01
To do
Upload dataset
- CSV
- Comma, semicolon, tab or pipe separated
- UTF-8 or Windows-1252
CSV only. Size limits are shown once you are signed in. Files are stored in your private workspace.
No dataset uploaded.
Dashboard
Ready
Dataset—
Target—
Training run—
Active model—
Dashboard
Your projects, training runs and saved models.
Get started
Create your first project
Every dataset, training run and model version belongs to a project. Create one, then upload a CSV and follow the workflow to a model you can predict with.
- 1Create a project
- 2Upload a dataset
- 3Profile, choose a target, validate
- 4Automatic preprocessing
- 5Train and compare models
- 6Evaluate the one you choose
- 7Use a version for prediction
Recent training runs
Loading…
Recent projects
Loading…
Projects
Each project keeps its own training runs, model versions and active model.
Saved projects
0Loading projects…
Project overview
No project selected
Choose a project to see its overview.
Workflow
From a dataset to an evaluated, versioned model.
Project required
Create a project to get started
Every dataset and model belongs to a project. Create or select a project before uploading a dataset; its training runs, model versions and active model stay together.
02
Locked
Dataset profile
Column types, missing values, duplicates and potential targets.
No profile generated yet.
Column details
| Column | Type | Missing | Unique | ID-like | Target candidate |
|---|
03
Locked
Select target
Only numeric columns with enough distinct values can be regression targets. Review the candidates and choose one explicitly.
Upload a dataset to find target candidates.
No target selected.
04
Locked
Dataset validation
Checks the target and the features. Leave the feature list empty to use every column except the target.
No validation run yet.
05
Locked
Automatic preprocessing
Feature typenumeric and categorical columns get their own handling
Missing valuesimputed from the training rows only
Skewnesspower transforms considered for strongly skewed columns
Cardinalityone-hot for few levels, cross-fitted target encoding for many
Model familyscaling for linear and distance models, none for trees
Cross-validationeach model's pipeline is chosen by 5-fold CV
The fitted preprocessing is saved inside every model version and reused unchanged for evaluation and prediction.
Validate the dataset before preprocessing.
06
Locked
Baseline training
Every selected model is trained with its default settings on the same train/test split. Each model gets preprocessing suited to it, chosen by cross-validation on the training rows. No hyperparameter tuning.
Models to train
Loading available models…
No training run yet.
07
Locked
Model comparison
Train baseline models to compare them here.
| Choose | Model | Test R² | RMSE | MAE | CV R² | CV std | Training time | Preprocessing | Status |
|---|
Test R²/RMSE/MAE: held-out test rows. CV R²: 5-fold cross-validation on training rows only. Each model uses its own preprocessing, chosen by CV. The recommendation uses CV only — the test set is never used to choose — and the final choice is yours.
Selected model
None — choose a row above
08
Locked
Final model evaluation
- Selected model
- None — choose one in step 07
- Training run
- —
- Evaluation status
- Not ready
Choose a model in step 07 and evaluate it.
Held-out test set
Cross-validation training rows only
Saved model version · next step
—
Training runs
Every run keeps its own models, metrics and recommendation. Open a run to compare its models.
Loading…
Models
Every saved version, grouped by the run that trained it. The active version is the one used for prediction.
Active model
No active model for this project. Choose a version in the registry below.
09
Model registry
Versions are never overwritten. Use a version for prediction to make it the project's active model.
Select a project to see its saved versions.
Prediction
Predict with the active model version, its saved preprocessing and its feature schema. Nothing is trained here.
10
Locked
Make a prediction
Choose a project to see its active model.
Input features
Select a model version to load its input features.
No prediction yet. Fill in the features and run a prediction.
Settings
Your account, session and the limits configured on this server.
Account
- Signed in as
- —
- Name
- —
- Google sign-in
- —
To change your password, sign out and use “Forgot password?” on the sign-in page. Changing it signs you out on every device.
Session
Your session renews automatically while you use InferLabs. It ends when you sign out, after a period of inactivity, or when it reaches its maximum lifetime. A temporary network problem never signs you out.
- Current access token renews before
- —
Dataset limits
- Limits
- Loading…
About
InferLabsVersion 1
- API reference
- Interactive API documentation