{
  "mode": "fixed",
  "seed": 1,
  "hf_repo": "Likithp/v10_fixed_s1",
  "base_model": "Qwen/Qwen2.5-0.5B",
  "dataset": "data/cs7_fixed_v3",
  "dataset_version": "cs7_v3",
  "trained_at": "2026-06-06T07:48:24.312420+00:00",
  "optimizer": "AdamW",
  "lr": 3e-05,
  "weight_decay": 0.0,
  "batch_size": 8,
  "grad_accum": 8,
  "effective_batch": 64,
  "epochs": 5,
  "warmup_steps": 200,
  "grad_clip": 0.5,
  "lr_schedule": "cosine",
  "checkpoint_criterion": "val_em",
  "max_seq_len": 256,
  "dtype": "torch.bfloat16",
  "eval_method": "teacher_forcing_argmax",
  "alias_groups": {
    "T1": [
      "act",
      "cst",
      "emp",
      "inv",
      "ord",
      "spl",
      "txn"
    ],
    "T2": [
      "brc",
      "ctg",
      "dpt",
      "empl",
      "ordr",
      "prd",
      "prj",
      "rgn",
      "shp",
      "tsk",
      "whs"
    ]
  },
  "log_every_steps": 50,
  "eval_n": 500,
  "dry_run": false,
  "best_epoch": 5,
  "best_train_loss": null,
  "val_exact_match": 1.0,
  "val_outer_alias_acc": 1.0,
  "val_inner_alias_acc": 1.0,
  "val_t1_inner_alias_acc": 1.0,
  "val_t2_inner_alias_acc": 1.0,
  "val_t1_t2_gap_pp": 0.0,
  "val_inner_by_alias": {
    "act": {
      "correct": 64,
      "total": 64,
      "pct": 100.0,
      "token_group": "T1"
    },
    "brc": {
      "correct": 50,
      "total": 50,
      "pct": 100.0,
      "token_group": "T2"
    },
    "cst": {
      "correct": 45,
      "total": 45,
      "pct": 100.0,
      "token_group": "T1"
    },
    "ctg": {
      "correct": 61,
      "total": 61,
      "pct": 100.0,
      "token_group": "T2"
    },
    "dpt": {
      "correct": 56,
      "total": 56,
      "pct": 100.0,
      "token_group": "T2"
    },
    "emp": {
      "correct": 51,
      "total": 51,
      "pct": 100.0,
      "token_group": "T1"
    },
    "empl": {
      "correct": 59,
      "total": 59,
      "pct": 100.0,
      "token_group": "T2"
    },
    "inv": {
      "correct": 45,
      "total": 45,
      "pct": 100.0,
      "token_group": "T1"
    },
    "ord": {
      "correct": 67,
      "total": 67,
      "pct": 100.0,
      "token_group": "T1"
    },
    "ordr": {
      "correct": 59,
      "total": 59,
      "pct": 100.0,
      "token_group": "T2"
    },
    "prd": {
      "correct": 57,
      "total": 57,
      "pct": 100.0,
      "token_group": "T2"
    },
    "prj": {
      "correct": 52,
      "total": 52,
      "pct": 100.0,
      "token_group": "T2"
    },
    "rgn": {
      "correct": 70,
      "total": 70,
      "pct": 100.0,
      "token_group": "T2"
    },
    "shp": {
      "correct": 60,
      "total": 60,
      "pct": 100.0,
      "token_group": "T2"
    },
    "spl": {
      "correct": 59,
      "total": 59,
      "pct": 100.0,
      "token_group": "T1"
    },
    "tsk": {
      "correct": 44,
      "total": 44,
      "pct": 100.0,
      "token_group": "T2"
    },
    "txn": {
      "correct": 51,
      "total": 51,
      "pct": 100.0,
      "token_group": "T1"
    },
    "whs": {
      "correct": 50,
      "total": 50,
      "pct": 100.0,
      "token_group": "T2"
    }
  },
  "train_log": [
    {
      "epoch": 1,
      "loss": 0.026036,
      "val_em": 1.0
    },
    {
      "epoch": 2,
      "loss": 1e-05,
      "val_em": 1.0
    },
    {
      "epoch": 3,
      "loss": 1e-05,
      "val_em": 1.0
    },
    {
      "epoch": 4,
      "loss": 1e-05,
      "val_em": 1.0
    },
    {
      "epoch": 5,
      "loss": 1e-05,
      "val_em": 1.0
    }
  ]
}