ldos's picture
End of training
52287d0
metadata
license: mit
base_model: facebook/bart-large-xsum
tags:
  - generated_from_trainer
metrics:
  - rouge
model-index:
  - name: text_shortening_model_v46
    results: []

text_shortening_model_v46

This model is a fine-tuned version of facebook/bart-large-xsum on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 1.8536
  • Rouge1: 0.485
  • Rouge2: 0.271
  • Rougel: 0.4374
  • Rougelsum: 0.4371
  • Bert precision: 0.8676
  • Bert recall: 0.8761
  • Average word count: 9.1032
  • Max word count: 17
  • Min word count: 4
  • Average token count: 15.8254
  • % shortened texts with length > 12: 9.5238

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: 0.0003
  • train_batch_size: 32
  • eval_batch_size: 32
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 7

Training results

Training Loss Epoch Step Validation Loss Rouge1 Rouge2 Rougel Rougelsum Bert precision Bert recall Average word count Max word count Min word count Average token count % shortened texts with length > 12
1.8213 1.0 42 2.1030 0.4561 0.2433 0.4131 0.4131 0.8606 0.8724 9.0529 15 5 14.7249 7.4074
0.8874 2.0 84 1.8034 0.4778 0.2609 0.4316 0.4316 0.8569 0.8787 10.6323 21 5 16.2963 19.3122
0.603 3.0 126 1.6613 0.4749 0.2594 0.425 0.4253 0.8576 0.8796 10.5106 21 5 16.2751 23.0159
0.5413 4.0 168 1.5975 0.4729 0.249 0.4258 0.4254 0.8635 0.8696 8.6481 16 4 14.3677 4.2328
0.3393 5.0 210 1.6755 0.4959 0.28 0.4476 0.4473 0.8687 0.8772 8.8942 20 5 15.8915 8.4656
0.2573 6.0 252 1.6908 0.4775 0.2589 0.4309 0.4307 0.866 0.873 8.9868 22 4 15.4339 10.3175
0.173 7.0 294 1.8536 0.485 0.271 0.4374 0.4371 0.8676 0.8761 9.1032 17 4 15.8254 9.5238

Framework versions

  • Transformers 4.33.1
  • Pytorch 2.0.1+cu118
  • Datasets 2.14.5
  • Tokenizers 0.13.3