ChadGPT
This model is a fine-tuned version of GPT-2 on the minhalvp/islamqa dataset, which consists of Islamic question-answer pairs. It is designed for generating answers to Islamic questions.
License
Apache 2.0
Datasets
The model is fine-tuned on the minhalvp/islamqa dataset from Hugging Face.
Language
English
Metrics
- Perplexity: 170.31 (calculated on the validation set of the minhalvp/islamqa dataset)
- Training Loss: 2.3 (final training loss after fine-tuning)
Base Model
The model is based on the gpt2 architecture.
Pipeline Tag
text-generation
Library Name
transformers
Tags
GPT-2, Islamic, QA, Fine-tuned, Text Generation
Eval Results
The model achieved a training loss of approximately 2.3 after 3 epochs of fine-tuning.