels import zero_experts_compute_triton
 from sglang.srt.layers.moe.ep_moe.layer import DeepEPMoE, get_moe_impl_class
 from sglang.srt.layers.moe.fused_moe_triton.layer import FusedMoE
 from sglang.srt.layers.moe.topk import StandardTopKOutput, TopK
+from sglang.srt.layers.moe.utils import filter_moe_weight_param_global_expert
 from sglang.srt.layers.quantization.base_config import QuantizationConfig
 from sglang.srt.layers.quantization.fp8_kernel import is_fp8_fnuz
 from sglang.srt.layers.quantization.fp8_utils import (
@@ -295,6 +296,9 @@ class LongcatFlashMoE(nn.Module):
             x.data
             for name, x in self.experts.named_parameters()
             if name not in ["correction_bias"]
+            and filter_moe_weight_param_global_expert(
+                name, x, self.experts.num_local_experts
+            )
         ]