app.py

import gradio as gr
import torch
from transformers import GPT2Tokenizer, GPT2LMHeadModel
import os

# Get the Hugging Face token from the environment variable
HF_TOKEN = os.environ.get("HF_TOKEN")

# Load the tokenizer and model
tokenizer = GPT2Tokenizer.from_pretrained('gpt2', use_auth_token=HF_TOKEN)
model = GPT2LMHeadModel.from_pretrained('skylersterling/TopicGPT', use_auth_token=HF_TOKEN)
model.eval()
model.to('cpu')

# Define the function that generates text from a prompt
def generate_text(prompt, temperature):

    print(prompt)
    
    prompt_with_eos = "#CONTEXT# " + prompt + " #TOPIC# "  # Add the string "EOS" to the end of the prompt
    input_tokens = tokenizer.encode(prompt_with_eos, return_tensors='pt')

    input_tokens = input_tokens.to('cpu')

    generated_text = prompt_with_eos  # Start with the initial prompt plus "EOS"
    prompt_length = len(generated_text)

    for _ in range(80):  # Adjust the range to control the number of tokens generated
        with torch.no_grad():
            outputs = model(input_tokens)
            predictions = outputs.logits[:, -1, :] / temperature
            next_token = torch.multinomial(torch.softmax(predictions, dim=-1), 1)
        
        input_tokens = torch.cat((input_tokens, next_token), dim=1)
        
        decoded_token = tokenizer.decode(next_token.item())
        generated_text += decoded_token  # Append the new token to the generated text
        if decoded_token == "#":  # Stop if the end of sequence token is generated
            break
        yield generated_text[prompt_length:]  # Yield the generated text excluding the initial prompt plus "EOS"

# Create a Gradio interface with a text input and a slider for temperature
interface = gr.Interface(
    fn=generate_text,
    inputs=[
        gr.Textbox(lines=2, placeholder="Enter your prompt here..."),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.3, label="Temperature"),
    ],
    outputs=gr.Textbox(),
    live=False,
    description="TopicGPT processes the input and returns a reasonably accurate guess of the topic/theme of a given conversation."
)

interface.launch()