Lightweight ML training monitor.
Runlog is an ML training monitor built for how training actually works.
Most monitoring tools assume perfect connectivity, treat terminal logs as an afterthought, and require complex setup before you can see a single metric. Runlog is different.
Every metric and terminal log line — stdout, stderr, framework output — streams live to your dashboard the moment training starts. Three lines of code. No configuration overhead.
Built around offline-first reliability: lose your connection mid-run, start fully offline, or come back online mid-training — not a single metric or log line is ever lost. Data buffers locally and syncs in order the moment you reconnect.
What you get:
Real-time metric streaming with zero polling
Live terminal capture — every print statement and framework log
Offline-first SDK with automatic sync and deduplication
Crash and dead run detection with instant alerts
Pause or stop training remotely from the dashboard
Team workspaces with role-based access control
Cross-run comparison with dynamic metric discovery
Checkpoint ledger with metric snapshots
Publicly shareable run links
Works with PyTorch, HuggingFace Trainer, Keras, XGBoost — anything Python
Currently in beta. 25 spots open, rest join the waitlist.
Loading updates…
Add the badge once and get a single, clean link to your Shipit listing. It helps with SEO and lets visitors discover your product. One snippet: no need to change anything when your standing changes.
The badge updates automatically so it always reflects your current standing. Visitors see your latest status and get a trusted link to your listing. You never have to update the code yourself.
Showing that you're featured or ranked builds trust. Visitors see that real people are engaging with your product, and the badge gives you a credible link back to Shipit.
A single badge looks good on any page and keeps your Shipit presence visible. It's small, loads fast, and gives you one professional link to your listing without clutter.
Sign in to ask the maker a question.
