debiased / README.md
samzirbo's picture
End of training
bd26fcc verified
metadata
base_model: samzirbo/gendered
tags:
  - generated_from_trainer
metrics:
  - bleu
model-index:
  - name: debiased
    results: []

debiased

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

  • Loss: 1.2202
  • Bleu: 43.2797
  • Meteor: 0.6854
  • Chrf++: 62.2643

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: 0.0001
  • 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: 10000

Training results

Training Loss Epoch Step Validation Loss Bleu Meteor Chrf++
1.4077 4.0 2500 1.2011 43.3975 0.685 62.313
1.2934 8.0 5000 1.2125 43.223 0.6845 62.1664
1.2352 12.0 7500 1.2173 43.2517 0.685 62.2467
1.2137 16.0 10000 1.2202 43.2797 0.6854 62.2643

Framework versions

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