twitter-data-xlm-roberta-base-sentiment-finetuned-memes-final
This model is a fine-tuned version of jayantapaul888/twitter-data-xlm-roberta-base-sentiment-finetuned-memes on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.5884
- Accuracy: 0.8310
- Precision: 0.8314
- Recall: 0.8310
- F1: 0.8311
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: 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: 6
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 |
|---|---|---|---|---|---|---|---|
| No log | 1.0 | 294 | 0.3981 | 0.8136 | 0.8185 | 0.8136 | 0.8132 |
| 0.4388 | 2.0 | 588 | 0.4114 | 0.8220 | 0.8275 | 0.8220 | 0.8221 |
| 0.4388 | 3.0 | 882 | 0.4203 | 0.8263 | 0.8285 | 0.8263 | 0.8266 |
| 0.2731 | 4.0 | 1176 | 0.4815 | 0.8235 | 0.8276 | 0.8235 | 0.8221 |
| 0.2731 | 5.0 | 1470 | 0.5090 | 0.8330 | 0.8335 | 0.8330 | 0.8332 |
| 0.1883 | 6.0 | 1764 | 0.5884 | 0.8310 | 0.8314 | 0.8310 | 0.8311 |
Framework versions
- Transformers 4.24.0.dev0
- Pytorch 1.11.0+cu102
- Datasets 2.6.1
- Tokenizers 0.13.1