{
  "snapshot_at": "2026-10-23T11:36:27.083966+08:00",
  "full_teacher": {
    "steps": 900,
    "python_s": 669.5377961610211,
    "native_s": 166.75689902703743,
    "speedup": 4.0150530506834645,
    "samples_identical": true,
    "decisions_identical": true
  },
  "full_gpu0_smoke": {
    "steps": 900,
    "total_s": 223.927179,
    "ppo_epochs": 4,
    "minibatches": 36
  },
  "training": [
    {
      "from_episode": 33,
      "to_episode": 34,
      "mean_new_rollout_s": 175.02491699999882,
      "remaining_h": 8.070593395
    },
    {
      "from_episode": 33,
      "to_episode": 34,
      "mean_new_rollout_s": 164.21448999999848,
      "remaining_h": 7.572112594444445
    }
  ],
  "scope": "same physical and reward contracts; remaining training time estimated from resumed rollouts",
  "vectorized": {
    "full_paired_steps": 900,
    "all_step_arrays_equal": true,
    "native_env_state_s": 167.053739,
    "vector_env_state_s": 95.280159,
    "additional_speedup": 1.75328989,
    "gpu0_smoke_s": 146.255596,
    "regression_tests": 98,
    "resume_episode": 33,
    "state": "training",
    "first_update_verified": true
  }
}