{
  "model_id": "mobtranslate/kuku-yalanji-nllb-lora",
  "version": "0.1.0",
  "created_at": "2026-07-01T22:06:47.629659+00:00",
  "base_model": "facebook/nllb-200-distilled-1.3B",
  "direction": "eng-gvn",
  "source_lang": "eng_Latn",
  "target_lang": "tpi_Latn",
  "target_lang_init_from": null,
  "source_lang_token_id": 256047,
  "target_lang_token_id": 256179,
  "dataset": {
    "train_file": "/workspace/data/kuku_yalanji_v9_7_tagged_direct_plus_reference_bible/v8_2048row/train.eng-gvn.jsonl",
    "validation_file": "/workspace/data/kuku_yalanji_v9_7_tagged_direct_plus_reference_bible/v8_2048row/eval_train.eng-gvn.jsonl",
    "test_file": "/workspace/data/kuku_yalanji_v9_7_tagged_direct_plus_reference_bible/v8_2048row/eval_train.eng-gvn.jsonl",
    "split_rows": {
      "train": 4096,
      "validation": 128,
      "test": 4096
    },
    "max_train_samples": null,
    "max_validation_samples": 128,
    "max_test_samples": null,
    "shuffle_before_cap": false
  },
  "training_args": {
    "epochs": 25.0,
    "batch_size": 16,
    "gradient_accumulation_steps": 1,
    "learning_rate": 0.0002,
    "lora_r": 64,
    "lora_alpha": 128,
    "lora_dropout": 0.0,
    "lora_target_modules": [
      "q_proj",
      "k_proj",
      "v_proj",
      "out_proj",
      "fc1",
      "fc2"
    ],
    "modules_to_save": [],
    "seed": 42,
    "ensure_weight_tying": true,
    "effective_ensure_weight_tying": true,
    "generation_num_beams": 1,
    "generation_no_repeat_ngram_size": 0,
    "generation_repetition_penalty": 1.0,
    "generation_length_penalty": 1.0,
    "save_total_limit": 2,
    "gradient_checkpointing": false
  },
  "metrics": {
    "train": {
      "train_runtime": 1020.1721,
      "train_samples_per_second": 100.375,
      "train_steps_per_second": 6.273,
      "total_flos": 4.551924976975872e+16,
      "train_loss": 0.4988648035377264,
      "epoch": 25.0
    },
    "validation": {
      "eval_loss": 0.002455136040225625,
      "eval_bleu": 100.00000000000004,
      "eval_chrf": 100.0,
      "eval_runtime": 8.9762,
      "eval_samples_per_second": 14.26,
      "eval_steps_per_second": 0.891,
      "epoch": 25.0
    },
    "test": {
      "test_loss": 0.002999512478709221,
      "test_bleu": 99.7547842886766,
      "test_chrf": 99.88157586353353,
      "test_runtime": 303.7513,
      "test_samples_per_second": 13.485,
      "test_steps_per_second": 0.843,
      "epoch": 25.0
    }
  },
  "artifacts": {
    "adapter_dir": "/workspace/models/kuku-yalanji-nllb-lora/v9.7-tagged-direct-plus-reference-bible-tpi-4096row-25epoch-batch16/adapter",
    "merged_dir": "/workspace/models/kuku-yalanji-nllb-lora/v9.7-tagged-direct-plus-reference-bible-tpi-4096row-25epoch-batch16/merged"
  }
}