save processing class correctly

Author: deep1401

transformerlab/plugins/dllm_server/setup.sh

@@ -5,3 +5,4 @@ cd dllm
 uv pip install -e .
 uv pip install "transformers>4.57.0"
 
+

transformerlab/plugins/dllm_trainer_multi_gpu/main.py

@@ -338,7 +338,7 @@ def convert_to_messages(row):
             adaptor_output_dir = getattr(tlab_trainer.params, "adaptor_output_dir", None)
             if adaptor_output_dir:
                 trainer.save_model(output_dir=adaptor_output_dir)
-                trainer.processing_class.save_pretrained(output_dir=adaptor_output_dir)
+                trainer.processing_class.save_pretrained(adaptor_output_dir)
         else:
             if "/" in model_id:
                 model_id_short = model_id.split("/")[-1]
@@ -351,7 +351,7 @@ def convert_to_messages(row):
 
             trainer.save_model(output_dir=fused_model_location)
             if hasattr(trainer, "processing_class") and trainer.processing_class:
-                trainer.processing_class.save_pretrained(output_dir=fused_model_location)
+                trainer.processing_class.save_pretrained(fused_model_location)
             model_config = getattr(trainer.model, "config", None)
             if model_config and getattr(model_config, "architectures", None):
                 model_architecture = model_config.architectures[0]