A running version of the model is hosted on Huggingface.
Collection of Jupyter Notebooks focusing on the fine-tuning of T5 for text summarization and title generation. The idea is to generate a summary short enough to be a possible title. A version of the fine-tuned model is hosted on Huggingface and can be tested directly from the browser (hosted inference API). The fine-tuned model can be downloaded and used through the Huggingface API.
This model is a fine-tuned version of t5-small on CShorten/ML-ArXiv-Papers.
It achieves the following results on the evaluation set:
- Train Loss: 1.9469
- Train Lr: 0.0004
- Validation Loss: 1.8462
- Validation Lr: 0.0002
The following hyperparameters were used during training:
- optimizer: {'name': 'Adam', 'learning_rate': 0.00015378147, 'decay': 0.0, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False}
- training_precision: float32
| Train Loss | Train Lr | Validation Loss | Validation Lr | Epoch |
|---|---|---|---|---|
| 2.2534 | 0.0005 | 1.9839 | 0.0007 | 0 |
| 1.9469 | 0.0004 | 1.8462 | 0.0002 | 1 |
- Transformers 4.21.3
- TensorFlow 2.10.0
- Datasets 2.4.0
- Tokenizers 0.12.1