import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import AutoPeftModelForCausalLM
import gradio as gr

# Load the fine-tuned model and tokenizer
model_path = "BoburAmirov/test-llama-uz"  # Adjust this to the path where your fine-tuned model is saved

model = AutoPeftModelForCausalLM.from_pretrained(model_path, device_map='auto')
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)

# Ensure the tokenizer settings match those used during training
tokenizer.pad_token = tokenizer.eos_token
tokenizer.padding_side = "right"

# Set the model to evaluation mode
model.eval()

def generate_text(input_prompt):
    # Tokenize the input
    input_ids = tokenizer(input_prompt, return_tensors="pt")

    # Generate text
    with torch.no_grad():
        output = model.generate(
            input_ids,
            max_length=400,  # Adjust max_length as needed
            num_return_sequences=1,
            temperature=0.7,  # Control randomness
            top_p=0.9,  # Control diversity
            top_k=50,  # Control diversity
        )

    # Decode the generated text
    generated_text = tokenizer.decode(output[0], skip_special_tokens=True)
    return generated_text

# Create a Gradio interface
iface = gr.Interface(
    fn=generate_text,
    inputs=gr.inputs.Textbox(lines=2, placeholder="Enter your prompt here..."),
    outputs="text",
    title="Text Generation with LLaMA",
    description="Generate text using a fine-tuned LLaMA model."
)

if __name__ == "__main__":
    iface.launch(server_name="0.0.0.0", server_port=7860)