June 12, 2025
Day 13 – Social Enrichment, Games, and Model Training
What I Learned
As usual, Thursday was pretty chill as we spent half the day in the social enrichment program. At first, each group stood up, introduced themselves, and shared their progress in front of Dr. Mack and the audience. For my turn, I shared that I have been working on and training my ensemble model, which I built by stacking MobileNetV2 and EfficientNetV2B3. After all the groups shared their progress, we discussed potential activities for the rest of the social enrichment programs. Most people suggested different places like Six Flags, NASA for trips. I personally suggested that we go to DC to visit the Congress building and other monuments. After this, we played a game called Taboo, where my team 2 won. Unfortunately, in the next game, Jenga, our team lost :(
After the games, we headed to another room where Michael went over some basics of Overleaf, which I was already familiar with, so I decided to help those around me. After this short program, we returned to our respective buildings, had lunch, and got back to work. I continued my work on training my ensemble model for the rest of the day.
Blockers
No issues faced.
Reflection
Even though today was chill, I felt kind of bored during the social enrichment program and really missed the lab work. I was excited to start training the model since I found the solution for GPU issue on the lab PC, so I eagerly wanted to go back and work on it. I also helped Ignatius fix his GPU problem. Now he can train each epoch in 9 mins, compared to his usual 1 hr and 15 mins per epoch. It always feels good to help others. As I’m writing this blog, the model is still training and the accuracy is already showing ~99%.
The training might not stop before 5 PM, but I will update that part in tomorrow’s blog.