{
  "model_id": "mobtranslate/kuku-yalanji-nllb-lora",
  "version": "0.1.0",
  "created_at": "2026-07-02T02:54:50.734031+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_v10_0_bible_glossary_usage/train.eng-gvn.jsonl",
    "validation_file": "/workspace/data/kuku_yalanji_v10_0_bible_glossary_usage/validation.eng-gvn.jsonl",
    "test_file": "/workspace/data/kuku_yalanji_v10_0_bible_glossary_usage/eval_train.eng-gvn.jsonl",
    "split_rows": {
      "train": 5556,
      "validation": 128,
      "test": 4461
    },
    "max_train_samples": null,
    "max_validation_samples": 128,
    "max_test_samples": null,
    "shuffle_before_cap": false
  },
  "training_args": {
    "epochs": 20.0,
    "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": 2142.6696,
      "train_samples_per_second": 51.861,
      "train_steps_per_second": 3.248,
      "total_flos": 4.811476256400998e+16,
      "train_loss": 0.5432022249561617,
      "epoch": 20.0
    },
    "validation": {
      "eval_loss": 0.004856833256781101,
      "eval_bleu": 99.37879241219281,
      "eval_chrf": 99.78663010512842,
      "eval_runtime": 20.7646,
      "eval_samples_per_second": 6.164,
      "eval_steps_per_second": 0.385,
      "epoch": 20.0
    },
    "test": {
      "test_loss": 0.005108408164232969,
      "test_bleu": 99.52680633613102,
      "test_chrf": 99.73610084617087,
      "test_runtime": 784.5282,
      "test_samples_per_second": 5.686,
      "test_steps_per_second": 0.356,
      "epoch": 20.0
    }
  },
  "artifacts": {
    "adapter_dir": "/workspace/models/kuku-yalanji-nllb-lora/v10.0-tagged-bible-plus-glossary-usage-tpi-20epoch-batch16/adapter",
    "merged_dir": "/workspace/models/kuku-yalanji-nllb-lora/v10.0-tagged-bible-plus-glossary-usage-tpi-20epoch-batch16/merged"
  }
}
