{
  "mode": "executor",
  "args": {
    "mode": "executor",
    "seed": 11,
    "modulus": 97,
    "train_min_len": 1,
    "train_max_len": 4,
    "eval_lengths": "2,4,6",
    "eval_k": "0,1,2,4,6",
    "dim": 64,
    "heads": 4,
    "compiler_layers": 1,
    "num_bases": 8,
    "rank": 8,
    "mem_dim": 24,
    "dropout": 0.05,
    "disable_memory": false,
    "disable_dynamic": false,
    "train_steps": 5,
    "batch_size": 32,
    "eval_batch_size": 32,
    "eval_examples": 64,
    "lr": 0.0003,
    "weight_decay": 0.01,
    "max_grad_norm": 1.0,
    "log_every": 1,
    "eval_every": 5,
    "output_dir": "runs_latent_executor/smoke_executor"
  },
  "train": [
    {
      "step": 1,
      "loss": 4.856504440307617,
      "pair_acc": 0.0,
      "grad_norm": 1.3153740167617798,
      "gate_entropy": 5.516522169113159,
      "mem_rms": 0.09313304722309113
    },
    {
      "step": 2,
      "loss": 4.708869457244873,
      "pair_acc": 0.0,
      "grad_norm": 1.2834947109222412,
      "gate_entropy": 5.5154924392700195,
      "mem_rms": 0.09358807653188705
    },
    {
      "step": 3,
      "loss": 4.726237773895264,
      "pair_acc": 0.0,
      "grad_norm": 1.1562975645065308,
      "gate_entropy": 5.515860676765442,
      "mem_rms": 0.10348620265722275
    },
    {
      "step": 4,
      "loss": 4.693971633911133,
      "pair_acc": 0.0,
      "grad_norm": 1.2223085165023804,
      "gate_entropy": 5.516818881034851,
      "mem_rms": 0.0928007960319519
    },
    {
      "step": 5,
      "loss": 4.728033542633057,
      "pair_acc": 0.0,
      "grad_norm": 1.1779452562332153,
      "gate_entropy": 5.513905882835388,
      "mem_rms": 0.08401723206043243
    }
  ],
  "eval": [
    {
      "step": 5,
      "rows": [
        {
          "model": "executor",
          "length": 2,
          "k": 0,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0078125
        },
        {
          "model": "executor",
          "length": 2,
          "k": 1,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0078125
        },
        {
          "model": "executor",
          "length": 2,
          "k": 2,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.015625
        },
        {
          "model": "executor",
          "length": 2,
          "k": 4,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.015625
        },
        {
          "model": "executor",
          "length": 2,
          "k": 6,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.015625
        },
        {
          "model": "executor",
          "length": 4,
          "k": 0,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0078125
        },
        {
          "model": "executor",
          "length": 4,
          "k": 1,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.015625
        },
        {
          "model": "executor",
          "length": 4,
          "k": 2,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0078125
        },
        {
          "model": "executor",
          "length": 4,
          "k": 4,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0078125
        },
        {
          "model": "executor",
          "length": 4,
          "k": 6,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0078125
        },
        {
          "model": "executor",
          "length": 6,
          "k": 0,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.015625
        },
        {
          "model": "executor",
          "length": 6,
          "k": 1,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0078125
        },
        {
          "model": "executor",
          "length": 6,
          "k": 2,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0
        },
        {
          "model": "executor",
          "length": 6,
          "k": 4,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0234375
        },
        {
          "model": "executor",
          "length": 6,
          "k": 6,
          "n": 64,
          "pair_accuracy": 0.0,
          "register_accuracy": 0.0
        }
      ]
    }
  ]
}