Spaces:
Sleeping
Sleeping
File size: 19,173 Bytes
5c6379b | 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 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 | #!/usr/bin/env python3
"""
Enhanced Code Analyzer with Fine-tuned Model Support
This version supports:
- Base CodeT5+ model (original)
- Fine-tuned DeepSeek Coder model (new)
- Easy toggle between models
- All existing optimizations (caching, streaming, etc.)
Author: AI Code Analyzer Project
Date: 2025
"""
import torch
import time
import hashlib
import json
import os
from typing import Dict, Any, Optional, Generator, Literal
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, AutoModelForCausalLM, BitsAndBytesConfig
from peft import PeftModel
from tqdm import tqdm
import streamlit as st
ModelType = Literal["codet5", "deepseek-finetuned"]
class EnhancedCodeAnalyzer:
"""
Enhanced analyzer supporting multiple model types including fine-tuned models.
"""
def __init__(
self,
model_type: ModelType = "codet5",
model_id: Optional[str] = None,
adapter_path: Optional[str] = None,
cache_dir: str = "./cache",
precision: str = "fp16",
quick_max_new_tokens: int = 180,
detailed_max_new_tokens: int = 300,
):
"""
Initialize the enhanced analyzer.
Args:
model_type: Type of model to use ("codet5" or "deepseek-finetuned")
model_id: Hugging Face model ID (auto-selected if None)
adapter_path: Path to LoRA adapters for fine-tuned models
cache_dir: Directory to store cached results
precision: Model precision (fp16, int8, int4)
quick_max_new_tokens: Max tokens for quick analysis
detailed_max_new_tokens: Max tokens for detailed analysis
"""
self.model_type = model_type
self.cache_dir = cache_dir
self.precision = precision.lower().strip()
self.quick_max_new_tokens = quick_max_new_tokens
self.detailed_max_new_tokens = detailed_max_new_tokens
# Auto-select model_id based on type
if model_id is None:
if model_type == "codet5":
self.model_id = "Salesforce/codet5p-220m"
elif model_type == "deepseek-finetuned":
self.model_id = "deepseek-ai/deepseek-coder-1.3b-instruct"
else:
raise ValueError(f"Unknown model_type: {model_type}")
else:
self.model_id = model_id
# Set adapter path
self.adapter_path = adapter_path
if model_type == "deepseek-finetuned" and adapter_path is None:
self.adapter_path = "./fine-tuned-analyst"
self.model = None
self.tokenizer = None
self.cache = {}
# Create cache directory
os.makedirs(cache_dir, exist_ok=True)
# Load cache if exists
self._load_cache()
def _create_quantization_config(self) -> BitsAndBytesConfig:
"""Create quantization configuration."""
return BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_quant_type="nf4",
bnb_4bit_compute_dtype=torch.float16,
bnb_4bit_use_double_quant=True,
)
def _load_model(self):
"""Load the model with optimizations."""
if self.model is not None:
return
print(f"π Loading {self.model_type} model...")
# Load tokenizer
self.tokenizer = AutoTokenizer.from_pretrained(self.model_id, trust_remote_code=True)
if self.tokenizer.pad_token is None:
self.tokenizer.pad_token = self.tokenizer.eos_token
# Configure precision
quantization_config = None
dtype = None
banner = ""
if self.precision == "fp16":
dtype = torch.float16
banner = "FP16 precision"
elif self.precision == "int8":
quantization_config = BitsAndBytesConfig(load_in_8bit=True)
banner = "INT8 quantization"
elif self.precision == "int4":
quantization_config = self._create_quantization_config()
banner = "INT4 (nf4) quantization"
else:
dtype = torch.float16
banner = f"Unknown precision '{self.precision}', defaulting to FP16"
# Load base model based on type
if self.model_type == "codet5":
self.model = AutoModelForSeq2SeqLM.from_pretrained(
self.model_id,
device_map="auto",
torch_dtype=dtype,
quantization_config=quantization_config,
trust_remote_code=True,
)
print(f"β
CodeT5+ loaded with {banner}!")
elif self.model_type == "deepseek-finetuned":
# Load base DeepSeek model
base_model = AutoModelForCausalLM.from_pretrained(
self.model_id,
device_map="auto",
torch_dtype=dtype,
quantization_config=quantization_config,
trust_remote_code=True,
)
# Load LoRA adapters if path exists
if self.adapter_path and os.path.exists(self.adapter_path):
print(f"π Loading LoRA adapters from {self.adapter_path}...")
self.model = PeftModel.from_pretrained(base_model, self.adapter_path)
print(f"β
Fine-tuned DeepSeek loaded with {banner}!")
else:
print(f"β οΈ Adapter path not found: {self.adapter_path}")
print(f"β
Using base DeepSeek model with {banner}")
self.model = base_model
def _create_prompt(self, code: str) -> str:
"""Create prompt based on model type."""
if self.model_type == "codet5":
return f"""Analyze this code for bugs, performance issues, and security concerns:
{code}
Analysis:"""
elif self.model_type == "deepseek-finetuned":
return f"""<s>[INST] Analyze this code for bugs, performance, and security issues. Give a quality score from 1-100 and provide a detailed analysis.
