Lightning Talks 2
summary
Short lightning slate: concrete claims and tools, not platitudes. Expect Moulding Data for ML: three cycles and bias corrections, more data ≠ better data, a demo of pytest parametrize vs unittest subTest to deduplicate tests, a tool called commit canvas predicts conventional-commit labels with an ML model (project-specific training, mid‑June 2021 release), a Reddit‑based sarcasm classifier using embeddings and a small neural net, Jupyter magic tips (%store, ?doc, post‑mortem pdb), api‑flask (Flask+Marshmallow auto‑validation/docs), direnv per‑dir envs, and all‑repos batching that updates ~300 repos and opens PRs.
AI-written summary · about
chapters (13)
- 0:00Lightning talks kickoff
- 0:51Data for ML models
- 6:09Paikolatam conference intro
- 10:28Python Chile initiatives
- 15:42Parameterizing tests
- 20:24Commit messages automation
- 26:22Sarcasm detection in text
- 32:34Advent of Code tips
- 38:15Python f-strings mastery
- 41:55FastAPI vs Flask
- 47:25Direnv per-directory config
- 53:05Managing Git repos
- 57:46Closing remarks
