Mary154/distilbert-emotion-classifier

DistilBERT Emotion Classifier

Fine-tuned DistilBERT model for six-way Twitter emotion classification: sadness, joy, love, anger, fear, and surprise.

Validation metrics:

• Accuracy: 0.9370
• Weighted F1: 0.9371
• Macro F1: 0.9113

Test metrics:

• Accuracy: 0.9230
• Weighted F1: 0.9227
• Macro F1: 0.8767

Training setup: 3 epochs, learning rate 2e-5, batch size 64, weight decay 0.01, fp16 training, and best-checkpoint loading by validation weighted F1.