🛠️ I have trained a model to predict my blood sugar
Summary
A developer has trained a custom 31,253 parameter encoder-only transformer model to predict blood sugar levels using a synthetic continuous glucose monitor dataset and personal data. The model can predict two hours into the future, infill missing traces, and perform backcasts.
Why it’s interesting
It demonstrates a custom-trained transformer model built for personal continuous glucose monitor data prediction using synthetic pretraining and LoRA.
Source metrics: Points 4 · Comments 2
HN discussion · Project
Source: #HackerNews / Show HN
Summary
A developer has trained a custom 31,253 parameter encoder-only transformer model to predict blood sugar levels using a synthetic continuous glucose monitor dataset and personal data. The model can predict two hours into the future, infill missing traces, and perform backcasts.
Why it’s interesting
It demonstrates a custom-trained transformer model built for personal continuous glucose monitor data prediction using synthetic pretraining and LoRA.
Source metrics: Points 4 · Comments 2
HN discussion · Project
Source: #HackerNews / Show HN