bigbird-base-health-fact
This model is a fine-tuned version of google/bigbird-roberta-base on the health_fact dataset. It achieves the following results on the evaluation set:
- Loss: 0.5864
- Micro F1: 0.8130
- Macro F1: 0.6874
- False F1: 0.8114
- Mixture F1: 0.4557
- True F1: 0.9154
- Unproven F1: 0.5672
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: 8
- eval_batch_size: 32
- seed: 18
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-06
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 3
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Micro F1 | Macro F1 | False F1 | Mixture F1 | True F1 | Unproven F1 |
---|---|---|---|---|---|---|---|---|---|
0.5563 | 1.0 | 1226 | 0.5020 | 0.7949 | 0.6062 | 0.7926 | 0.4591 | 0.8986 | 0.2745 |
0.5048 | 2.0 | 2452 | 0.4969 | 0.8180 | 0.6846 | 0.8202 | 0.4342 | 0.9126 | 0.5714 |
0.3454 | 3.0 | 3678 | 0.5864 | 0.8130 | 0.6874 | 0.8114 | 0.4557 | 0.9154 | 0.5672 |
Framework versions
- Transformers 4.19.0.dev0
- Pytorch 1.11.0a0+17540c5
- Datasets 2.1.1.dev0
- Tokenizers 0.12.1
- Downloads last month
- 41