Welcome to chAI Cockpit

Multi-User ML Pipeline Interface

🎯 Training Pipeline Configuration

Configure and train machine learning models with automated feature engineering.

Working Directory Configuration
This directory will be used as the base for file downloads and ZIP extraction.
Browse Configuration File
Browse and select a JSON or YAML configuration file. All paths in the config will be treated as absolute paths.
Configuration File Ready
Pipeline Configuration
Select the columns you want to predict (required)
Data Source
CSV, Parquet, or Excel file
YAML dataset specification
Directory for models and outputs
Path for execution logs
Feature Configuration
▼
Specific features to include (leave empty for all)
▼
Features to exclude from training
▼
Features that must always be included
▼
Features excluded from training but used in pipeline
Data Preview
Model Architecture
Model hyperparameters catalog
Available Base Models
Loading models from catalog...
Neural Network Configuration
Neural Network models require more computational resources
Requires CUDA installation
AutoML Configuration
FLAML will automatically select and optimize models
AutoGluon provides advanced automated machine learning
Training Configuration
Number of folds for cross-validation
Parallel processing threads
Optimization Settings
Current: Light
Current: Normal

Enable models cross calibration
Use existing trained model for training modes
Enable range-based predictions
Number of range intervals (2-10)
Comma-separated interval boundaries. Leave empty to use automatic bounds based on Number of Range Bounds.
Current: 0.95

Minimum tolerance prediction for regression
Maximum tolerance prediction for regression
Regression tolerance range split value
Regression tolerance method
Regression target scaling method
Regression metric for model training
Evaluation Configuration
Enable detailed regression debugging
Cost Profile Configuration
Select cost profile for analysis (requires cost profile catalog)
Create comprehensive model analysis
Model Query Configuration
Directory containing trained models
Storage Mode
Persistent: Job files stored on disk (default).
Non-Persistent: Job files stored in memory (tmpfs) for ephemeral execution - faster but data deleted after completion.
Server Status
Connected
Loading server info...
Working Directory
Multi-user mode enabled
Files are browsed using local file service