---
license: apache-2.0
datasets:
- tatsu-lab/alpaca
base_model:
- distilbert/distilgpt2
tags:
- text-generation-inference
- fine-tuned
- alpaca
- pytorch
- instruction-following
language:
- en
---

# DistilGPT2 Fine-tuned Test

## Model Description

This repository contains a DistilGPT2 model that has been fine-tuned on the `tatsu-lab/alpaca` dataset. The Alpaca dataset is known for its collection of instruction-following examples, making this model capable of generating responses that adhere to given instructions.

The base model, DistilGPT2, is a distilled version of GPT-2, making it smaller and faster while retaining much of the original's language generation capabilities. By fine-tuning it on the Alpaca dataset, we aim to enhance its ability to understand and execute various natural language instructions.

## Intended Uses

This model is intended for text generation tasks, specifically those that involve following instructions or generating creative content based on a prompt. It can be used for:

* **Instruction Following:** Generating responses to specific commands or queries.
* **Creative Writing:** Assisting with story generation, poetry, or other forms of creative text.
* **Question Answering (Generative):** Providing answers to questions where the answer might require synthesis of information or generation of a coherent response.
* **Chatbots:** As a component in conversational AI systems for generating contextually relevant replies.

### Limitations and Bias

As with any large language model, this model may exhibit biases present in its training data (both DistilGPT2's original training data and the Alpaca dataset). It may also generate nonsensical, factually incorrect, or harmful content. It is crucial to monitor its outputs and use it responsibly. It should not be used for critical applications without thorough validation.

## How to Get Started

### Installation
Make sure you have the transformers library installed:
```python
pip install transformers torch
```

### Code Example
```python
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
import torch

# Specify the model ID on Hugging Face Hub
model_id = "your-username/my-fine-tuned-distilgpt2-alpaca" # <-- REPLACE WITH YOUR ACTUAL HUGGING FACE MODEL ID

# Determine the device (GPU if available, otherwise CPU)
device = 0 if torch.cuda.is_available() else -1

print(f"Loading model from Hugging Face Hub: {model_id}")

# Load the tokenizer and model
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

# Create a text generation pipeline
generator = pipeline('text-generation', model=model, tokenizer=tokenizer, device=device)

# Example 1: Simple instruction
prompt_1 = "Instruction: Write a short positive affirmation.\nInput: None\nOutput:"
print(f"\nPrompt 1: {prompt_1}")
output_1 = generator(prompt_1, max_new_tokens=30, num_return_sequences=1, do_sample=True, top_k=50, top_p=0.95)
print("Generated Output 1:")
print(output_1[0]['generated_text'])

# Example 2: More complex instruction with input
prompt_2 = "Instruction: Summarize the following text in one sentence.\nInput: Photosynthesis is the process by which green plants and some other organisms convert light energy into chemical energy. This chemical energy is stored in carbohydrate molecules, such as sugars, which are synthesized from carbon dioxide and water. Most life on Earth ultimately depends on photosynthesis as a source of energy.\nOutput:"
print(f"\nPrompt 2: {prompt_2}")
output_2 = generator(prompt_2, max_new_tokens=50, num_return_sequences=1, do_sample=True, top_k=50, top_p=0.95)
print("Generated Output 2:")
print(output_2[0]['generated_text'])

# Example 3: Creative writing prompt
prompt_3 = "Instruction: Continue the story:\nInput: A lone astronaut floated through the silent cosmos, gazing at the distant blue marble. Suddenly, a faint light shimmered...\nOutput:"
print(f"\nPrompt 3: {prompt_3}")
output_3 = generator(prompt_3, max_new_tokens=70, num_return_sequences=1, do_sample=True, top_k=50, top_p=0.95)
print("Generated Output 3:")
print(output_3[0]['generated_text'])
```

## Training Data

The model was fine-tuned using the **`tatsu-lab/alpaca`** dataset.
This dataset consists of 52K instruction-following demonstrations generated by OpenAI's `text-davinci-003` instruction-tuned language model. Each instance in the dataset is composed of an `instruction`, an optional `input`, and a desired `output`.

**Example from the Alpaca dataset:**

```json
{
  "instruction": "Identify the odd one out.",
  "input": "Lion, Tiger, Bear, Salmon",
  "output": "Salmon"
}
```