Improving QA Performance with NLI as a FR Intermediate Task
Collection
This is the model implementation of Intermediate Task Transfer Learning from NLI models to QA models with freezing layer -> train only output layer. • 9 items • Updated
fine-tuned-DatasetQAS-Squad-ID-with-indobert-large-p2-with-ITTL-without-freeze-LR-1e-05
This model is a fine-tuned version of indobenchmark/indobert-large-p2 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.5867
- Exact Match: 47.7296
- F1: 64.3850
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: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 128
- total_train_batch_size: 128
- 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 | Exact Match | F1 |
|---|---|---|---|---|---|
| 1.8839 | 0.5 | 463 | 1.7873 | 39.9512 | 56.0205 |
| 1.6682 | 1.0 | 926 | 1.6243 | 44.2651 | 60.9585 |
| 1.5129 | 1.5 | 1389 | 1.5722 | 45.6609 | 61.7661 |
| 1.4634 | 2.0 | 1852 | 1.5185 | 47.1493 | 63.5348 |
| 1.3128 | 2.5 | 2315 | 1.5212 | 46.9475 | 63.4277 |
| 1.323 | 3.0 | 2778 | 1.5052 | 47.6118 | 64.2591 |
| 1.1824 | 3.5 | 3241 | 1.5352 | 47.5950 | 64.2896 |
| 1.2013 | 4.0 | 3704 | 1.5302 | 47.9566 | 64.5453 |
| 1.0842 | 4.5 | 4167 | 1.5678 | 47.5362 | 64.2029 |
| 1.0811 | 5.0 | 4630 | 1.5590 | 47.7632 | 64.1309 |
| 1.0138 | 5.5 | 5093 | 1.5867 | 47.7296 | 64.3850 |
Framework versions
- Transformers 4.26.1
- Pytorch 1.13.1+cu117
- Datasets 2.2.0
- Tokenizers 0.13.2
- Downloads last month
- 2