nllb-600m-atekervoices-lr2e-4
This model is a fine-tuned version of facebook/nllb-200-distilled-600M on the generator dataset. It achieves the following results on the evaluation set:
- Loss: 1.0903
- Bleu Eng Teo: 23.114
- Bleu Teo Eng: 45.249
- Bleu Eng Kdj: 20.259
- Bleu Kdj Eng: 38.822
- Bleu Mean: 31.861
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: 0.0002
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Use OptimizerNames.ADAFACTOR and the args are: No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 10
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Bleu Eng Teo | Bleu Teo Eng | Bleu Eng Kdj | Bleu Kdj Eng | Bleu Mean |
|---|---|---|---|---|---|---|---|---|
| 3.4524 | 0.0610 | 500 | 2.6771 | 3.705 | 14.878 | 0.138 | 7.858 | 6.645 |
| 2.5471 | 0.1220 | 1000 | 2.2137 | 7.471 | 22.128 | 1.152 | 14.67 | 11.355 |
| 2.2262 | 0.1829 | 1500 | 1.9732 | 10.018 | 25.563 | 1.822 | 18.597 | 14.0 |
| 2.0268 | 0.2439 | 2000 | 1.8295 | 12.738 | 28.164 | 2.46 | 20.894 | 16.064 |
| 1.9124 | 0.3049 | 2500 | 1.7191 | 13.582 | 29.739 | 6.429 | 22.311 | 18.015 |
| 1.8194 | 0.3659 | 3000 | 1.6466 | 14.753 | 31.651 | 6.447 | 24.279 | 19.283 |
| 1.7330 | 0.4268 | 3500 | 1.5826 | 15.577 | 32.773 | 8.682 | 25.061 | 20.523 |
| 1.6707 | 0.4878 | 4000 | 1.5272 | 16.201 | 33.346 | 5.977 | 25.799 | 20.331 |
| 1.6139 | 0.5488 | 4500 | 1.4867 | 16.926 | 34.457 | 7.783 | 27.613 | 21.694 |
| 1.5626 | 0.6098 | 5000 | 1.4525 | 17.304 | 35.262 | 10.539 | 28.048 | 22.788 |
| 1.5479 | 0.6707 | 5500 | 1.4155 | 17.447 | 36.205 | 9.62 | 29.035 | 23.077 |
| 1.5153 | 0.7317 | 6000 | 1.3882 | 17.757 | 36.498 | 10.201 | 28.756 | 23.303 |
| 1.4755 | 0.7927 | 6500 | 1.3603 | 18.393 | 36.981 | 8.136 | 30.601 | 23.528 |
| 1.4714 | 0.8537 | 7000 | 1.3398 | 18.845 | 37.646 | 12.947 | 30.423 | 24.965 |
| 1.4449 | 0.9146 | 7500 | 1.3142 | 18.742 | 38.041 | 12.336 | 31.389 | 25.127 |
| 1.4265 | 0.9756 | 8000 | 1.2971 | 19.496 | 38.259 | 12.304 | 31.882 | 25.485 |
| 1.2616 | 1.0366 | 8500 | 1.2906 | 19.09 | 39.256 | 14.39 | 31.911 | 26.162 |
| 1.1602 | 1.0976 | 9000 | 1.2701 | 19.699 | 39.637 | 14.385 | 32.205 | 26.482 |
| 1.1539 | 1.1585 | 9500 | 1.2659 | 20.19 | 39.603 | 14.995 | 32.316 | 26.776 |
| 1.1516 | 1.2195 | 10000 | 1.2528 | 19.971 | 39.571 | 15.136 | 32.541 | 26.805 |
| 1.1529 | 1.2805 | 10500 | 1.2338 | 20.495 | 40.301 | 14.188 | 33.425 | 27.102 |
