Fixed Ruff check and small redme update

README.md

@@ -30,7 +30,7 @@ Trained for 8 epochs on an RTX 4070 12GB (~9 hours) with BFloat16 mixed precisio
 
 ## Key Findings
 
-From the [research paper](docs/research_paper.tex):
+From my research paper:
 
 - **Naive MTL produces mixed results**: topic classification benefits (+3.7% accuracy), but emotion detection suffers negative transfer (−0.02 F1) under mean pooling with round-robin scheduling.
 - **Learned attention pooling + temperature sampling eliminates negative transfer entirely**: emotion F1 improves from 0.199 → 0.352 (+77%), surpassing the single-task baseline (0.218).

scripts/train_bert_baseline.py

@@ -57,7 +57,6 @@ from src.data.dataset import (
     load_emotion_jsonl,
     load_topic_jsonl,
 )
-
 from src.training.metrics import (
     bootstrap_confidence_interval,
     multilabel_f1,
@@ -67,7 +66,6 @@ from src.training.metrics import (
     tune_per_class_thresholds,
 )
 
-
 # Configuration
 
 @dataclass
@@ -439,7 +437,6 @@ class BertTrainer:
         self.optimizer.zero_grad()
 
         epoch_losses: Dict[str, List[float]] = {t: [] for t in self.train_loaders}
-        epoch_metrics: Dict[str, List[float]] = {}
 
         if len(self.train_loaders) > 1:
             # Multi-task: temperature sampling
@@ -951,12 +948,12 @@ def run_experiment(mode: str, config: BertBaselineConfig) -> Dict[str, Any]:
     # Load best checkpoint for final evaluation
     best_path = config.checkpoint_dir / mode / "best.pt"
     if best_path.exists():
-        print(f"\n  Loading best checkpoint for final evaluation...")
+        print("\n  Loading best checkpoint for final evaluation...")
         checkpoint = torch.load(best_path, map_location=device, weights_only=False)
         model.load_state_dict(checkpoint["model_state_dict"])
 
     # Full evaluation
-    print(f"\n  Running final evaluation...")
+    print("\n  Running final evaluation...")
     eval_results = evaluate_bert_model(
         model,
         val_loaders,