<!-- 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. -->
t5-small-t5small-gigaword
This model is a fine-tuned version of t5-small on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 1.4052
- Rouge1: 50.1555
- Rouge2: 25.5096
- Rougel: 46.5771
- Rougelsum: 46.5827
- Gen Len: 14.246
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: 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: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
|---|---|---|---|---|---|---|---|---|
| 1.9066 | 1.0 | 118874 | 1.4971 | 49.2994 | 24.75 | 45.8251 | 45.8162 | 14.3197 |
| 1.8339 | 2.0 | 237748 | 1.4449 | 49.6767 | 25.1673 | 46.1631 | 46.156 | 14.2557 |
| 1.8067 | 3.0 | 356622 | 1.4220 | 50.043 | 25.4886 | 46.4577 | 46.437 | 14.2857 |
| 1.8141 | 4.0 | 475496 | 1.4097 | 50.11 | 25.4327 | 46.502 | 46.5001 | 14.2653 |
| 1.7985 | 5.0 | 594370 | 1.4052 | 50.1555 | 25.5096 | 46.5771 | 46.5827 | 14.246 |
Framework versions
- Transformers 4.11.0.dev0
- Pytorch 1.8.1+cu101
- Datasets 1.12.1
- Tokenizers 0.10.3
