{
  "config": {
    "pins": {
      "torch": "2.7.0",
      "unsloth": "2026.4.4",
      "unsloth_zoo": "2026.4.4",
      "trl": "0.21.0",
      "transformers": "4.56.2",
      "vllm": "0.9.2",
      "peft": "0.18.0",
      "accelerate": "1.13.0",
      "bitsandbytes": "0.49.2",
      "torchao": "0.17.0"
    },
    "grpo": {
      "num_generations": 4,
      "max_completion_length": 512,
      "gradient_accumulation_steps": 8,
      "learning_rate": 5e-06,
      "beta": 0.04,
      "num_train_epochs": 1,
      "max_steps": 400,
      "logging_steps": 5,
      "save_steps": 25,
      "eval_steps": 25,
      "lr_scheduler_type": "cosine",
      "warmup_ratio": 0.05
    },
    "sft": {
      "num_train_epochs": 1,
      "per_device_train_batch_size": 4,
      "gradient_accumulation_steps": 4,
      "learning_rate": 2e-05,
      "max_seq_length": 1024
    },
    "task_filter": "action_screen",
    "smoke_steps": 5,
    "abort_step100_min_reward": 0.05,
    "abort_step200_min_reward": 0.7
  },
  "f1_per_tier": {
    "action_screen": {
      "precision": 1.0,
      "recall": 0.9744,
      "f1": 0.987
    },
    "war_room": {
      "precision": 1.0,
      "recall": 0.9624,
      "f1": 0.9808
    },
    "drift_ops": {
      "precision": 0.9205,
      "recall": 0.7864,
      "f1": 0.8482
    }
  },
  "baseline_qwen3_1_7b_zeroshot_f1_per_tier": {},
  "abort_path": null,
  "wall_clock_s": 5078.9,
  "best_checkpoint_step": 350,
  "model_repo": "Elliot89/sentinel-overseer-qwen3-1.7b-grpo400",
  "git_repo": "https://github.com/MrEinsteinE/sentinel-openenv",
  "sentinel_url": "https://elliot89-sentinel.hf.space"
}