Hi! This is Ian Prazak's fine-tuned DeBERTa Large model for fact-checking.
- Dataset: The model was trained over 3 epochs with a dataset of approximately 30,000 formatted examples in JSONL; validation took place with approximately 8,000 other unique examples.
- Tokenizer: The base DeBERTa Large tokenizer was used.
- Metrics: The metrics listed above represent the final validation loss and accuracy before any signs of overfitting were observed.
Thank you for checking it out! :)
