{
  "model_id": "mobtranslate/kuku-yalanji-nllb-lora",
  "version": "0.1.0",
  "created_at": "2026-07-01T00:50:43.481319+00:00",
  "base_model": "facebook/nllb-200-distilled-600M",
  "direction": "eng-gvn",
  "source_lang": "eng_Latn",
  "target_lang": "gvn_Latn",
  "target_lang_init_from": "tpi_Latn",
  "source_lang_token_id": 256047,
  "target_lang_token_id": 256204,
  "dataset": {
    "train_file": "/workspace/data/kuku_yalanji_ebible_parallel_v0.1.0/train.eng-gvn.jsonl",
    "validation_file": "/workspace/data/kuku_yalanji_ebible_parallel_v0.1.0/validation.eng-gvn.jsonl",
    "test_file": "/workspace/data/kuku_yalanji_ebible_parallel_v0.1.0/test.eng-gvn.jsonl",
    "split_rows": {
      "train": 13270,
      "validation": 512,
      "test": 512
    },
    "max_train_samples": null,
    "max_validation_samples": 512,
    "max_test_samples": 512,
    "shuffle_before_cap": true
  },
  "training_args": {
    "epochs": 8.0,
    "batch_size": 8,
    "gradient_accumulation_steps": 4,
    "learning_rate": 0.0001,
    "lora_r": 32,
    "lora_alpha": 64,
    "lora_dropout": 0.05,
    "lora_target_modules": [
      "q_proj",
      "k_proj",
      "v_proj",
      "out_proj",
      "fc1",
      "fc2"
    ],
    "modules_to_save": [
      "model.shared",
      "lm_head"
    ],
    "seed": 42,
    "ensure_weight_tying": true,
    "effective_ensure_weight_tying": false,
    "generation_num_beams": 4,
    "generation_no_repeat_ngram_size": 4,
    "generation_repetition_penalty": 1.15,
    "generation_length_penalty": 0.8
  },
  "metrics": {
    "train": {
      "train_runtime": 3260.5311,
      "train_samples_per_second": 32.559,
      "train_steps_per_second": 1.016,
      "total_flos": 3.769786406957875e+16,
      "train_loss": 2.5613213374418913,
      "epoch": 7.998191681735985
    },
    "validation": {
      "eval_loss": 1.9663472175598145,
      "eval_bleu": 5.5394505873313955,
      "eval_chrf": 33.13722868754415,
      "eval_runtime": 214.4208,
      "eval_samples_per_second": 2.388,
      "eval_steps_per_second": 0.298,
      "epoch": 7.998191681735985
    },
    "test": {
      "test_loss": 1.9890578985214233,
      "test_bleu": 5.274383048216052,
      "test_chrf": 32.65721839454287,
      "test_runtime": 232.3938,
      "test_samples_per_second": 2.203,
      "test_steps_per_second": 0.275,
      "epoch": 7.998191681735985
    }
  },
  "artifacts": {
    "adapter_dir": "/workspace/models/kuku-yalanji-nllb-lora/gvn-token-full-v0.4.0/adapter",
    "merged_dir": "/workspace/models/kuku-yalanji-nllb-lora/gvn-token-full-v0.4.0/merged"
  }
}