bbc-to-ind-nmt-v1 / README.md
kepinsam's picture
End of training
38ab85f
metadata
license: cc-by-nc-4.0
base_model: facebook/nllb-200-distilled-600M
tags:
  - generated_from_trainer
datasets:
  - nusatranslation_mt
metrics:
  - sacrebleu
model-index:
  - name: bataknese_to_indonesian_translation_model
    results:
      - task:
          name: Sequence-to-sequence Language Modeling
          type: text2text-generation
        dataset:
          name: nusatranslation_mt
          type: nusatranslation_mt
          config: nusatranslation_mt_btk_ind_source
          split: test
          args: nusatranslation_mt_btk_ind_source
        metrics:
          - name: Sacrebleu
            type: sacrebleu
            value: 38.1114

bataknese_to_indonesian_translation_model

This model is a fine-tuned version of facebook/nllb-200-distilled-600M on the nusatranslation_mt dataset. It achieves the following results on the evaluation set:

  • Loss: 1.1353
  • Sacrebleu: 38.1114
  • Gen Len: 37.3265

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-05
  • train_batch_size: 4
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 5

Training results

Training Loss Epoch Step Validation Loss Sacrebleu Gen Len
1.36 1.0 1650 1.2432 34.4418 37.63
1.0823 2.0 3300 1.1404 36.8766 37.312
0.9761 3.0 4950 1.1281 37.9027 37.2895
0.8241 4.0 6600 1.1267 38.1879 37.1395
0.7579 5.0 8250 1.1353 38.1114 37.3265

Framework versions

  • Transformers 4.34.1
  • Pytorch 2.1.0+cu118
  • Datasets 2.13.1
  • Tokenizers 0.14.1