Overview

Configuration presets and custom settings for optimization behavior, device usage, and performance tuning.

DefaultConfigs

Pre-configured optimization settings for common scenarios:

python 
from MAT_HPO_LIB.utils import DefaultConfigs

# Quick test (10 steps, fast evaluation)
config = DefaultConfigs.quick_test()

# Standard optimization (100 steps, balanced)  
config = DefaultConfigs.standard()

# CPU-only mode (no GPU usage)
config = DefaultConfigs.cpu_only()

⚙️ Custom Configuration

python 
from MAT_HPO_LIB.utils import OptimizationConfig

config = OptimizationConfig(
    max_steps=50,              # Number of optimization steps
    batch_size=32,             # Agent batch size
    use_cuda=True,             # Enable GPU acceleration  
    gpu_device=0,              # GPU device ID
    verbose=True,              # Show progress
    save_interval=10,          # Save every N steps
    early_stop_patience=15,    # Stop if no improvement
    output_dir="./results"     # Results directory
)

Configuration Options

Optimization Control

  • max_steps: Total optimization steps
  • early_stop_patience: Early stopping patience
  • convergence_threshold: Convergence criteria

Performance Settings

  • use_cuda: Enable GPU acceleration
  • gpu_device: GPU device selection
  • batch_size: Agent training batch size

Output & Logging

  • verbose: Progress display
  • save_interval: Checkpoint frequency
  • output_dir: Results directory

Preset Comparison

Setting quick_test() standard() cpu_only()
max_steps 10 100 100
use_cuda Auto-detect Auto-detect False
batch_size 16 32 32
Use Case Testing setup Production runs CPU-only systems

Usage Example

python 
# Choose based on your needs:

# 1. Quick testing/debugging
config = DefaultConfigs.quick_test()

# 2. Standard optimization run  
config = DefaultConfigs.standard()

# 3. Custom configuration for your specific needs
config = OptimizationConfig(
    max_steps=200,           # Longer optimization
    use_cuda=True,          
    gpu_device=1,           # Use GPU 1
    verbose=True,           # Show progress
    output_dir="./my_results"
)

# Use with optimizer
optimizer = MAT_HPO_Optimizer(environment, space, config)
results = optimizer.optimize()