trainyourcomputer.com
Train Your Computer
Learn to train models. Start with problems that matter in cancer.
Promised labs
-
Histopathology patches
Train a classifier on a public patch set. Publish the notebook, the split, and the mistakes — not just the accuracy.
-
Immuno-oncology Q&A, with citations
LoRA a small model on public IO questions. Require sources. Score it on whether it invents papers.
-
Guideline RAG, then catch the lies
Retrieve over one open guideline. Measure hallucination against the source text. Keep the eval when it fails.
Labs are not published yet. When they ship, they will land here as notebooks and write-ups.
Work in public on X: @train_computer