<!-- 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. -->
gbert-large
This model is a fine-tuned version of deepset/gbert-large
It was fine-tuned on poetry from Projekt Gutenberg in order to do masked language modeling tasks in poetry generation (synonym creation for rythm and to find rhyming pairs)
- Loss: 2.1519
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: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 10
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 2.7515 | 1.0 | 2481 | 2.5493 |
| 2.5626 | 2.0 | 4962 | 2.4392 |
| 2.4839 | 3.0 | 7443 | 2.3692 |
| 2.4082 | 4.0 | 9924 | 2.3425 |
| 2.3109 | 5.0 | 12405 | 2.2934 |
| 2.2551 | 6.0 | 14886 | 2.2582 |
| 2.2154 | 7.0 | 17367 | 2.2062 |
| 2.2003 | 8.0 | 19848 | 2.1962 |
| 2.1616 | 9.0 | 22329 | 2.1991 |
| 2.1462 | 10.0 | 24810 | 2.1519 |
Framework versions
- Transformers 4.19.4
- Pytorch 1.11.0+cu113
- Datasets 2.2.2
- Tokenizers 0.12.1
