fix

llama_lora/lib/finetune.py

@@ -277,10 +277,12 @@ def train(
         model.is_parallelizable = True
         model.model_parallel = True
 
+    # https://huggingface.co/docs/transformers/main/en/main_classes/trainer#transformers.Trainer
     trainer = transformers.Trainer(
         model=model,
         train_dataset=train_data,
         eval_dataset=val_data,
+        # https://huggingface.co/docs/transformers/main/en/main_classes/trainer#transformers.TrainingArguments
         args=transformers.TrainingArguments(
             per_device_train_batch_size=micro_batch_size,
             gradient_accumulation_steps=gradient_accumulation_steps,

llama_lora/ui/finetune_ui.py

@@ -545,8 +545,6 @@ Train data (first 10):
         result_message = f"Training ended:\n{str(train_output)}\n\nLogs:\n{logs_str}"
         print(result_message)
 
-        del base_model
-        del tokenizer
         clear_cache()
 
         return result_message