======================================================================
MODEL 4.2: XLM-RoBERTa with Data Augmentation
======================================================================
Date: 2025-12-25 16:18:05

SPECIFICATION:
  Model: xlm-roberta-large
  Batch Size: 8 (Grad Accum: 2)
  Learning Rate: 1e-05
  Max Length: 128
  Epochs: 3/5
  Seed: 42
  Augmentation: Yes (Target minority 5%)

DATA AUGMENTATION:
  Original train size: 15699
  Augmented train size: 16156
  Samples added: 457

TEST RESULTS:
  Macro F1: 0.6822
  Weighted F1: 0.7843
  Accuracy: 0.7832

TRAINING TIME:
  Total: 81.88 minutes
  Per epoch: 982.5s