| 1.1336 | 1.3415 | 11000 | 1.2275 | 20.454 | 40.893 | 15.525 | 33.861 | 27.683 |
| 1.1220 | 1.4024 | 11500 | 1.2142 | 20.896 | 40.68 | 15.514 | 33.949 | 27.76 |
| 1.1382 | 1.4634 | 12000 | 1.2060 | 20.805 | 41.135 | 15.329 | 34.067 | 27.834 |
| 1.1104 | 1.5244 | 12500 | 1.1931 | 21.09 | 41.2 | 16.949 | 34.11 | 28.338 |
| 1.0994 | 1.5854 | 13000 | 1.1825 | 20.928 | 41.329 | 17.083 | 34.448 | 28.447 |
| 1.0953 | 1.6463 | 13500 | 1.1727 | 21.565 | 41.66 | 16.797 | 34.008 | 28.508 |
| 1.0878 | 1.7073 | 14000 | 1.1644 | 21.484 | 42.252 | 17.121 | 34.971 | 28.957 |
| 1.0728 | 1.7683 | 14500 | 1.1551 | 21.588 | 42.45 | 17.458 | 35.701 | 29.299 |
| 1.0691 | 1.8293 | 15000 | 1.1468 | 21.968 | 42.48 | 17.124 | 35.854 | 29.356 |
| 1.0688 | 1.8902 | 15500 | 1.1388 | 21.972 | 43.037 | 18.038 | 36.617 | 29.916 |
| 1.0626 | 1.9512 | 16000 | 1.1307 | 22.032 | 43.379 | 18.294 | 36.49 | 30.049 |
| 1.0238 | 2.0122 | 16500 | 1.1504 | 22.136 | 43.58 | 18.3 | 36.527 | 30.136 |
| 0.8657 | 2.0732 | 17000 | 1.1431 | 21.977 | 43.624 | 17.755 | 37.209 | 30.141 |
| 0.8479 | 2.1341 | 17500 | 1.1469 | 22.355 | 43.536 | 18.045 | 36.934 | 30.217 |
| 0.8479 | 2.1951 | 18000 | 1.1394 | 22.46 | 43.804 | 18.646 | 37.439 | 30.587 |
| 0.8615 | 2.2561 | 18500 | 1.1329 | 22.517 | 43.771 | 18.849 | 37.609 | 30.687 |
| 0.8506 | 2.3171 | 19000 | 1.1268 | 22.405 | 43.904 | 18.71 | 37.902 | 30.73 |
| 0.8447 | 2.3780 | 19500 | 1.1198 | 22.603 | 44.329 | 18.874 | 38.037 | 30.961 |
| 0.8440 | 2.4390 | 20000 | 1.1185 | 22.592 | 44.074 | 19.28 | 37.82 | 30.941 |
| 0.8329 | 2.5 | 20500 | 1.1166 | 22.514 | 44.414 | 19.362 | 38.71 | 31.25 |
| 0.8403 | 2.5610 | 21000 | 1.1115 | 22.7 | 44.7 | 19.744 | 38.115 | 31.315 |
| 0.8170 | 2.6220 | 21500 | 1.1067 | 22.835 | 44.824 | 20.021 | 38.178 | 31.464 |
| 0.8389 | 2.6829 | 22000 | 1.1035 | 22.846 | 45.016 | 19.835 | 38.601 | 31.574 |
| 0.8320 | 2.7439 | 22500 | 1.0983 | 22.935 | 45.083 | 20.299 | 38.45 | 31.692 |
| 0.8234 | 2.8049 | 23000 | 1.0950 | 23.173 | 45.099 | 20.172 | 38.923 | 31.842 |
| 0.8223 | 2.8659 | 23500 | 1.0939 | 23.209 | 45.203 | 20.413 | 38.881 | 31.926 |
| 0.8215 | 2.9268 | 24000 | 1.0903 | 23.114 | 45.249 | 20.259 | 38.822 | 31.861 |
| 0.8237 | 2.9878 | 24500 | 1.0905 | 23.164 | 45.244 | 20.312 | 38.834 | 31.889 |
| 0.8237 | 3.0 | 24600 | 1.0903 | 23.13 | 45.241 | 20.374 | 38.839 | 31.896 |
Framework versions
- Transformers 5.17.0
- Pytorch 2.14.0+cu130
- Datasets 5.0.1
- Tokenizers 0.23.2