Skip to main content
This page provides a comprehensive reference for all CLI arguments available in ML Experiment Autopilot.

​
Global Options

These options are available for the main autopilot command:
​
--version
boolean
Display the version number and exit.Short form: -VExample:
autopilot --version
# Output: ML Experiment Autopilot v0.1.0
​
--help
Show help message with available commands and exit.Example:
autopilot --help

​
Run Command Arguments

All arguments for the autopilot run command.

​
Required Arguments

​
--data
path
required
Path to dataset file (CSV or Parquet format).Short form: -dValidation:
  • File must exist
  • File must be readable
  • Must be a file (not a directory)
Supported formats:
  • .csv - Comma-separated values
  • .parquet - Apache Parquet
Examples:
# CSV file
autopilot run --data ./data/train.csv -t price --task regression

# Parquet file
autopilot run -d ~/datasets/housing.parquet -t SalePrice --task regression

# Absolute path
autopilot run --data /home/user/data/customers.csv -t churned --task classification
​
--target
string
required
Name of the target column for prediction.Short form: -tValidation:
  • Column must exist in the dataset
  • Column name is case-sensitive
Examples:
# Simple column name
autopilot run -d data.csv --target price --task regression

# Column with spaces (use quotes)
autopilot run -d data.csv --target "Sale Price" --task regression

# Case-sensitive matching
autopilot run -d data.csv --target SalePrice --task regression
​
--task
enum
required
Type of machine learning task.Allowed values:
  • classification - Categorical target variable
  • regression - Continuous target variable
Examples:
# Classification
autopilot run -d data.csv -t churned --task classification

# Regression
autopilot run -d data.csv -t price --task regression

​
Optional Arguments

​
--constraints
path
Path to constraints file (Markdown format).Short form: -cDefault: NoneValidation:
  • File must exist if specified
  • File must be readable
  • Must be a file (not a directory)
Examples:
# With constraints
autopilot run -d data.csv -t price --task regression -c constraints.md

# Without constraints
autopilot run -d data.csv -t price --task regression
See Constraints for file format details.
​
--max-iterations
integer
Maximum number of experiment iterations.Short form: -nDefault: 20Range: 1-100Validation:
  • Must be at least 1
  • Cannot exceed 100
Examples:
# Default (20 iterations)
autopilot run -d data.csv -t price --task regression

# Custom iterations
autopilot run -d data.csv -t price --task regression --max-iterations 50

# Short form
autopilot run -d data.csv -t price --task regression -n 30

# Minimum value
autopilot run -d data.csv -t price --task regression -n 1

# Maximum value
autopilot run -d data.csv -t price --task regression -n 100
​
--time-budget
integer
Time budget in seconds for the experiment session.Default: 3600 (1 hour)Range: 60-86400 (1 minute to 24 hours)Validation:
  • Minimum: 60 seconds (1 minute)
  • Maximum: 86400 seconds (24 hours)
Examples:
# Default (1 hour)
autopilot run -d data.csv -t price --task regression

# 30 minutes
autopilot run -d data.csv -t price --task regression --time-budget 1800

# 2 hours
autopilot run -d data.csv -t price --task regression --time-budget 7200

# 6 hours
autopilot run -d data.csv -t price --task regression --time-budget 21600

# 24 hours (maximum)
autopilot run -d data.csv -t price --task regression --time-budget 86400
Common durations:
  • 30 minutes: 1800
  • 1 hour: 3600
  • 2 hours: 7200
  • 4 hours: 14400
  • 8 hours: 28800
  • 12 hours: 43200
  • 24 hours: 86400
​
--output-dir
path
Directory where results will be saved.Short form: -oDefault: Auto-generated timestamped directory in ./outputs/Auto-generated format: ./outputs/experiment_YYYYMMDD_HHMMSS/Examples:
# Auto-generated output directory
autopilot run -d data.csv -t price --task regression
# Creates: ./outputs/experiment_20240315_143022/

# Custom output directory
autopilot run -d data.csv -t price --task regression -o ./my_experiment

# Absolute path
autopilot run -d data.csv -t price --task regression --output-dir /home/user/results

# Relative path
autopilot run -d data.csv -t price --task regression -o ../results/housing_exp
The directory will be created if it doesn’t exist.
​
--verbose
boolean
Show detailed Gemini reasoning and decision-making process.Short form: -vDefault: falseExamples:
# Normal output
autopilot run -d data.csv -t price --task regression

# Verbose output
autopilot run -d data.csv -t price --task regression --verbose

# Short form
autopilot run -d data.csv -t price --task regression -v
Verbose output includes:
  • Gemini’s reasoning for each decision
  • Detailed hypothesis generation process
  • Full experiment planning steps
  • Extended error messages and stack traces
​
--resume
path
Resume from a previously saved state file.Default: NoneValidation:
  • File must exist
  • File must be readable
  • Must be a valid state file (JSON format)
Examples:
# Resume from auto-generated state file
autopilot run --resume ./outputs/experiment_20240315_143022/state.json

# Resume from custom output directory
autopilot run --resume ./my_experiment/state.json
When using --resume, all other arguments are ignored as they are loaded from the state file.

​
Argument Combinations

​
Quick Start

Minimal required arguments: 
autopilot run -d train.csv -t price --task regression

​
Production Run

All common arguments for a production scenario: 
autopilot run \
  --data production_data.parquet \
  --target conversion_rate \
  --task regression \
  --constraints prod_constraints.md \
  --max-iterations 100 \
  --time-budget 28800 \
  --output-dir ./prod_run_$(date +%Y%m%d) \
  --verbose

​
Short Forms

Using all short-form arguments: 
autopilot run -d data.csv -t price --task regression -c constraints.md -n 30 -o ./results -v

​
Help Commands

Get help at any level: 
# Main help
autopilot --help

# Run command help
autopilot run --help

​
Terminal Output Example

When you run a command, you’ll see output like this: 
$ autopilot run -d housing.csv -t SalePrice --task regression -n 25 -v

╭───────────────────────────────────────────╮
│  ML Experiment Autopilot v0.1.0           │
│  Powered by Google Gemini                 │
╰───────────────────────────────────────────╯

Configuration:
  Data: housing.csv
  Target: SalePrice
  Task: regression
  Max Iterations: 25
  Time Budget: 3600s (1h 0m)
  Constraints: None
  Verbose: Yes

[1/6] Profiling dataset...
✓ Dataset profiling complete (1461 rows, 81 columns)

[2/6] Designing experiments...
✓ Experiment plan created (5 initial experiments)

...

​
See Also

Build docs developers (and LLMs) love

Get started for freeTalk to us