vijayvizag's picture
readme update
bcb80f2
raw
history blame
15.3 kB
import streamlit as st
import torch
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import re
import time
# Model constants
CODET5_MODEL = "Salesforce/codet5-base-multi-sum"
class CodeT5Summarizer:
def __init__(self, device=None):
"""Initialize CodeT5 summarization model."""
self.device = device if device else ('cuda' if torch.cuda.is_available() else 'cpu')
# Initialize model and tokenizer
with st.spinner("Loading CodeT5 model... this may take a minute..."):
self.tokenizer = AutoTokenizer.from_pretrained(CODET5_MODEL)
self.model = AutoModelForSeq2SeqLM.from_pretrained(CODET5_MODEL).to(self.device)
def preprocess_code(self, code):
"""Clean and preprocess the Python code."""
# Remove empty lines
code = re.sub(r'\n\s*\n', '\n', code)
# Remove excessive comments (keeping docstrings)
code_lines = []
in_docstring = False
docstring_delimiter = None
for line in code.split('\n'):
# Check for docstring delimiters
if '"""' in line or "'''" in line:
delimiter = '"""' if '"""' in line else "'''"
if not in_docstring:
in_docstring = True
docstring_delimiter = delimiter
elif docstring_delimiter == delimiter:
in_docstring = False
docstring_delimiter = None
# Keep docstrings and non-comment lines
if in_docstring or not line.strip().startswith('#'):
code_lines.append(line)
processed_code = '\n'.join(code_lines)
# Normalize whitespace
processed_code = re.sub(r' +', ' ', processed_code)
return processed_code
def extract_functions(self, code):
"""Extract individual functions for summarization"""
# Simple regex to find function definitions
function_pattern = r'def\s+([a-zA-Z_][a-zA-Z0-9_]*)\s*\(.*?\).*?:'
function_matches = re.finditer(function_pattern, code, re.DOTALL)
functions = []
for match in function_matches:
start_pos = match.start()
# Find the function body
function_name = match.group(1)
lines = code[start_pos:].split('\n')
# Skip the function definition line
body_start = 1
while body_start < len(lines) and not lines[body_start].strip():
body_start += 1
if body_start < len(lines):
# Get the indentation of the function body
body_indent = len(lines[body_start]) - len(lines[body_start].lstrip())
# Gather all lines with at least this indentation
function_body = [lines[0]] # The function definition
i = 1
while i < len(lines):
line = lines[i]
if line.strip() and (len(line) - len(line.lstrip())) < body_indent and not line.strip().startswith('#'):
break
function_body.append(line)
i += 1
function_code = '\n'.join(function_body)
functions.append((function_name, function_code))
# Simple regex to find class methods
class_pattern = r'class\s+([a-zA-Z_][a-zA-Z0-9_]*)'
class_matches = re.finditer(class_pattern, code, re.DOTALL)
for match in class_matches:
class_name = match.group(1)
start_pos = match.start()
# Find class methods using the function pattern
class_code = code[start_pos:]
method_matches = re.finditer(function_pattern, class_code, re.DOTALL)
for method_match in method_matches:
method_name = method_match.group(1)
# Skip if this is not a method (i.e., it's a function outside the class)
if method_match.start() > 200: # Simple heuristic to check if method is within class scope
break
# Get the full method code
method_start = method_match.start()
method_lines = class_code[method_start:].split('\n')
# Skip the method definition line
body_start = 1
while body_start < len(method_lines) and not method_lines[body_start].strip():
body_start += 1
if body_start < len(method_lines):
# Get the indentation of the method body
body_indent = len(method_lines[body_start]) - len(method_lines[body_start].lstrip())
# Gather all lines with at least this indentation
method_body = [method_lines[0]] # The method definition
i = 1
while i < len(method_lines):
line = method_lines[i]
if line.strip() and (len(line) - len(line.lstrip())) < body_indent and not line.strip().startswith('#'):
break
method_body.append(line)
i += 1
method_code = '\n'.join(method_body)
functions.append((f"{class_name}.{method_name}", method_code))
return functions
def extract_classes(self, code):
"""Extract class definitions for summarization"""
class_pattern = r'class\s+([a-zA-Z_][a-zA-Z0-9_]*)'
class_matches = re.finditer(class_pattern, code, re.DOTALL)
classes = []
for match in class_matches:
class_name = match.group(1)
start_pos = match.start()
# Extract class body
class_lines = code[start_pos:].split('\n')
# Skip the class definition line
body_start = 1
while body_start < len(class_lines) and not class_lines[body_start].strip():
body_start += 1
if body_start < len(class_lines):
# Get the indentation of the class body
body_indent = len(class_lines[body_start]) - len(class_lines[body_start].lstrip())
# Gather all lines with at least this indentation
class_body = [class_lines[0]] # The class definition
i = 1
while i < len(class_lines):
line = class_lines[i]
if line.strip() and (len(line) - len(line.lstrip())) < body_indent:
break
class_body.append(line)
i += 1
class_code = '\n'.join(class_body)
classes.append((class_name, class_code))
return classes
def summarize(self, code, max_length=50):
"""Generate summary using CodeT5."""
