Pixie.30.32
This model is a fine-tuned version of t5-small on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.1623
- Bleu: 47.6437
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: 1e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 121
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 30
Training results
| Training Loss | Epoch | Step | Validation Loss | Bleu |
|---|---|---|---|---|
| No log | 9.09 | 100 | 1.5563 | 21.3462 |
| No log | 18.18 | 200 | 1.2493 | 29.2353 |
| No log | 27.27 | 300 | 1.1670 | 32.5700 |
Framework versions
- Transformers 4.18.0
- Pytorch 1.11.0a0+17540c5
- Datasets 2.1.0
- Tokenizers 0.12.1