bug fixed: optimizer

mlora.py

@@ -171,8 +171,6 @@ def train(config: Dict[str, any], llm_model: mlora.LLMModel, dispatcher: mlora.D
     step_cnt = 0
     while not dispatcher.check_task_done():
         input: mlora.MultiLoraBatchData = dispatcher.get_train_data()
-        for lora in input.lora_batch_data_config_:
-            all_optimizer[lora.adapter_name_].zero_grad()
 
         step_cnt += 1
 
@@ -201,6 +199,7 @@ def train(config: Dict[str, any], llm_model: mlora.LLMModel, dispatcher: mlora.D
         for lora in input.lora_batch_data_config_:
             if step_cnt % accumulation_step[lora.adapter_name_] == 0:
                 all_optimizer[lora.adapter_name_].step()
+                all_optimizer[lora.adapter_name_].zero_grad()
 
         if step_cnt % config["save_step"] == 0:
             mlora.save_lora_model(llm_model, config, f"{step_cnt}")