Navigation

  • Home
  • About Me
  • About My Mentors
  • About My Project
  • My Blog

June 27, 2025

Day 24 – Preparing for model fine-tuning and presentation

By Aayam Raj Shakya

What I Learned

Today, I finished my codebase from last week. I made 2 versions: one with Optuna and another without. Training with Optuna might take over 24 hrs, so I will train over the weekend. I also helped my labmates Ignatius and Michelle fix their code bugs and issues. Mr. Pelumi came back from Canada, and we met him in the morning. I showed him my accomplishments from the past few days, and he was really impressed with my work. Although my current results look perfect, they seem too good to be true.

Fig. 1 - Metrics curve (training phase)
Fig. 2 - Metrics on validation set
Fig. 3 - Confusion matrix for test set

I then worked on further optimizing the Optuna framework; I applied pruning to the codebase, and I will know the results once I start training. Additionally, I worked on the mid-summer presentation slides and the weekly presentation slides and video.

Blockers

N/A

Reflection

Today went well. I finally finished my codebase. Now, the only things left to do are to make a few changes to the ensemble model layer and run the bayesian optimization (hyperparameter tuning). These tasks are simple but time-consuming. Over the weekend, I will resume working on the mid-summer presentation slides and the presentation itself.

Tags: