logs/trial_0013/run_config.json

{
  "elo_min": null,
  "elo_max": null,
  "max_games": null,
  "val_games": 50000,
  "min_ply": 10,
  "total_steps": 200000,
  "batch_size": 256,
  "lr": 0.0003,
  "weight_decay": 0.0,
  "warmup_frac": 0.05,
  "warmup_steps": null,
  "max_grad_norm": 1.0,
  "patience": null,
  "eval_interval": 2500,
  "log_interval": 100,
  "pause_after_steps": 5000,
  "mate_boost": 1.0,
  "no_outcome_token": false,
  "discard_ply_limit": false,
  "amp_dtype": "bfloat16",
  "no_compile": true,
  "sdpa_math": true,
  "num_workers": 4,
  "device": "cuda",
  "log_dir": "/workspace/logs/trial_0013",
  "hf_repo": null,
  "local_checkpoints": true,
  "resume": null,
  "wandb": false,
  "cache_dir": null,
  "run_type": "pretrain",
  "variant": "base",
  "d_model": null,
  "n_layers": null,
  "n_heads": null,
  "d_ff": null,
  "accumulation_steps": 1,
  "checkpoint_interval": 5000
}