[Bugfix]: Fix unquantized gpt-oss weight loading broken by FusedMoE r… (#45818)

Signed-off-by: priyansh jain <priyansh.jain2@amd.com>
Co-authored-by: Li, Jiang <jiang1.li@intel.com>

vllm/model_executor/models/gpt_oss.py

@@ -979,7 +979,11 @@ class GptOssModel(nn.Module, EagleModelMixin):
         tp_rank_start = tp_rank * per_rank_intermediate_size
         tp_rank_end = min((tp_rank + 1) * per_rank_intermediate_size, intermediate_size)
 
-        for name, weight in weights:
+        # Use centralized weight remapping for MoE expert parameters.
+        # The FusedMoE refactor moved expert params under
+        # `mlp.experts.routed_experts.*`; this remaps checkpoint names so
+        # MoE weight/bias keys resolve against params_dict.
+        for name, weight in remap_moe_expert_weights(weights, params_dict):
             # Skip layers on other devices.
             if is_pp_missing_parameter(name, self):
                 continue