| language: en | |
| license: mit | |
| tags: | |
| - financial-qa | |
| - distilgpt2 | |
| - fine-tuned | |
| datasets: | |
| - financial-qa | |
| metrics: | |
| - perplexity | |
| # Financial QA Fine-Tuned Model | |
| This model is a fine-tuned version of `distilgpt2` on financial question-answering data from Allstate's financial reports. | |
| ## Model description | |
| The model was fine-tuned to answer questions about Allstate's financial reports and performance. | |
| ## Intended uses & limitations | |
| This model is intended to be used for answering factual questions about Allstate's financial reports for 2022-2023. | |
| It should not be used for financial advice or decision-making without verification from original sources. | |
| ## Training data | |
| The model was trained on a custom dataset of financial QA pairs derived from Allstate's 10-K reports. | |
| ## Training procedure | |
| The model was fine-tuned using the `Trainer` class from Hugging Face's Transformers library with the following parameters: | |
| - Learning rate: default | |
| - Batch size: 2 | |
| - Number of epochs: 3 | |
| ## Evaluation results | |
| The model achieved a final training loss of 0.44 and validation loss of 0.43. | |
| ## Limitations and bias | |
| This model has limited knowledge only of Allstate's financial data and cannot answer questions about other companies or financial topics outside its training data. | |