Implementation:Facebookresearch Habitat lab Get config baselines

Knowledge Sources	Habitat-Lab Hydra Documentation
Domains	Configuration_Management, Training
Last Updated	2026-02-15 02:00 GMT

Overview

Concrete function for composing Habitat-Baselines experiment configurations via Hydra YAML composition with optional CLI overrides.

Description

The get_config function in habitat-baselines loads a YAML config file, applies Hydra composition (merging habitat-lab core config with baselines training config), and applies any runtime overrides. It returns a frozen `DictConfig` containing the complete experiment setup: environment settings, task parameters, policy architecture, and training hyperparameters.

Usage

Call this function at the start of any training or evaluation script to load the experiment configuration. Typically invoked via `habitat_baselines.run.execute_exp` or directly in custom training scripts.

Code Reference

Source Location

• Repository: habitat-lab
• File: habitat-baselines/habitat_baselines/config/default.py
• Lines: L27-43

Signature

def get_config(
    config_path: str,
    overrides: Optional[List[str]] = None,
    configs_dir: str = _BASELINES_CFG_DIR,
) -> DictConfig:
    """
    Compose a Habitat-Baselines config from YAML + overrides.

    Args:
        config_path: Relative path to experiment YAML (e.g., "pointnav/ppo_pointnav.yaml")
        overrides: List of Hydra override strings (e.g., ["habitat.seed=1"])
        configs_dir: Base directory for config search (defaults to habitat_baselines/config/)
    Returns:
        DictConfig: Frozen composed configuration
    """

Import

from habitat_baselines.config.default import get_config

I/O Contract

Inputs

Name	Type	Required	Description
config_path	str	Yes	Relative path to experiment YAML config file
overrides	Optional[List[str]]	No	Hydra override strings (e.g., `["habitat.seed=1", "habitat_baselines.num_environments=4"]`)
configs_dir	str	No	Base directory for Hydra config search (defaults to baselines config dir)

Outputs

Name	Type	Description
config	DictConfig	Frozen composed configuration with habitat + baselines settings merged

Usage Examples

Load PointNav PPO Config

from habitat_baselines.config.default import get_config

# Load PointNav PPO training configuration
config = get_config(
    config_path="pointnav/ppo_pointnav.yaml",
    overrides=[
        "habitat.seed=42",
        "habitat_baselines.num_environments=8",
        "habitat_baselines.rl.ppo.num_steps=128",
    ],
)

Related Pages

Implements Principle

• Principle:Facebookresearch_Habitat_lab_Configuration_Composition

Requires Environment

• Environment:Facebookresearch_Habitat_lab_Python_3_9_Core_Dependencies