{
  "mode": "categorical",
  "args": {
    "mode": "categorical",
    "seed": 11,
    "modulus": 31,
    "op_family": "const",
    "train_min_len": 1,
    "train_max_len": 4,
    "eval_lengths": "2,4,6",
    "eval_k": "0,1,2,4,6",
    "dim": 128,
    "heads": 4,
    "compiler_layers": 1,
    "num_bases": 16,
    "rank": 16,
    "mem_dim": 48,
    "dropout": 0.05,
    "disable_memory": false,
    "disable_dynamic": false,
    "transition_temperature": 1.0,
    "train_steps": 5,
    "batch_size": 64,
    "eval_batch_size": 64,
    "eval_examples": 128,
    "lr": 0.01,
    "weight_decay": 0.01,
    "max_grad_norm": 1.0,
    "log_every": 1,
    "eval_every": 5,
    "output_dir": "runs_latent_executor/smoke_categorical"
  },
  "train": [
    {
      "step": 1,
      "loss": 1.7917513847351074,
      "pair_acc": 0.0,
      "grad_norm": 0.04268414527177811,
      "transition_entropy": 3.4339401721954346
    },
    {
      "step": 2,
      "loss": 1.926443099975586,
      "pair_acc": 0.015625,
      "grad_norm": 0.04407985508441925,
      "transition_entropy": 3.4339383840560913
    },
    {
      "step": 3,
      "loss": 1.8829529285430908,
      "pair_acc": 0.015625,
      "grad_norm": 0.04318876191973686,
      "transition_entropy": 3.4339332580566406
    },
    {
      "step": 4,
      "loss": 1.8496863842010498,
      "pair_acc": 0.046875,
      "grad_norm": 0.04323245957493782,
      "transition_entropy": 3.43392413854599
    },
    {
      "step": 5,
      "loss": 1.8672714233398438,
      "pair_acc": 0.046875,
      "grad_norm": 0.041534170508384705,
      "transition_entropy": 3.4339141249656677
    }
  ],
  "eval": [
    {
      "step": 5,
      "rows": [
        {
          "model": "executor",
          "length": 2,
          "k": 0,
          "n": 128,
          "pair_accuracy": 0.015625,
          "register_accuracy": 0.2734375
        },
        {
          "model": "executor",
          "length": 2,
          "k": 1,
          "n": 128,
          "pair_accuracy": 0.03125,
          "register_accuracy": 0.296875
        },
        {
          "model": "executor",
          "length": 2,
          "k": 2,
          "n": 128,
          "pair_accuracy": 0.015625,
          "register_accuracy": 0.30078125
        },
        {
          "model": "executor",
          "length": 2,
          "k": 4,
          "n": 128,
          "pair_accuracy": 0.015625,
          "register_accuracy": 0.30078125
        },
        {
          "model": "executor",
          "length": 2,
          "k": 6,
          "n": 128,
          "pair_accuracy": 0.015625,
          "register_accuracy": 0.30078125
        },
        {
          "model": "executor",
          "length": 4,
          "k": 0,
          "n": 128,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.09765625
        },
        {
          "model": "executor",
          "length": 4,
          "k": 1,
          "n": 128,
          "pair_accuracy": 0.0078125,
          "register_accuracy": 0.0859375
        },
        {
          "model": "executor",
          "length": 4,
          "k": 2,
          "n": 128,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.09765625
        },
        {
          "model": "executor",
          "length": 4,
          "k": 4,
          "n": 128,
          "pair_accuracy": 0.0078125,
          "register_accuracy": 0.1171875
        },
        {
          "model": "executor",
          "length": 4,
          "k": 6,
          "n": 128,
          "pair_accuracy": 0.0078125,
          "register_accuracy": 0.1171875
        },
        {
          "model": "executor",
          "length": 6,
          "k": 0,
          "n": 128,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.04296875
        },
        {
          "model": "executor",
          "length": 6,
          "k": 1,
          "n": 128,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0390625
        },
        {
          "model": "executor",
          "length": 6,
          "k": 2,
          "n": 128,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0390625
        },
        {
          "model": "executor",
          "length": 6,
          "k": 4,
          "n": 128,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.04296875
        },
        {
          "model": "executor",
          "length": 6,
          "k": 6,
          "n": 128,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0546875
        }
      ]
    }
  ]
}