t5-small-finetuned-xsum
This model is a fine-tuned version of t5-small on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.0531
- Rouge1: 97.0969
- Rouge2: 95.8095
- Rougel: 96.7452
- Rougelsum: 96.7363
- Gen Len: 15.3151
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: 16
- eval_batch_size: 16
- 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 |
|---|---|---|---|---|---|---|---|---|
| 1.033 | 1.0 | 519 | 0.2212 | 95.3952 | 91.6915 | 94.8024 | 94.7963 | 14.9826 |
| 0.2732 | 2.0 | 1038 | 0.1233 | 96.3758 | 94.121 | 95.9352 | 95.9297 | 15.2223 |
| 0.1814 | 3.0 | 1557 | 0.0940 | 96.7098 | 94.7563 | 96.2577 | 96.2413 | 15.2133 |
| 0.144 | 4.0 | 2076 | 0.0757 | 96.6801 | 95.0173 | 96.2782 | 96.2691 | 15.2679 |
| 0.1213 | 5.0 | 2595 | 0.0688 | 96.8498 | 95.2702 | 96.5014 | 96.485 | 15.2515 |
| 0.1043 | 6.0 | 3114 | 0.0620 | 96.8951 | 95.3824 | 96.5526 | 96.5419 | 15.2808 |
| 0.0938 | 7.0 | 3633 | 0.0561 | 97.0021 | 95.6205 | 96.6811 | 96.6711 | 15.3163 |
| 0.0877 | 8.0 | 4152 | 0.0546 | 97.016 | 95.7049 | 96.6736 | 96.6688 | 15.3044 |
| 0.0873 | 9.0 | 4671 | 0.0534 | 97.0697 | 95.7894 | 96.7221 | 96.7192 | 15.3123 |
| 0.0841 | 10.0 | 5190 | 0.0531 | 97.0969 | 95.8095 | 96.7452 | 96.7363 | 15.3151 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2