You click “Train.” Then what?

In most ML tools, training is a waiting game. A spinner spins. Maybe a progress bar inches forward. You don’t know what’s happening, how long it will take, or whether something went wrong.

Kanva now shows you the whole process as it unfolds.

Why visibility matters

Training machine learning models takes time. Depending on your data size and the models you’re comparing, it might be seconds or minutes.

During that time, things are happening:

  • Data is being prepared
  • Multiple model types are being trained
  • Each model is being evaluated
  • The best performer is being selected

When you can see this process, you learn:

  • Which model types train fastest
  • Which ones perform best
  • Whether training is actually progressing or stuck
  • When you can expect results

Transparency builds understanding. And understanding builds trust.

What you’ll see

The training progress panel shows:

Overall progress. An animated bar showing how far through the training batch you are. Not just a spinner—actual progress.

Model-by-model updates. As each model type trains, you see it start and complete. LightGBM finished. XGBoost is running. CatBoost is next.

Live metrics. Performance scores appear as each model completes. You can see early results before training finishes.

Time estimates. Based on progress so far, roughly how much longer you’ll wait.

The details

Here’s what happens when you click “Train”:

  1. Preparation. The data pipeline runs—feature engineering, encoding, splitting. You see this stage start and complete.

  2. Baseline models. If enabled, simple baselines train first. These give you a reference point for comparison.

  3. Main models. Each selected model type trains in sequence. You see them start, train, and report results.

  4. Selection. The best model is identified based on your configured metric. Training completes.

Throughout this process, the project shows a “busy” indicator. You can navigate away and come back—the progress panel remembers where you are.

What this changes

Real-time progress changes the training experience:

Less anxiety. You know something is happening. The process feels responsive, not mysterious.

Better intuition. Over time, you learn how long different model types take on your data. You develop expectations.

Earlier feedback. If an early model looks great, you know before training finishes. If everything looks bad, you can investigate sooner.

Team visibility. If you’re sharing your screen or collaborating, everyone sees what’s happening. No more “I think it’s still running?”

A note on SignalR

This feature relies on Kanva’s real-time architecture. The hub server pushes updates to your client as they happen—not on refresh, not on polling, but instantly.

The same infrastructure that powers live collaboration powers training visibility. Your client stays in sync with what the server is doing.

Try it

Next time you train a model, watch the progress panel. You’ll see:

  • The overall progress bar filling
  • Individual models starting and completing
  • Metrics appearing as training proceeds
  • The final results landing

Training is no longer a black box. You see the work as it happens.

Real-time training progress is live now. Questions? Reach out at hello@human-driven.ai