fix: load_return

modeling_hf_nomic_bert.py

@@ -458,7 +458,9 @@ class NomicBertPreTrainedModel(PreTrainedModel):
                 state_dict = filter_shapes(state_dict, model)
 
             load_return = model.load_state_dict(state_dict, strict=strict)
-        logger.warning(load_return)
+        # moe models load new weights 
+        if getattr(config, "moe_top_k", 0) == 0:
+            logger.warning(load_return)
         return model
 
     def _set_gradient_checkpointing(self, module, value=False):