|
--- |
|
license: apache-2.0 |
|
base_model: distilbert/distilgpt2 |
|
tags: |
|
- generated_from_trainer |
|
datasets: |
|
- eli5_category |
|
model-index: |
|
- name: gpt2-funetuned-eli5 |
|
results: [] |
|
language: |
|
- en |
|
metrics: |
|
- perplexity |
|
library_name: transformers |
|
pipeline_tag: text-generation |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
|
|
# gpt2-finetuned-eli5 |
|
|
|
This model is a fine-tuned version of [distilbert/distilgpt2](https://huggingface.co/distilbert/distilgpt2), fine-tuned on the `eli5_category` dataset. It has been trained to generate human-like responses to questions, specifically tailored to the Explain Like I'm 5 (ELI5) community. This model aims to provide clear and concise answers suitable for a general audience. |
|
|
|
## Model Description |
|
|
|
The `gpt2-finetuned-eli5` model is based on the DistilGPT-2 architecture, which is a smaller, faster, and more efficient version of GPT-2. It retains most of GPT-2's capabilities while being more computationally efficient. The model is particularly adept at generating text that resembles human-written responses, making it suitable for tasks involving natural language understanding and generation. |
|
|
|
### Key Features: |
|
- **Architecture**: DistilGPT-2, a distilled version of GPT-2. |
|
- **Purpose**: Generating clear and concise explanations suitable for general audiences, particularly in response to questions typical of the ELI5 community. |
|
- **Model Size**: Smaller and more efficient than the original GPT-2, with reduced computational requirements. |
|
|
|
## Intended Uses & Limitations |
|
|
|
### Intended Uses: |
|
- **Question Answering**: Provide simplified and easy-to-understand answers to a wide range of questions. |
|
- **Text Generation**: Generate coherent and contextually relevant text based on a given prompt. |
|
- **Educational Tools**: Assist in educational content creation by generating simple explanations of complex topics. |
|
- **Chatbots**: Improve the conversational abilities of chatbots by providing human-like responses. |
|
|
|
### Limitations: |
|
- **Simplification Risks**: While the model excels at providing simplified explanations, it might oversimplify or miss nuances, especially with complex topics. |
|
- **Dataset Bias**: The model's behavior reflects the data it was trained on. It might exhibit biases present in the training data, leading to inappropriate or biased responses. |
|
- **Factually Inaccurate Responses**: The model does not have real-time access to factual databases, and its knowledge is based on the data it was trained on. It might produce outdated or incorrect information. |
|
- **Limited Knowledge Cut-off**: The model's training data only includes information up to a certain date, and it does not know about events or developments beyond that time. |
|
|
|
## Training and Evaluation Data |
|
|
|
### Training Data: |
|
- **Dataset**: The model was fine-tuned on the `eli5_category` dataset, which consists of questions and answers from the Explain Like I'm 5 (ELI5) community. This dataset contains a variety of topics where users seek simple and clear explanations. |
|
|
|
### Evaluation Data: |
|
- The evaluation data consisted of a subset of the ELI5 dataset that was held out during training. The model's performance was assessed based on its ability to generate coherent and contextually appropriate responses. |
|
|
|
## Training Procedure |
|
|
|
### Training Hyperparameters: |
|
- **Learning Rate**: 2e-05 |
|
- **Train Batch Size**: 8 |
|
- **Eval Batch Size**: 8 |
|
- **Seed**: 42 |
|
- **Optimizer**: Adam with betas=(0.9, 0.999) and epsilon=1e-08 |
|
- **Learning Rate Scheduler Type**: Linear |
|
- **Number of Epochs**: 3.0 |
|
|
|
### Training Results: |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | |
|
|:-------------:|:-----:|:----:|:---------------:| |
|
| 3.8522 | 1.0 | 1289 | 3.8307 | |
|
| 3.8093 | 2.0 | 2578 | 3.8280 | |
|
| 3.7661 | 3.0 | 3867 | 3.8269 | |
|
|
|
- The model achieved a final validation loss of 3.8269, indicating a consistent improvement in training performance. |
|
|
|
### Framework Versions: |
|
- **Transformers**: 4.42.4 |
|
- **PyTorch**: 2.3.1+cu121 |
|
- **Datasets**: 2.21.0 |
|
- **Tokenizers**: 0.19.1 |
|
|
|
## Ethical Considerations |
|
|
|
- **Bias and Fairness**: The model's responses might reflect biases present in the training data. Users should be aware of potential biases and verify the information generated. |
|
- **Privacy**: The model was trained on publicly available data. However, care should be taken to avoid using the model for generating content that may violate privacy norms. |
|
|
|
## Example Usage |
|
|
|
To generate text using the `gpt2-finetuned-eli5` model, you can use the following code: |
|
|
|
```python |
|
from transformers import pipeline |
|
|
|
# Load the text generation pipeline |
|
generator = pipeline("text-generation", model="ashaduzzaman/gpt2-funetuned-eli5") |
|
|
|
# Provide a prompt |
|
prompt = "Somatic hypermutation allows the immune system to" |
|
|
|
# Generate text |
|
generator(prompt) |
|
``` |
|
|
