app.py

import gradio as gr
from transformers import pipeline
import torch

# Check if a GPU is available
device = 0 if torch.cuda.is_available() else -1

# Load the text-generation pipeline with the appropriate device
model = pipeline(
    "text-generation",
    model="rish13/polymers",
    device=device  # Automatically use GPU if available, otherwise CPU
)

def remove_duplicate_sentences(text):
    # Split text into sentences (assuming sentences are separated by periods)
    sentences = text.split('. ')
    unique_sentences = list(dict.fromkeys(sentences))  # Remove exact duplicates
    return '. '.join(unique_sentences)

def generate_response(prompt):
    # Generate text from the model
    response = model(
        prompt, 
        max_length=130,  # Adjusted to generate shorter text
        num_return_sequences=1, 
        temperature=0.7,  # Increased to add more randomness
        top_k=130,  # Increased to allow a wider selection of words
        top_p=0.95  # Slightly increased cumulative probability threshold
    )
    
    # Get the generated text from the response
    generated_text = response[0]['generated_text']
    
    # Post-process to remove duplicate sentences
    processed_text = remove_duplicate_sentences(generated_text)
    
    return processed_text

# Define the Gradio interface
interface = gr.Interface(
    fn=generate_response,
    inputs=gr.Textbox(lines=2, placeholder="Enter your prompt here...", label="Prompt"),
    outputs="text",
    title="Polymer Knowledge Model",
    description="A model fine-tuned for generating text related to polymers."
)

# Launch the interface
interface.launch()