Fine-tuned BanglaBERT tokenizer for hate speech detection (Subtask 1B)

Fine-tuned BanglaBERT for hate speech detection (Subtask 1B) - Accuracy: 0.7253

initial commit