matekadlicsko's picture
Update README.md
812c7da verified
|
raw
history blame
2.33 kB
metadata
license: cc-by-4.0
base_model: Helsinki-NLP/opus-mt-tc-big-hu-en
tags:
  - generated_from_trainer
metrics:
  - bleu
model-index:
  - name: output
    results: []
datasets:
  - matekadlicsko/hungarian-news-translations
language:
  - hu
  - en
pipeline_tag: translation

opus-mt-tc-big-hu-en-finetuned-news

This model is a fine-tuned version of Helsinki-NLP/opus-mt-tc-big-hu-en on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 1.1956
  • Bleu: 39.6514
  • Gen Len: 51.7972

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: 16
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 3

Training results

Training Loss Epoch Step Validation Loss Bleu Gen Len
No log 0.23 150 1.2390 37.7285 51.2344
No log 0.46 300 1.2198 38.0141 50.9959
No log 0.7 450 1.1969 38.4344 51.464
1.2949 0.93 600 1.1898 38.6658 51.379
1.2949 1.16 750 1.1968 38.646 51.34
1.2949 1.39 900 1.1948 38.9869 51.7807
1.0345 1.62 1050 1.1866 39.1708 51.7604
1.0345 1.85 1200 1.1792 39.4682 51.661
1.0345 2.09 1350 1.1974 39.5432 51.882
0.8679 2.32 1500 1.1971 39.5958 51.7895
0.8679 2.55 1650 1.1980 39.5926 52.0244
0.8679 2.78 1800 1.1961 39.6369 51.8143

Framework versions

  • Transformers 4.35.2
  • Pytorch 2.1.0+cu121
  • Datasets 2.16.1
  • Tokenizers 0.15.0