config_benchmark.yaml

# Benchmark Pipeline Configuration
# This file configures the automated HPO and basic training pipeline

# Mode: "hpo", "basic", or "both"
mode: "both"

# Datasets to run (7 total: 4 ADME + 3 Toxicity)
datasets:
  # ADME (regression)
  - "Half_Life_Obach"
  - "Clearance_Hepatocyte_AZ"
  - "Clearance_Microsome_AZ"
  - "Caco2_Wang"
  # Toxicity (classification)
  - "tox21"
  - "herg"
  - "clintox"

# HPO Configuration
hpo:
  enabled: true
  # Algorithm: single string (e.g., "random"), list (e.g., ["random", "pso"]), or "all" for all algorithms
  algorithm: "all"  # Options: "random", "pso", "ga", "sa", "hc", "abc", or ["random", "pso", ...], or "all"
  trials: 10           # Number of HPO trials per dataset
  epochs: 50           # Training epochs per trial
  patience: 12         # Early stopping patience
  batch_train: 32      # Training batch size
  batch_eval: 64       # Evaluation batch size
  val_fraction: 0.1    # Validation set fraction

# Basic Training Configuration
basic:
  enabled: true
  epochs: 100          # Training epochs
  patience: 20         # Early stopping patience

# General Settings
general:
  seed: 42             # Random seed
  device: "auto"       # Device: "auto", "cpu", or "cuda"
  no_confirm: false    # Skip confirmation prompt
  runs_dir: "runs"     # Directory for HPO results
  reports_dir: "reports" # Directory for benchmark reports

# Report Generation
reporting:
  generate_reports: true
  save_plots: true
  plot_dpi: 300