This dataset was created as part of an MSc Computer Science dissertation focused on supporting learners during the transition from block-based programming (such as Blockly) to text-based Python coding.
The dataset is designed for instruction tuning of lightweight language models, particularly Microsoft Phi-2, to act as an educational assistant for beginner programmers rather than a full code generator.
It… See the full description on the dataset page: https://huggingface.co/datasets/uma-siddareddy/phi2-block-to-text-python-education-dataset.
