Data from https://www.kaggle.com/datasets/astraszab/facial-expression-dataset-image-folders-fer2013 Final_project is the latest file with cuda/gpu support from colab and 63% acc at 0.00008 l_r and 6 layers at around 70-80 epoch. Next steps: Change model sizing because of inut image size issue(48 vs 284) and experiment again