bart_baseline_1024 / README.md
jordanfan's picture
training completed[dev]: 1024 128
1c57d1c verified
metadata
license: apache-2.0
base_model: facebook/bart-large
tags:
  - generated_from_trainer
metrics:
  - rouge
  - wer
model-index:
  - name: bart_baseline_1024
    results: []

bart_baseline_1024

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

  • Loss: 0.9558
  • Rouge1: 0.7069
  • Rouge2: 0.4544
  • Rougel: 0.6489
  • Rougelsum: 0.6489
  • Wer: 0.4398

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: 2e-05
  • train_batch_size: 6
  • eval_batch_size: 6
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 2
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Rouge1 Rouge2 Rougel Rougelsum Wer
No log 0.13 250 1.2224 0.665 0.3911 0.6 0.6001 0.4993
2.0905 0.27 500 1.1190 0.6743 0.4083 0.6103 0.6104 0.4809
2.0905 0.4 750 1.0832 0.6818 0.418 0.6178 0.6178 0.4726
1.188 0.53 1000 1.0541 0.6871 0.4246 0.6242 0.6242 0.4675
1.188 0.66 1250 1.0352 0.6881 0.4283 0.6269 0.6268 0.4628
1.1172 0.8 1500 1.0291 0.6912 0.4319 0.6303 0.6303 0.4586
1.1172 0.93 1750 1.0079 0.7001 0.4406 0.6396 0.6397 0.4543
1.0803 1.06 2000 0.9957 0.6939 0.4396 0.6359 0.6359 0.4511
1.0803 1.2 2250 0.9891 0.6972 0.443 0.6383 0.6383 0.4479
0.9849 1.33 2500 0.9800 0.7009 0.4467 0.6425 0.6425 0.4464
0.9849 1.46 2750 0.9771 0.7017 0.4479 0.6426 0.6426 0.4437
0.9867 1.6 3000 0.9638 0.7085 0.4541 0.6495 0.6495 0.4422
0.9867 1.73 3250 0.9675 0.7013 0.4495 0.6438 0.6438 0.4413
0.9556 1.86 3500 0.9565 0.707 0.4544 0.6493 0.6492 0.4401
0.9556 1.99 3750 0.9558 0.7069 0.4544 0.6489 0.6489 0.4398

Framework versions

  • Transformers 4.38.2
  • Pytorch 2.2.1+cu121
  • Datasets 2.18.0
  • Tokenizers 0.15.2