<!-- 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-uncased-sentiment-reddit-crypto
This model is a fine-tuned version of distilbert-base-uncased on the crypto-related reddit comments dataset. It achieves the following results on the evaluation set:
- Loss: 0.3070
- Accuracy: 0.8915
Accuracy on the final test set was: 0.8641
Training and evaluation data
Training and validation data collected from 2 sources:
- Kaggle reddit cryptocurrency posts and comments
- Kaggle reddit cryptocurrency related posts from various subreddits. Comments from subreddits:
'cryptocurrency', 'bitcoin', 'ethereum', 'dogecoin'were extracted.
Final test data source is from https://www.surgehq.ai/datasets/crypto-sentiment-dataset.
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 0.2823 | 1.0 | 5109 | 0.2658 | 0.8840 |
| 0.1905 | 2.0 | 10218 | 0.3070 | 0.8915 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.1+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2
