rgou-cloudflare/evaluation_results.json at main · tre-systems/rgou-cloudflare · 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
{
  "model_path": "ml/data/weights/ml_ai_weights_pytorch_v5.json",
  "num_games": 100,
  "evaluation_time_seconds": 15.18,
  "ml_wins": 49,
  "classic_wins": 51,
  "draws": 0,
  "ml_win_rate": 0.49,
  "classic_win_rate": 0.51,
  "draw_rate": 0.0,
  "avg_moves_per_game": 166.6,
  "games_per_second": 6.59,
  "ml_avg_time_per_move_ms": 0.7,
  "classic_avg_time_per_move_ms": 0.2,
  "ml_total_moves": 8628,
  "classic_total_moves": 8303,
  "ml_wins_playing_first": 19,
  "ml_wins_playing_second": 30,
  "performance_rating": "⭐⭐⭐⭐⭐",
  "model_rank": 1,
  "notes": "PyTorch V5 model shows excellent performance with 49% win rate against expectiminimax AI and 61.7% average win rate against other ML models. Ready for production use.",
  "training_config": {
    "epochs": 2000,
    "games_per_epoch": 100,
    "learning_rate": 0.001,
    "batch_size": 64,
    "framework": "PyTorch",
    "gpu_acceleration": "MPS"
  },
  "individual_games": [
    {
      "game_id": 0,
      "winner": "Classic",
      "moves_played": 165,
      "p1_finished": 7,
      "p2_finished": 7
    }
  ]
}