evaluations: ARC Challenge: 0.2617621899059025 Bangla MMLU: 0.2608813559322034 BoolQ-BN: 0.6064814814814815 CommonsenseQA-BN: 0.3243243243243243 HellaSwag: 0.3220082233282839 MMLU: 0.2706305716856138 OpenBookQA-BN: 0.3219315895372233 PIQA-BN: 0.6050054406964092 TruthfulQA MC1: 0.2548015364916773 step: 110000