Code:
```
{code}
``` [/INST]"""
return code
def _extract_response(self, full_text: str, prompt: str) -> str:
"""Extract the actual response from generated text."""
if self.model_type == "codet5":
# Remove prompt from response
return full_text[len(prompt):].strip()
elif self.model_type == "deepseek-finetuned":
# Extract response after [/INST]
if '[/INST]' in full_text:
return full_text.split('[/INST]')[-1].strip()
return full_text.strip()
return full_text.strip()
def _get_cache_key(self, code: str) -> str:
"""Generate cache key for code."""
# Include model type in cache key
combined = f"{self.model_type}:{code}"
return hashlib.md5(combined.encode()).hexdigest()
def _load_cache(self):
"""Load cached results from disk."""
cache_file = os.path.join(self.cache_dir, "analysis_cache_enhanced.json")
if os.path.exists(cache_file):
try:
with open(cache_file, 'r') as f:
self.cache = json.load(f)
print(f"π Loaded {len(self.cache)} cached analyses")
except:
self.cache = {}
def _save_cache(self):
"""Save cache to disk."""
cache_file = os.path.join(self.cache_dir, "analysis_cache_enhanced.json")
with open(cache_file, 'w') as f:
json.dump(self.cache, f)
def _check_cache(self, code: str) -> Optional[Dict[str, Any]]:
"""Check if analysis is cached."""
cache_key = self._get_cache_key(code)
return self.cache.get(cache_key)
def _save_to_cache(self, code: str, result: Dict[str, Any]):
"""Save analysis result to cache."""
cache_key = self._get_cache_key(code)
self.cache[cache_key] = result
self._save_cache()
def analyze_code_streaming(
self,
code: str,
show_progress: bool = True,
mode: str = "detailed",
) -> Generator[str, None, Dict[str, Any]]:
"""
Analyze code with streaming response and progress indicators.
Args:
code: Code to analyze
show_progress: Whether to show progress indicators
mode: Analysis mode ("quick" or "detailed")
Yields:
str: Partial analysis results
"""
# Check cache first
cached_result = self._check_cache(code)
if cached_result:
print("β‘ Using cached result!")
yield cached_result["analysis"]
return cached_result
# Load model if not loaded
self._load_model()
# Create prompt
prompt = self._create_prompt(code)
# Tokenize input
inputs = self.tokenizer(
prompt,
return_tensors="pt",
truncation=True,
max_length=1024 if self.model_type == "deepseek-finetuned" else 512,
padding=True,
)
device = next(self.model.parameters()).device
inputs = {k: v.to(device) for k, v in inputs.items()}
# Generate analysis
start_time = time.time()
if show_progress:
print(f"π Analyzing code with {self.model_type}...")
progress_bar = tqdm(total=100, desc="Analysis Progress")
try:
with torch.no_grad():
max_new = self.detailed_max_new_tokens if mode == "detailed" else self.quick_max_new_tokens
# Generation parameters based on model type
if self.model_type == "codet5":
num_beams = 2 if mode == "detailed" else 1
outputs = self.model.generate(
inputs["input_ids"],
attention_mask=inputs.get("attention_mask"),
max_new_tokens=max_new,
num_beams=num_beams,
do_sample=False,
pad_token_id=self.tokenizer.eos_token_id,
use_cache=True,
)
else: # deepseek-finetuned
outputs = self.model.generate(
**inputs,
max_new_tokens=max_new,
temperature=0.7,
do_sample=True,
pad_token_id=self.tokenizer.eos_token_id,
eos_token_id=self.tokenizer.eos_token_id,
)
if show_progress:
progress_bar.update(50)
# Decode analysis
full_text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
analysis_text = self._extract_response(full_text, prompt)
if show_progress:
progress_bar.update(50)
progress_bar.close()
# Calculate quality score
quality_score = self._calculate_quality_score(analysis_text)
total_time = time.time() - start_time
# Create result
result = {
"analysis": analysis_text,
"quality_score": quality_score,
"execution_time": total_time,
"model": self.model_id,
"model_type": self.model_type,
"cached": False
}
# Save to cache
self._save_to_cache(code, result)
# Yield the analysis
yield analysis_text
return result
except Exception as e:
if show_progress:
progress_bar.close()
raise e
def analyze_code_fast(self, code: str, mode: str = "quick") -> Dict[str, Any]:
"""
Fast analysis without streaming (for batch processing).
