File size: 2,664 Bytes
0040626 159c5e2 0040626 b595fa7 93c3822 0040626 b595fa7 0040626 b595fa7 6bb18d3 b595fa7 6bb18d3 b6e81a2 6bb18d3 b6e81a2 6bb18d3 b6e81a2 b595fa7 0040626 b595fa7 159c5e2 0040626 6bb18d3 b595fa7 0040626 159c5e2 0040626 159c5e2 0040626 159c5e2 0040626 b595fa7 0040626 b595fa7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 |
from transformers import AutoModelForCausalLM, AutoTokenizer
import gradio as gr
import torch
import tempfile # ✅ Import tempfile to create temp files
# ✅ Load the fastest model on CPU
model_name = "Salesforce/codegen-350M-mono" # Fastest model for code review
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name).to("cpu") # Force CPU mode
def review_code(code_snippet):
print("✅ Received Code:", code_snippet) # Debugging log
# ✅ Add a clear instruction to the model
prompt = f"### Instruction: Review and fix the Python code below.\n### Input Code:\n{code_snippet}\n### Fixed Code:\n"
# Process input
inputs = tokenizer(prompt, return_tensors="pt").to("cpu") # Move to CPU
outputs = model.generate(
**inputs,
max_length=80, # ✅ Increased length for full function review
do_sample=False,
num_beams=5, # ✅ Higher beams for better decision-making
repetition_penalty=1.8 # ✅ Lower penalty to avoid weird token removal
)
# Check if the model generated output
if outputs is None:
print("❌ Model did not generate output!") # Debugging log
return "Error: Model did not generate output."
reviewed_code = tokenizer.decode(outputs[0], skip_special_tokens=True)
print("✅ Generated Code:", reviewed_code) # Debugging log
# ✅ Write reviewed code to a temporary file for download
temp_file_path = tempfile.NamedTemporaryFile(delete=False, suffix=".txt").name
with open(temp_file_path, "w") as temp_file:
temp_file.write(reviewed_code)
return reviewed_code, temp_file_path # ✅ Return reviewed code & file path
# ✅ Handle user input and return reviewed code
def check_code(input_code):
reviewed_code, file_path = review_code(input_code)
return input_code, reviewed_code, file_path # ✅ Correctly return file path
# ✅ Gradio UI with Side-by-Side Comparison & Fixed Download Option
interface = gr.Interface(
fn=check_code,
inputs=gr.Textbox(label="Enter Python Code"),
outputs=[
gr.Textbox(label="Original Code", interactive=False), # Left side
gr.Textbox(label="Reviewed Code", interactive=False), # Right side
gr.File(label="Download Reviewed Code") # ✅ Fixed Download Button
],
title="🚀 AI Code Reviewer",
description="📌 Enter Python code and get a reviewed version. Download the reviewed code as a file.",
allow_flagging="never"
)
# ✅ Launch app (Fixes font issues and removes `share=True`)
interface.launch(server_name="0.0.0.0", server_port=7860, show_error=True)
|