{
  "model_id": "mobtranslate/kuku-yalanji-nllb-lora",
  "version": "0.1.0",
  "created_at": "2026-07-02T15:51:37.641886+00:00",
  "base_model": "/workspace/base_models/v12.0-tagged-direct-plus-reference-bible-gvn-token-4096row-25epoch-batch16/merged",
  "direction": "eng-gvn",
  "source_lang": "eng_Latn",
  "target_lang": "gvn_Latn",
  "target_lang_init_from": null,
  "source_lang_token_id": 256047,
  "target_lang_token_id": 256204,
  "dataset": {
    "train_file": "/workspace/data/kuku_yalanji_v20_0_full_candidate_corpus_gvn/train.eng-gvn.jsonl",
    "validation_file": "/workspace/data/kuku_yalanji_v20_0_full_candidate_corpus_gvn/validation.eng-gvn.jsonl",
    "test_file": "/workspace/data/kuku_yalanji_v20_0_full_candidate_corpus_gvn/test.eng-gvn.jsonl",
    "split_rows": {
      "train": 35394,
      "validation": 1024,
      "test": 1024
    },
    "max_train_samples": null,
    "max_validation_samples": 1024,
    "max_test_samples": 1024,
    "shuffle_before_cap": false
  },
  "training_args": {
    "epochs": 3.0,
    "batch_size": 8,
    "gradient_accumulation_steps": 2,
    "learning_rate": 2e-05,
    "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": 6767.6911,
      "train_samples_per_second": 15.69,
      "train_steps_per_second": 0.981,
      "total_flos": 8.119575264288768e+16,
      "train_loss": 1.4851415328335662,
      "epoch": 2.998870056497175
    },
    "validation": {
      "eval_loss": 1.5804394483566284,
      "eval_bleu": 13.406777236161425,
      "eval_chrf": 39.882101420536515,
      "eval_runtime": 556.9744,
      "eval_samples_per_second": 1.839,
      "eval_steps_per_second": 0.23,
      "epoch": 2.998870056497175
    },
    "test": {
      "test_loss": 1.5612670183181763,
      "test_bleu": 13.157394044826804,
      "test_chrf": 40.2505720878999,
      "test_runtime": 539.4346,
      "test_samples_per_second": 1.898,
      "test_steps_per_second": 0.237,
      "epoch": 2.998870056497175
    }
  },
  "artifacts": {
    "adapter_dir": "/workspace/models/kuku-yalanji-nllb-lora/v20.0-full-candidate-corpus-gvn-3epoch-lr2e-5/adapter",
    "merged_dir": "/workspace/models/kuku-yalanji-nllb-lora/v20.0-full-candidate-corpus-gvn-3epoch-lr2e-5/merged"
  }
}