app.py

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

# 🔹 Load tokenizer and base model
base_model_id = "deepseek-ai/deepseek-coder-1.3b-base"
lora_model_id = "brijmansuriya/deepseek-lora"  # ✅ Your LoRA fine-tuned model repo

tokenizer = AutoTokenizer.from_pretrained(lora_model_id, trust_remote_code=True)
base_model = AutoModelForCausalLM.from_pretrained(
    base_model_id,
    device_map="auto",
    torch_dtype=torch.float16,
    trust_remote_code=True
)

# 🔹 Load LoRA adapter
model = PeftModel.from_pretrained(base_model, lora_model_id)

# 🔹 Define the function
def generate_code(prompt):
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    outputs = model.generate(
        **inputs,
        max_new_tokens=200,
        temperature=0.7,
        do_sample=True,
        top_p=0.95,
    )
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

# 🔹 Gradio UI
demo = gr.Interface(
    fn=generate_code,
    inputs=gr.Textbox(label="Enter your coding prompt"),
    outputs=gr.Textbox(label="Generated Code"),
    title="🤖 DeepSeek Code Generator (LoRA)",
    description="This app uses DeepSeek-Coder with Brijbhai's fine-tuned LoRA model to generate code from natural language prompts."
)

demo.launch()