Quantization made by Richard Erkhov.
finetuned-gpt2 - GGUF
Original model description:
library_name: transformers
tags: []
Model Card for Model ID
This model card documents a GPT-2 model fine-tuned on a dataset of multiple-choice medical exam questions. The model has been trained to generate answers based on a given question and corresponding options.
Model Details
Model Description
- Developed by: Jaimin Shah
- Model type: GPT-2 Language Model, fine-tuned for question-answering tasks
- Language(s) (NLP): English
- License: [More Information Needed]
- Finetuned from model [optional]: GPT-2 (openai-community/gpt2)
Uses
Direct Use
This model can be used to automatically generate answers for multiple-choice medical exam questions. It is suited for educational purposes, quiz generation, and training applications.
Training Details
Training Data
182,822 training samples
Preprocessing
The input was tokenized with Hugging Face's GPT-2 tokenizer. Data was formatted as a question followed by the answer options.
Training Hyperparameters
Number of Training Epochs: 20
Batch Size (Training): 64
Learning Rate: 1e-5
Mixed Precision (FP16): True
Evaluation
Batch Size (Evaluation): 64
Testing Data, Factors & Metrics
Testing Data: The model was evaluated on a held-out validation set of medical exam questions.
Factors: The model's ability to select the correct answer from the given options was the key evaluation factor.
Metrics: Cross-entropy loss was used to monitor performance during training, with validation loss used to evaluate generalization.
Results
Training Loss: Decreased from 1.8577 to 1.6961 over 20 epochs.
Validation Loss: Decreased from 1.8692 to 1.7394, demonstrating good generalization performance.
