<!-- 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. -->
distilbert-base-spanish-uncased-finetuned-opinion
This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 6.2078
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: 64
- eval_batch_size: 64
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss |
---|---|---|---|
12.7318 | 1.0 | 899 | 7.5402 |
7.8137 | 2.0 | 1798 | 7.0341 |
7.2367 | 3.0 | 2697 | 6.7184 |
6.9153 | 4.0 | 3596 | 6.5694 |
6.7086 | 5.0 | 4495 | 6.4376 |
6.5769 | 6.0 | 5394 | 6.3584 |
6.4833 | 7.0 | 6293 | 6.2887 |
6.4076 | 8.0 | 7192 | 6.2511 |
6.362 | 9.0 | 8091 | 6.2257 |
6.3296 | 10.0 | 8990 | 6.2064 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2