{
  "model_id": "mobtranslate/kuku-yalanji-nllb-lora",
  "version": "0.1.0",
  "created_at": "2026-07-01T06:58:14.575356+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_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": 4,
    "gradient_accumulation_steps": 8,
    "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": [],
    "seed": 42,
    "ensure_weight_tying": true,
    "effective_ensure_weight_tying": true,
    "generation_num_beams": 4,
    "generation_no_repeat_ngram_size": 4,
    "generation_repetition_penalty": 1.15,
    "generation_length_penalty": 0.8,
    "save_total_limit": 2,
    "gradient_checkpointing": false
  },
  "metrics": {
    "train": {
      "train_runtime": 10005.437,
      "train_samples_per_second": 10.61,
      "train_steps_per_second": 0.331,
      "total_flos": 5.11259258573783e+16,
      "train_loss": 2.1281005921571152,
      "epoch": 7.981314044605184
    },
    "validation": {
      "eval_loss": 1.6889560222625732,
      "eval_bleu": 8.745046033766966,
      "eval_chrf": 37.51884191837713,
      "eval_runtime": 505.3277,
      "eval_samples_per_second": 1.013,
      "eval_steps_per_second": 0.253,
      "epoch": 7.981314044605184
    },
    "test": {
      "test_loss": 1.6784048080444336,
      "test_bleu": 8.400941927903686,
      "test_chrf": 37.094134686675,
      "test_runtime": 522.6972,
      "test_samples_per_second": 0.98,
      "test_steps_per_second": 0.245,
      "epoch": 7.981314044605184
    }
  },
  "artifacts": {
    "adapter_dir": "/workspace/models/kuku-yalanji-nllb-lora/tpi-proxy-full-v0.7.0-1.3b/adapter",
    "merged_dir": "/workspace/models/kuku-yalanji-nllb-lora/tpi-proxy-full-v0.7.0-1.3b/merged"
  }
}