<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
bart-paraphrase-finetuned-xsum-v4
This model is a fine-tuned version of eugenesiow/bart-paraphrase on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 1.1765
- Rouge1: 49.972
- Rouge2: 49.85
- Rougel: 49.9165
- Rougelsum: 49.7819
- Gen Len: 8.3061
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: 2
- eval_batch_size: 2
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
|---|---|---|---|---|---|---|---|---|
| No log | 1.0 | 263 | 0.5050 | 47.9628 | 47.7085 | 47.8625 | 47.772 | 6.9639 |
| 0.676 | 2.0 | 526 | 0.5793 | 49.6085 | 49.3495 | 49.5196 | 49.4173 | 7.4715 |
| 0.676 | 3.0 | 789 | 0.7011 | 49.8635 | 49.6937 | 49.8155 | 49.6604 | 7.576 |
| 0.322 | 4.0 | 1052 | 0.7585 | 49.8851 | 49.7578 | 49.8526 | 49.6977 | 7.6654 |
| 0.322 | 5.0 | 1315 | 0.6615 | 49.861 | 49.7185 | 49.7978 | 49.6669 | 8.3023 |
| 0.2828 | 6.0 | 1578 | 0.6233 | 49.916 | 49.7819 | 49.8861 | 49.7384 | 7.6084 |
| 0.2828 | 7.0 | 1841 | 0.9380 | 49.916 | 49.7819 | 49.8861 | 49.7384 | 8.2433 |
| 0.2073 | 8.0 | 2104 | 0.8497 | 49.9624 | 49.8355 | 49.91 | 49.7666 | 7.6331 |
| 0.2073 | 9.0 | 2367 | 0.7715 | 49.972 | 49.85 | 49.9165 | 49.7819 | 7.9772 |
| 0.1744 | 10.0 | 2630 | 1.1765 | 49.972 | 49.85 | 49.9165 | 49.7819 | 8.3061 |
Framework versions
- Transformers 4.19.2
- Pytorch 1.11.0+cu113
- Datasets 2.2.2
- Tokenizers 0.12.1
