HyperTunner/default.yaml at main · info-gallary/HyperTunner · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
# CNN HyperTuner Configuration
# Use this file to customize your tuning run.
# Pass with: python run_tuner.py --config configs/default.yaml

experiment:
  name: "cnn_tune_experiment"
  output_dir: "./experiments"
  seed: 42
  verbose: true

dataset:
  name: "cifar10"          # "cifar10", "mnist", or "custom"
  data_root: "./data"
  n_epochs_per_eval: 5     # Training epochs per config evaluation
  device: null             # null = auto-detect (cuda > mps > cpu)

# ─── Algorithm Selection ───────────────────────────────────────────────────
# Choose one: "ga", "pso", "sa"
algorithm: "pso"
n_iterations: 30

# ─── Algorithm-Specific Settings ───────────────────────────────────────────

genetic_algorithm:
  population_size: 20
  crossover_rate: 0.85
  mutation_rate: 0.15
  tournament_size: 3
  elitism_count: 2
  blx_alpha: 0.5
  mutation_sigma: 0.1
  adaptive_mutation: true
  stagnation_restart_after: 20

particle_swarm:
  n_particles: 25
  w_max: 0.9
  w_min: 0.4
  c1: 2.0
  c2: 2.0
  v_max_fraction: 0.2
  topology: "global"        # "global" or "ring"
  ring_neighbors: 2
  stagnation_restart_after: 20

simulated_annealing:
  T0: 1.0
  T_min: 0.00001
  alpha: 0.97
  perturbation: "gaussian"  # "gaussian" or "cauchy"
  perturbation_scale: 0.1
  reheat_factor: 2.0
  stagnation_reheat_after: 30

# ─── Search Space Overrides ────────────────────────────────────────────────
# Leave empty to use defaults. Override specific params only.
search_space:
  # Example overrides:
  # learning_rate: {type: float, low: 1e-4, high: 1e-2, log_scale: true}
  # batch_size: {type: categorical, choices: [32, 64, 128]}