DeBERTa is a fine-tuned version of DeBERTa model. The dataset used for fine tuning is the IMDB dataset from Hugging Face. The number of instances used for training is 12 500, and 2 500 for testing. Metrics used for evaluation are classification oriented, because the model is used for sentiment analysis. The number of epochs used for training is 5, and the learning rate was set to lr = 5e-5. The results obtained are: Accuracy: 0.9312 Precision: 0.9268 Recall: 0.9371 F1-score: 0.9319
- Downloads last month
- 0