# Truncate input if needed
max_input_length = 512 # CodeT5 typically accepts up to 512 tokens
tokenized_code = self.tokenizer(code, truncation=True, max_length=max_input_length, return_tensors="pt").to(self.device)
with torch.no_grad():
generated_ids = self.model.generate(
tokenized_code["input_ids"],
max_length=max_length,
num_beams=4,
early_stopping=True
)
summary = self.tokenizer.decode(generated_ids[0], skip_special_tokens=True)
return summary
def summarize_code(self, code, summarize_functions=True, summarize_classes=True):
"""
Generate full file summary and optionally function/class level summaries.
Returns a dictionary with summaries.
"""
preprocessed_code = self.preprocess_code(code)
results = {
"file_summary": None,
"function_summaries": {},
"class_summaries": {}
}
# Generate file-level summary
try:
file_summary = self.summarize(preprocessed_code)
results["file_summary"] = file_summary
except Exception as e:
results["file_summary"] = f"Error generating file summary: {str(e)}"
# Generate function-level summaries if requested
if summarize_functions:
functions = self.extract_functions(preprocessed_code)
for function_name, function_code in functions:
try:
summary = self.summarize(function_code)
results["function_summaries"][function_name] = summary
except Exception as e:
results["function_summaries"][function_name] = f"Error: {str(e)}"
# Generate class-level summaries if requested
if summarize_classes:
classes = self.extract_classes(preprocessed_code)
for class_name, class_code in classes:
try:
summary = self.summarize(class_code)
results["class_summaries"][class_name] = summary
except Exception as e:
results["class_summaries"][class_name] = f"Error: {str(e)}"
return results
def main():
st.set_page_config(
page_title="Python Code Summarizer",
page_icon="πŸ“",
layout="wide"
)
st.title("πŸ“ Python Code Summarizer using CodeT5")
st.markdown("""
Upload a Python file or paste code directly to generate summaries.
This app uses CodeT5, a pretrained model for code understanding and generation.
""")
# Initialize session state
if 'summarizer' not in st.session_state:
st.session_state.summarizer = None
# Load model if not already loaded
if st.session_state.summarizer is None:
st.session_state.summarizer = CodeT5Summarizer()
# Create tabs for different input methods
tab1, tab2 = st.tabs(["Upload Python File", "Paste Code"])
with tab1:
uploaded_file = st.file_uploader("Choose a Python file", type=['py'])
if uploaded_file is not None:
code = uploaded_file.getvalue().decode('utf-8')
with st.expander("View Uploaded Code", expanded=False):
st.code(code, language='python')
# Add summarization options
st.subheader("Summarization Options")
col1, col2 = st.columns(2)
with col1:
summarize_functions = st.checkbox("Generate function summaries", value=True)
with col2:
summarize_classes = st.checkbox("Generate class summaries", value=True)
if st.button("Summarize Code", key="summarize_file"):
with st.spinner("Generating summaries..."):
start_time = time.time()
summaries = st.session_state.summarizer.summarize_code(
code,
summarize_functions=summarize_functions,
summarize_classes=summarize_classes
)
end_time = time.time()
# Display summaries
st.success(f"Summarization completed in {end_time - start_time:.2f} seconds!")
# File summary
st.subheader("File Summary")
st.write(summaries["file_summary"])
# Function summaries
if summarize_functions and summaries["function_summaries"]:
st.subheader("Function Summaries")
for func_name, summary in summaries["function_summaries"].items():
with st.expander(f"Function: {func_name}"):
st.write(summary)
# Class summaries
if summarize_classes and summaries["class_summaries"]:
st.subheader("Class Summaries")
for class_name, summary in summaries["class_summaries"].items():
with st.expander(f"Class: {class_name}"):
st.write(summary)
with tab2:
code = st.text_area("Paste Python code here", height=300)
if code:
# Add summarization options
st.subheader("Summarization Options")
col1, col2 = st.columns(2)
with col1:
summarize_functions = st.checkbox("Generate function summaries", value=True, key="func_paste")
with col2:
summarize_classes = st.checkbox("Generate class summaries", value=True, key="class_paste")
if st.button("Summarize Code", key="summarize_paste"):
with st.spinner("Generating summaries..."):
start_time = time.time()
summaries = st.session_state.summarizer.summarize_code(
code,
summarize_functions=summarize_functions,
summarize_classes=summarize_classes
)
end_time = time.time()
# Display summaries
st.success(f"Summarization completed in {end_time - start_time:.2f} seconds!")
# File summary
st.subheader("File Summary")
st.write(summaries["file_summary"])
# Function summaries
if summarize_functions and summaries["function_summaries"]:
st.subheader("Function Summaries")
for func_name, summary in summaries["function_summaries"].items():
with st.expander(f"Function: {func_name}"):
st.write(summary)
# Class summaries
if summarize_classes and summaries["class_summaries"]:
st.subheader("Class Summaries")
for class_name, summary in summaries["class_summaries"].items():
with st.expander(f"Class: {class_name}"):
st.write(summary)
st.markdown("---")
st.markdown("### About")
st.markdown("""
This app uses the CodeT5 model to generate summaries of Python code. The model is trained on a large corpus of code and documentation.
**Features:**
- File-level summaries
- Function-level summaries
- Class-level summaries
**Limitations:**
- Summaries may not always be accurate
- Long files may be truncated
- Complex code structures might not be properly understood
""")
if __name__ == "__main__":
main()