T5-small-summarization
This model is a fine-tuned version of t5-small on the truncated version of Samsum dataset. It achieves the following results on the evaluation set:
- Loss: 1.9294
- Rouge1: 0.3772
- Rouge2: 0.1453
- Rougel: 0.3105
- Rougelsum: 0.3106
- Gen Len: 16.1832
Model description
This model performs the summarization of Texts.
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- 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: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
|---|---|---|---|---|---|---|---|---|
| No log | 1.0 | 63 | 2.1165 | 0.338 | 0.1186 | 0.2811 | 0.2813 | 16.7595 |
| No log | 2.0 | 126 | 2.0210 | 0.3612 | 0.1338 | 0.2982 | 0.2985 | 16.5592 |
| No log | 3.0 | 189 | 1.9838 | 0.3652 | 0.1384 | 0.3034 | 0.304 | 16.1197 |
| No log | 4.0 | 252 | 1.9623 | 0.3715 | 0.142 | 0.3077 | 0.3079 | 16.2308 |
| No log | 5.0 | 315 | 1.9513 | 0.3727 | 0.1441 | 0.308 | 0.3084 | 16.1453 |
| No log | 6.0 | 378 | 1.9419 | 0.375 | 0.1438 | 0.309 | 0.3093 | 16.2234 |
| No log | 7.0 | 441 | 1.9376 | 0.3748 | 0.144 | 0.3102 | 0.3104 | 16.1465 |
| 2.2452 | 8.0 | 504 | 1.9324 | 0.3754 | 0.1451 | 0.3098 | 0.3099 | 16.1893 |
| 2.2452 | 9.0 | 567 | 1.9302 | 0.3769 | 0.1459 | 0.3112 | 0.3113 | 16.1966 |
| 2.2452 | 10.0 | 630 | 1.9294 | 0.3772 | 0.1453 | 0.3105 | 0.3106 | 16.1832 |
Framework versions
- Transformers 4.37.0
- Pytorch 2.1.2
- Datasets 2.1.0
- Tokenizers 0.15.1