gendered_last / README.md
samzirbo's picture
End of training
e47ed28 verified
metadata
base_model: samzirbo/mT5.en-es.pretrained
tags:
  - generated_from_trainer
metrics:
  - bleu
model-index:
  - name: gendered_new
    results: []

gendered_new

This model is a fine-tuned version of samzirbo/mT5.en-es.pretrained on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 1.1752
  • Bleu: 43.4465
  • Meteor: 0.6886
  • Chrf++: 62.4831

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: 0.0005
  • train_batch_size: 64
  • eval_batch_size: 64
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 1000
  • training_steps: 50000

Training results

Training Loss Epoch Step Validation Loss Bleu Meteor Chrf++
4.5213 0.26 2500 2.0240 27.6825 0.5543 48.9268
2.422 0.53 5000 1.7336 33.2108 0.6032 54.1788
2.173 0.79 7500 1.5902 35.8768 0.6243 56.215
2.0274 1.05 10000 1.4993 37.3691 0.6371 57.5369
1.9096 1.32 12500 1.4399 38.4947 0.6495 58.5692
1.8532 1.58 15000 1.3892 39.6338 0.6586 59.4359
1.7999 1.84 17500 1.3481 40.1694 0.6639 59.8771
1.7366 2.11 20000 1.3057 41.1684 0.6702 60.6373
1.6849 2.37 22500 1.2913 41.2899 0.6702 60.7243
1.6608 2.64 25000 1.2600 41.9037 0.6749 61.1685
1.6367 2.9 27500 1.2382 42.2288 0.6806 61.5742
1.5943 3.16 30000 1.2196 42.9029 0.6828 61.9359
1.5647 3.43 32500 1.2091 42.7591 0.6826 61.9382
1.5553 3.69 35000 1.1987 43.2246 0.6845 62.2767
1.5466 3.95 37500 1.1888 43.3998 0.687 62.3713
1.5153 4.22 40000 1.1826 43.3886 0.6883 62.4512
1.5089 4.48 42500 1.1786 43.5134 0.6892 62.5449
1.5035 4.74 45000 1.1769 43.4891 0.6884 62.5178
1.5001 5.01 47500 1.1754 43.3885 0.6882 62.4596
1.4901 5.27 50000 1.1752 43.4465 0.6886 62.4831

Framework versions

  • Transformers 4.38.0
  • Pytorch 2.2.1+cu121
  • Datasets 2.19.1
  • Tokenizers 0.15.2