Worked on the pytorch/torchtune repository to deliver flexible training enhancements focused on deep learning model robustness and efficiency. Developed and integrated Early Exit Loss and Layer Dropout features into the model training process, providing configurable fine-tuning options for diverse deployment scenarios. The implementation involved creating new modules and utilities in Python using PyTorch, ensuring seamless integration with existing training pipelines. Emphasis was placed on maintainability and extensibility, laying groundwork for future improvements. No major bugs were addressed during this period, as the primary focus remained on feature development and enhancing the flexibility of machine learning model training strategies.