Update README.md

README.md

@@ -102,7 +102,7 @@ The training data was aggregated from multiple sources:
 - **Mixed Precision Training:** Native AMP
 
 **Label Imbalance Correction:**  
-A correction factor was computed for each topic based on the number of non-"unknown" samples to mitigate label imbalance. The correction weight for each topic was calculated as:  
+A correction factor was computed for each topic based on the number of non-'unknown' samples to mitigate label imbalance. The correction weight for each topic was calculated as:  
     **weight = (average non-unknown count) / (non-unknown count for the topic)**
 
 ---
@@ -152,13 +152,13 @@ After hyperparameter search, update your training arguments using the best hyper
 
 ```python
 # After hyperparameter search:
-best_run = trainer.hyperparameter_search(direction="maximize", hp_space=hp_space, n_trials=5)
-print("Best hyperparameters:", best_run.hyperparameters)
+best_run = trainer.hyperparameter_search(direction='maximize', hp_space=hp_space, n_trials=5)
+print('Best hyperparameters:', best_run.hyperparameters)
 
 # Update TrainingArguments accordingly
-training_args.learning_rate = best_run.hyperparameters["learning_rate"]
-training_args.num_train_epochs = best_run.hyperparameters["num_train_epochs"]
-training_args.gradient_accumulation_steps = best_run.hyperparameters["gradient_accumulation_steps"]
+training_args.learning_rate = best_run.hyperparameters['learning_rate']
+training_args.num_train_epochs = best_run.hyperparameters['num_train_epochs']
+training_args.gradient_accumulation_steps = best_run.hyperparameters['gradient_accumulation_steps']
 
 # Reinitialize the Trainer with the updated arguments
 trainer = CustomTrainer(