Args:
code: Code to analyze
mode: Analysis mode ("quick" or "detailed")
Returns:
Dict: Analysis result
"""
# Check cache first
cached_result = self._check_cache(code)
if cached_result:
cached_result["cached"] = True
return cached_result
# Load model if not loaded
self._load_model()
# Create prompt
prompt = self._create_prompt(code)
# Tokenize input
inputs = self.tokenizer(
prompt,
return_tensors="pt",
truncation=True,
max_length=1024 if self.model_type == "deepseek-finetuned" else 512,
padding=True,
)
device = next(self.model.parameters()).device
inputs = {k: v.to(device) for k, v in inputs.items()}
# Generate analysis
start_time = time.time()
with torch.no_grad():
max_new = self.quick_max_new_tokens if mode == "quick" else self.detailed_max_new_tokens
if self.model_type == "codet5":
num_beams = 1 if mode == "quick" else 2
outputs = self.model.generate(
inputs["input_ids"],
attention_mask=inputs.get("attention_mask"),
max_new_tokens=max_new,
num_beams=num_beams,
do_sample=False,
pad_token_id=self.tokenizer.eos_token_id,
use_cache=True,
)
else: # deepseek-finetuned
outputs = self.model.generate(
**inputs,
max_new_tokens=max_new,
temperature=0.7,
do_sample=True,
pad_token_id=self.tokenizer.eos_token_id,
eos_token_id=self.tokenizer.eos_token_id,
)
# Decode analysis
full_text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
analysis_text = self._extract_response(full_text, prompt)
# Calculate quality score
quality_score = self._calculate_quality_score(analysis_text)
total_time = time.time() - start_time
# Create result
result = {
"analysis": analysis_text,
"quality_score": quality_score,
"execution_time": total_time,
"model": self.model_id,
"model_type": self.model_type,
"cached": False
}
# Save to cache
self._save_to_cache(code, result)
return result
def _calculate_quality_score(self, analysis_text: str) -> int:
"""
Calculate quality score for analysis.
Args:
analysis_text: Analysis text
Returns:
int: Quality score (0-100)
"""
# Try to extract score from DeepSeek output first
if "Quality Score:" in analysis_text or "quality score" in analysis_text.lower():
import re
score_match = re.search(r'(?:Quality Score:|quality score:?)\s*(\d+)', analysis_text, re.IGNORECASE)
if score_match:
return int(score_match.group(1))
# Fallback to heuristic calculation
score = 0
analysis_lower = analysis_text.lower()
if any(word in analysis_lower for word in ['bug', 'error', 'issue', 'problem', 'flaw']):
score += 20
if any(word in analysis_lower for word in ['performance', 'slow', 'efficient', 'complexity', 'optimization']):
score += 20
if any(word in analysis_lower for word in ['security', 'vulnerability', 'safe', 'unsafe', 'risk']):
score += 20
if any(word in analysis_lower for word in ['suggest', 'improve', 'better', 'recommend', 'fix', 'solution']):
score += 20
if len(analysis_text) > 200:
score += 10
if len(analysis_text) > 500:
score += 10
return min(score, 100)
def get_model_info(self) -> Dict[str, Any]:
"""Get information about the loaded model."""
if self.model is None:
return {"status": "Model not loaded"}
param_count = sum(p.numel() for p in self.model.parameters())
trainable_params = sum(p.numel() for p in self.model.parameters() if p.requires_grad)
device = next(self.model.parameters()).device
info = {
"model_id": self.model_id,
"model_type": self.model_type,
"parameters": param_count,
"trainable_parameters": trainable_params,
"device": str(device),
"precision": self.precision,
"quick_max_new_tokens": self.quick_max_new_tokens,
"detailed_max_new_tokens": self.detailed_max_new_tokens,
"cache_size": len(self.cache)
}
if self.adapter_path:
info["adapter_path"] = self.adapter_path
info["using_adapters"] = os.path.exists(self.adapter_path) if self.adapter_path else False
return info
def main():
"""Demo of the enhanced analyzer."""
print("π Enhanced Code Analyzer Demo")
print("=" * 60)
# Test code
test_code = """
def calculate_fibonacci(n):
if n <= 0:
return 0
elif n == 1:
return 1
else:
return calculate_fibonacci(n-1) + calculate_fibonacci(n-2)
result = calculate_fibonacci(35)
print(result)
"""
print(f"Test Code:\n{test_code}")
print("=" * 60)
# Test with CodeT5+
print("\nπ΅ Testing with CodeT5+ (Base Model):")
print("-" * 60)
analyzer_codet5 = EnhancedCodeAnalyzer(model_type="codet5")
result_codet5 = analyzer_codet5.analyze_code_fast(test_code)
print(f"Analysis: {result_codet5['analysis'][:300]}...")
print(f"Quality Score: {result_codet5['quality_score']}/100")
print(f"Execution Time: {result_codet5['execution_time']:.2f}s")
# Test with Fine-tuned DeepSeek
print("\nπ’ Testing with Fine-tuned DeepSeek:")
print("-" * 60)
analyzer_deepseek = EnhancedCodeAnalyzer(
model_type="deepseek-finetuned",
adapter_path="./fine-tuned-analyst"
)
result_deepseek = analyzer_deepseek.analyze_code_fast(test_code)
print(f"Analysis: {result_deepseek['analysis'][:300]}...")
print(f"Quality Score: {result_deepseek['quality_score']}/100")
print(f"Execution Time: {result_deepseek['execution_time']:.2f}s")
# Show model comparison
print("\nπ Model Comparison:")
print("-" * 60)
print(f"CodeT5+ Info: {analyzer_codet5.get_model_info()}")
print(f"DeepSeek Info: {analyzer_deepseek.get_model_info()}")
if __name__ == "__main__":
main()
|