scenario-non-kd-from-scratch-data-hate_speech_filipino-model-xlm-roberta-base
This model is a fine-tuned version of xlm-roberta-base on the hate_speech_filipino dataset. It achieves the following results on the evaluation set:
- Loss: 1.0612
- Accuracy: 0.7247
- F1: 0.7257
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: 5e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 6969
Training results
Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 |
---|---|---|---|---|---|
No log | 0.32 | 100 | 0.6595 | 0.6307 | 0.6853 |
No log | 0.64 | 200 | 0.5676 | 0.7032 | 0.6620 |
No log | 0.96 | 300 | 0.5294 | 0.7358 | 0.7069 |
No log | 1.28 | 400 | 0.5112 | 0.7493 | 0.7084 |
0.585 | 1.6 | 500 | 0.5554 | 0.7283 | 0.7420 |
0.585 | 1.92 | 600 | 0.5201 | 0.7349 | 0.6679 |
0.585 | 2.24 | 700 | 0.5838 | 0.7361 | 0.7415 |
0.585 | 2.56 | 800 | 0.5693 | 0.7325 | 0.7421 |
0.585 | 2.88 | 900 | 0.5469 | 0.7517 | 0.7128 |
0.3954 | 3.19 | 1000 | 0.6406 | 0.7509 | 0.7361 |
0.3954 | 3.51 | 1100 | 0.5834 | 0.7401 | 0.7158 |
0.3954 | 3.83 | 1200 | 0.6038 | 0.7538 | 0.7324 |
0.3954 | 4.15 | 1300 | 0.7079 | 0.7436 | 0.7230 |
0.3954 | 4.47 | 1400 | 0.7422 | 0.7474 | 0.7182 |
0.2591 | 4.79 | 1500 | 0.6393 | 0.75 | 0.7307 |
0.2591 | 5.11 | 1600 | 0.7890 | 0.7481 | 0.7307 |
0.2591 | 5.43 | 1700 | 1.0788 | 0.7332 | 0.6651 |
0.2591 | 5.75 | 1800 | 0.8036 | 0.7353 | 0.7157 |
0.2591 | 6.07 | 1900 | 1.0868 | 0.7474 | 0.7167 |
0.1729 | 6.39 | 2000 | 1.3150 | 0.7441 | 0.7027 |
0.1729 | 6.71 | 2100 | 1.0097 | 0.7351 | 0.7268 |
0.1729 | 7.03 | 2200 | 1.0160 | 0.7389 | 0.7074 |
0.1729 | 7.35 | 2300 | 1.0612 | 0.7247 | 0.7257 |
Framework versions
- Transformers 4.33.3
- Pytorch 2.0.1
- Datasets 2.14.5
- Tokenizers 0.13.3
- Downloads last month
- 8
Finetuned from
Dataset used to train haryoaw/scenario-non-kd-from-scratch-data-hate_speech_filipino-model-xlm-roberta-base
Evaluation results
- Accuracy on hate_speech_filipinovalidation set self-reported0.725
- F1 on hate_speech_filipinovalidation set self-reported0.726