app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM

from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model and tokenizer directly from Hugging Face Hub
model = AutoModelForCausalLM.from_pretrained("Manasa1/Llama-2-7b-chat-finetune")
tokenizer = AutoTokenizer.from_pretrained("Manasa1/Llama-2-7b-chat-finetune")


def generate_tweet():
    prompt = "Write a concise, creative tweet reflecting the style and personality in the fine-tuned dataset."
    # Tokenize the input prompt
    inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=100, padding=True)
    
    # Explicitly set the pad_token_id
    model.config.pad_token_id = model.config.eos_token_id

    # Generate the tweet with the attention mask
    outputs = model.generate(
        inputs["input_ids"],
        attention_mask=inputs["attention_mask"],  # Pass attention_mask explicitly
        max_length=140,
        num_return_sequences=1,
        top_p=0.8,
        temperature=0.6,
        repetition_penalty=1.2,  # Penalize repetition
    )
    # Decode and return the generated tweet
    generated_tweet = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return generated_tweet.strip()


# Gradio Interface
with gr.Blocks() as app:
    gr.Markdown("# AI Tweet Generator")
    gr.Markdown("Click the button below to generate a tweet reflecting the fine-tuned personality.")
    generate_button = gr.Button("Generate")
    output_box = gr.Textbox(label="Generated Tweet")
    
    generate_button.click(generate_tweet, inputs=None, outputs=output_box)

# Launch the app locally
app.launch()