import gradio as gr
import torch
from transformers import CodeT5ForConditionalGeneration, CodeT5Tokenizer

# Load pre-trained CodeT5 model and tokenizer
model = CodeT5ForConditionalGeneration.from_pretrained("Salesforce/code-t5-small")
tokenizer = CodeT5Tokenizer.from_pretrained("Salesforce/code-t5-small")

def generate_code(prompt, code_file):
    # Read uploaded code file
    if code_file:
        code_text = code_file.read().decode("utf-8")
    else:
        code_text = ""

    # Tokenize input prompt
    input_ids = tokenizer.encode(prompt, return_tensors="pt")

    # Generate code using CodeT5 model
    output = model.generate(input_ids=input_ids, max_length=256)
    generated_code = tokenizer.decode(output[0], skip_special_tokens=True)

    # Return generated code and code preview
    return generated_code, f"```python\n{generated_code}\n```"

# Create Gradio interface
iface = gr.Interface(
    fn=generate_code,
    inputs=[
________gr.Textbox(label="Input_Prompt",_placeholder="Enter_a_prompt"),
________gr.Upload(label="Upload_Code_File",_file_types=["py"])
    ],
    outputs=[
________gr.Textbox(label="Generated_Code"),
________gr.Code(label="Code_Preview",_language="python")
____],
    title="Code Generation with CodeT5",
    description="Generate Python code based on input prompt and uploaded code file."
)

# Launch Gradio interface
iface.launch()