{
  "python": "3.12.14",
  "torch": "2.6.0+cpu",
  "device": "cpu",
  "threads": 1,
  "seed": 1337,
  "unique_parameters": 7056,
  "tied_storage": true,
  "transpose_import_error": 0.0,
  "attention_forward_max_error": 4.440892098500626e-16,
  "attention_grad_max_error": 4.884981308350689e-15,
  "weighted_accumulation_error": 5.551115123125783e-17,
  "wrong_equal_microbatch_error": 0.05279916199372996,
  "ddp_average_math_error": 5.551115123125783e-17,
  "batch_shifts": [
    {
      "rank": 0,
      "x": [
        [
          0,
          1,
          2
        ],
        [
          3,
          4,
          5
        ]
      ],
      "y": [
        [
          1,
          2,
          3
        ],
        [
          4,
          5,
          6
        ]
      ],
      "next_start": 12
    },
    {
      "rank": 1,
      "x": [
        [
          6,
          7,
          8
        ],
        [
          9,
          10,
          11
        ]
      ],
      "y": [
        [
          7,
          8,
          9
        ],
        [
          10,
          11,
          12
        ]
      ],
      "next_start": 18
    }
  ],
  "schedule": {
    "0": 0.00015,
    "3": 0.0006,
    "4": 0.0006,
    "12": 0.00032999999999999994,
    "20": 6e-05,
    "21": 6e-05
  },
  "completion_score_sum_argmin": 0,
  "completion_score_mean_argmin": 1,
  "shifted_completion_scores": [
    0.026594985342392085,
    6.387977430713088
  ],
  "wrong_unshifted_mask_scores": [
    0.026594985342392193,
    1.6094379124341003
  ],
  "prompt_loss_invariance_error": 0.0,
  "causality_prefix_error": 0.0,
  "checkpoint_resume_max_error": 0.0,
  "toy_training": {
    "task": "next token is (current+1) modulo 17, fixed batch",
    "initial_loss": 2.8402984142303467,
    "final_post_update_loss": 0.5357053875923157,
    "history": [
      {
        "step": 0,
        "pre_update_loss": 2.8402984142303467,
        "pre_clip_norm": 0.9957486987113953
      },
      {
        "step": 9,
        "pre_update_loss": 2.5020911693573,
        "pre_clip_norm": 1.0055875778198242
      },
      {
        "step": 49,
        "pre_update_loss": 1.5028687715530396,
        "pre_clip_norm": 0.7711174488067627
      },
      {
        "step": 99,
        "pre_update_loss": 0.5490099787712097,
        "pre_clip_norm": 0.5083446502685547
      }
    ]
  },
  "residual_noise_std_unscaled": 9.42806195885145,
  "residual_noise_std_scaled": 0.9428061958851452,
  "notebook_mlp_equal_accumulation_error": 1.6653345369377348e-16,
  "passed": true
}