t5-small-finetuned-xlsum-chinese-tradition
This model is a fine-tuned version of t5-small on the xlsum dataset. It achieves the following results on the evaluation set:
- Loss: 1.2061
- Rouge1: 0.8887
- Rouge2: 0.0671
- Rougel: 0.889
- Rougelsum: 0.8838
- Gen Len: 6.8779
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: 6
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
|---|---|---|---|---|---|---|---|---|
| 1.4231 | 1.0 | 2336 | 1.2586 | 0.711 | 0.0528 | 0.7029 | 0.7053 | 7.3368 |
| 1.378 | 2.0 | 4672 | 1.2281 | 0.9688 | 0.05 | 0.9574 | 0.9656 | 7.0392 |
| 1.3567 | 3.0 | 7008 | 1.2182 | 0.9534 | 0.1035 | 0.9531 | 0.9472 | 6.7437 |
| 1.3339 | 4.0 | 9344 | 1.2096 | 0.9969 | 0.0814 | 0.9969 | 0.9938 | 7.4503 |
| 1.3537 | 5.0 | 11680 | 1.2072 | 0.8429 | 0.0742 | 0.8372 | 0.838 | 6.8049 |
| 1.3351 | 6.0 | 14016 | 1.2061 | 0.8887 | 0.0671 | 0.889 | 0.8838 | 6.8779 |
Framework versions
- Transformers 4.20.1
- Pytorch 1.11.0+cu113
- Datasets 2.3.2
- Tokenizers 0.12.1