Update my_model/fine_tuner/fine_tuner.py

my_model/fine_tuner/fine_tuner.py

@@ -16,7 +16,7 @@
 #   OK-VQA dataset.
 
 #   Ensure all dependencies are installed and the required files are in place before running this script.
-#   The configurations for the fine-tuning process are defined in the 'fine_tuning_config.py' file.
+#   The configurations for the fine-tuning process are defined in the 'my_model/config/fine_tuning_config.py' file.
 
 #   ---------- Please run this file for the full fine-tuning process to start ----------#
 #   ---------- Please ensure this is run on a GPU ----------#
@@ -302,6 +302,11 @@ def fine_tune(save_fine_tuned_adapter=False, merge=False, delete_trainer_after_f
     This function encompasses the complete workflow of fine-tuning, including data handling, training,
     and optional steps like saving the fine-tuned model and merging weights.
 
+    The function initiates by preparing the training and evaluation datasets using the `FinetuningDataHandler`.
+    It then sets up the QLoRA configuration for the fine-tuning process. The actual training is carried out by
+    the `Finetuner` class. Post training, based on the arguments, the function can save the fine-tuned model,
+    merge the adapter weights with the base model, and clean up resources by deleting the trainer object.
+
     Args:
         save_fine_tuned_adapter (bool): If True, saves the fine-tuned adapter after training.
         merge (bool): If True, merges the weights of the fine-tuned adapter into the base model.
@@ -311,10 +316,6 @@ def fine_tune(save_fine_tuned_adapter=False, merge=False, delete_trainer_after_f
         The fine-tuned model after the fine-tuning process. This could be either the merged model
         or the trained model based on the provided arguments.
 
-    The function initiates by preparing the training and evaluation datasets using the `FinetuningDataHandler`.
-    It then sets up the QLoRA configuration for the fine-tuning process. The actual training is carried out by
-    the `Finetuner` class. Post training, based on the arguments, the function can save the fine-tuned model,
-    merge the adapter weights with the base model, and clean up resources by deleting the trainer object.
     """
 
     data_handler = FinetuningDataHandler()
@@ -344,4 +345,5 @@ def fine_tune(save_fine_tuned_adapter=False, merge=False, delete_trainer_after_f
 
 
 if __name__ == "__main__":
-    fine_tune()
+    # uncomment to run the fine-tuning process.
+    #fine_tune()