wikihow-t5-small-finetuned-pubmed
This model is a fine-tuned version of deep-learning-analytics/wikihow-t5-small on the pub_med_summarization_dataset dataset. It achieves the following results on the evaluation set:
- Loss: 2.2702
- Rouge1: 8.9619
- Rouge2: 3.2719
- Rougel: 8.1558
- Rougelsum: 8.5714
- Gen Len: 19.0
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: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
|---|---|---|---|---|---|---|---|---|
| 2.5984 | 1.0 | 4000 | 2.3696 | 10.237 | 3.8609 | 8.9776 | 9.677 | 19.0 |
| 2.5677 | 2.0 | 8000 | 2.3132 | 9.302 | 3.4499 | 8.3816 | 8.8831 | 19.0 |
| 2.5038 | 3.0 | 12000 | 2.2884 | 9.0578 | 3.3103 | 8.23 | 8.6723 | 19.0 |
| 2.4762 | 4.0 | 16000 | 2.2758 | 9.0001 | 3.2882 | 8.1845 | 8.6084 | 19.0 |
| 2.4393 | 5.0 | 20000 | 2.2702 | 8.9619 | 3.2719 | 8.1558 | 8.5714 | 19.0 |
Framework versions
- Transformers 4.17.0
- Pytorch 1.10.0+cu111
- Datasets 1.18.3
- Tokenizers 0.11.6