README.md

---
datasets:
- knkarthick/dialogsum
language:
- en
metrics:
- accuracy
pipeline_tag: text-generation
library_name: "torch, transformers"
---

Certainly! Here's a short README for using the pre-trained `distilgpt2` model for chatting:

---

# DistilGPT-2 Chatbot

This project demonstrates how to use the pre-trained `distilgpt2` model from Hugging Face for creating a simple chatbot. It includes code for loading the model, generating responses, and running an interactive conversation loop.

## Prerequisites

Ensure you have the following libraries installed:

```bash
pip install transformers torch
```

## Usage

1. **Load the Pre-trained Model and Tokenizer**

   ```python
   from transformers import GPT2LMHeadModel, GPT2Tokenizer

   model_name = "distilgpt2"
   model = GPT2LMHeadModel.from_pretrained(model_name)
   tokenizer = GPT2Tokenizer.from_pretrained(model_name)
   ```

2. **Generate a Response**

   Use the following function to generate a response based on user input:

   ```python
   def generate_response(prompt, max_length=100):
       input_ids = tokenizer.encode(prompt, return_tensors='pt')
       output = model.generate(
           input_ids,
           max_length=max_length,
           pad_token_id=tokenizer.eos_token_id,
           no_repeat_ngram_size=2,
           num_return_sequences=1,
           temperature=0.7,
           top_p=0.9,
           top_k=50
       )
       response = tokenizer.decode(output[0], skip_special_tokens=True)
       return response
   ```

3. **Interactive Conversation Loop**

   Run the following code to start a chat session:

   ```python
   while True:
       user_input = input("You: ")
       prompt = f"<user> {user_input}<AI>"
       response = generate_response(prompt)
       print(f"AI: {response}")

       if user_input.lower() in ["exit", "quit"]:
           break
   ```

## Configuration

- **Temperature**: Controls randomness. Lower values are more deterministic.
- **Top-p and top-k**: Limit word selection for balanced diversity and coherence.
- **Max_length**: Limits the length of the response.