{
  "model_id": "mobtranslate/kuku-yalanji-nllb-lora",
  "version": "0.1.0",
  "created_at": "2026-07-02T10:22:46.507632+00:00",
  "base_model": "/workspace/base_models/v10.0-tagged-bible-plus-glossary-usage-tpi-20epoch-batch16/merged",
  "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_v18_0_v10_plus_photo_story/db_usage_plus_photo/train_usage_photo_oversampled.eng-gvn.jsonl",
    "validation_file": "/workspace/data/kuku_yalanji_v18_0_v10_plus_photo_story/db_usage_plus_photo/eval_train_usage_photo.eng-gvn.jsonl",
    "test_file": "/workspace/data/kuku_yalanji_v18_0_v10_plus_photo_story/db_usage_plus_photo/eval_train_usage_photo.eng-gvn.jsonl",
    "split_rows": {
      "train": 1976,
      "validation": 128,
      "test": 408
    },
    "max_train_samples": null,
    "max_validation_samples": 128,
    "max_test_samples": 408,
    "shuffle_before_cap": false
  },
  "training_args": {
    "epochs": 10.0,
    "batch_size": 16,
    "gradient_accumulation_steps": 1,
    "learning_rate": 5e-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": 431.5132,
      "train_samples_per_second": 45.792,
      "train_steps_per_second": 2.874,
      "total_flos": 7312682600103936.0,
      "train_loss": 0.2773259439776021,
      "epoch": 10.0
    },
    "validation": {
      "eval_loss": 0.0007577191572636366,
      "eval_bleu": 100.00000000000004,
      "eval_chrf": 100.0,
      "eval_runtime": 15.4292,
      "eval_samples_per_second": 8.296,
      "eval_steps_per_second": 0.518,
      "epoch": 10.0
    },
    "test": {
      "test_loss": 0.0007674304069951177,
      "test_bleu": 100.00000000000004,
      "test_chrf": 100.0,
      "test_runtime": 52.0843,
      "test_samples_per_second": 7.833,
      "test_steps_per_second": 0.499,
      "epoch": 10.0
    }
  },
  "artifacts": {
    "adapter_dir": "/workspace/models/kuku-yalanji-nllb-lora/v18.0-usage-photo-continuation-from-v10-tpi-10epoch-lr5e-5/adapter",
    "merged_dir": "/workspace/models/kuku-yalanji-nllb-lora/v18.0-usage-photo-continuation-from-v10-tpi-10epoch-lr5e-5/merged"
  }
}