ataxx-zero / metadata_iter_024.json
dieg0code's picture
Checkpoint iteration 24
014f93e verified
raw
history blame
1.81 kB
{
"iteration": 24,
"timestamp": "2026-02-23T14:12:47.332609+00:00",
"buffer_size": 50000,
"config": {
"iterations": 40,
"episodes_per_iter": 70,
"mcts_sims": 420,
"c_puct": 1.5,
"temp_threshold": 15,
"add_noise": true,
"seed": 42,
"verbose_logs": false,
"episode_log_every": 25,
"epochs": 5,
"batch_size": 96,
"learning_rate": 0.0009,
"weight_decay": 0.0001,
"buffer_size": 50000,
"val_split": 0.1,
"d_model": 128,
"nhead": 8,
"num_layers": 6,
"dim_feedforward": 512,
"dropout": 0.1,
"log_dir": "logs",
"checkpoint_dir": "checkpoints",
"save_every": 3,
"keep_last_n_local_checkpoints": 2,
"keep_last_n_log_versions": 1,
"keep_last_n_hf_checkpoints": 3,
"onnx_path": "ataxx_model.onnx",
"export_onnx": false,
"hf_enabled": true,
"hf_repo_id": "dieg0code/ataxx-zero",
"hf_token_env": "HF_TOKEN",
"hf_local_dir": "hf_checkpoints",
"show_progress_bar": false,
"trainer_log_every_n_steps": 100,
"num_workers": 3,
"persistent_workers": true,
"strict_probs": true,
"trainer_devices": 1,
"trainer_strategy": "auto",
"opponent_self_prob": 0.85,
"opponent_heuristic_prob": 0.12,
"opponent_random_prob": 0.03,
"opponent_heuristic_level": "normal",
"opponent_heuristic_easy_prob": 0.05,
"opponent_heuristic_normal_prob": 0.2,
"opponent_heuristic_hard_prob": 0.75,
"model_side_swap_prob": 0.5,
"eval_enabled": true,
"eval_every": 3,
"eval_games": 12,
"eval_sims": 220,
"eval_heuristic_level": "hard"
},
"stats": {
"replay_size": 50000,
"best_eval_score": 0.0,
"games": 12,
"wins": 0,
"losses": 12,
"draws": 0,
"score": 0.0,
"heuristic_level": "hard",
"sims": 220
}
}