File size: 7,851 Bytes
1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e 1b77b3b 76cad5e |
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 |
from flask import Flask, request, jsonify
import torch
from transformers import RobertaTokenizer, RobertaForSequenceClassification
import os
from functools import lru_cache
app = Flask(__name__)
model = None
tokenizer = None
device = None
def setup_device():
if torch.cuda.is_available():
return torch.device('cuda')
elif hasattr(torch.backends, 'mps') and torch.backends.mps.is_available():
return torch.device('mps')
else:
return torch.device('cpu')
def load_tokenizer():
try:
tokenizer = RobertaTokenizer.from_pretrained('./tokenizer_vulnerability')
tokenizer.model_max_length = 512
return tokenizer
except Exception as e:
print(f"Error loading tokenizer: {e}")
return RobertaTokenizer.from_pretrained('microsoft/codebert-base')
def load_model():
global device
device = setup_device()
print(f"Using device: {device}")
try:
checkpoint = torch.load("codebert_vulnerability_scorer.pth", map_location=device)
if 'config' in checkpoint:
from transformers import RobertaConfig
config = RobertaConfig.from_dict(checkpoint['config'])
model = RobertaForSequenceClassification(config)
else:
model = RobertaForSequenceClassification.from_pretrained(
'microsoft/codebert-base',
num_labels=1
)
if 'model_state_dict' in checkpoint:
model.load_state_dict(checkpoint['model_state_dict'])
else:
model.load_state_dict(checkpoint)
model.to(device)
model.eval()
if device.type == 'cuda':
model.half()
return model
except Exception as e:
print(f"Error loading model: {e}")
raise e
@lru_cache(maxsize=1000)
def cached_tokenize(code_hash, max_length):
code = code_hash
return tokenizer(
code,
truncation=True,
padding='max_length',
max_length=max_length,
return_tensors='pt'
)
try:
print("Loading tokenizer...")
tokenizer = load_tokenizer()
print("Tokenizer loaded successfully!")
print("Loading model...")
model = load_model()
print("Model loaded successfully!")
except Exception as e:
print(f"Error during initialization: {str(e)}")
tokenizer = None
model = None
@app.route("/", methods=['GET'])
def home():
return jsonify({
"message": "CodeBERT Vulnerability Scorer API",
"status": "Model loaded" if model is not None else "Model not loaded",
"device": str(device) if device else "unknown",
"endpoints": {
"/predict": "POST with JSON body containing 'code' field",
"/predict_batch": "POST with JSON body containing 'codes' array",
"/predict_get": "GET with 'code' URL parameter"
}
})
@app.route("/predict", methods=['POST'])
def predict_post():
try:
if model is None or tokenizer is None:
return jsonify({"error": "Model not loaded properly"}), 500
data = request.get_json()
if not data or 'code' not in data:
return jsonify({"error": "Missing 'code' field in JSON body"}), 400
code = data['code']
if not code or not isinstance(code, str):
return jsonify({"error": "'code' field must be a non-empty string"}), 400
score = predict_vulnerability(code)
return jsonify({
"score": score,
"vulnerability_level": get_vulnerability_level(score),
"code_preview": code[:200] + "..." if len(code) > 200 else code
})
except Exception as e:
return jsonify({"error": f"Prediction error: {str(e)}"}), 500
@app.route("/predict_batch", methods=['POST'])
def predict_batch():
try:
if model is None or tokenizer is None:
return jsonify({"error": "Model not loaded properly"}), 500
data = request.get_json()
if not data or 'codes' not in data:
return jsonify({"error": "Missing 'codes' field in JSON body"}), 400
codes = data['codes']
if not isinstance(codes, list) or len(codes) == 0:
return jsonify({"error": "'codes' must be a non-empty array"}), 400
batch_size = min(len(codes), 16)
results = []
for i in range(0, len(codes), batch_size):
batch = codes[i:i+batch_size]
scores = predict_vulnerability_batch(batch)
for j, score in enumerate(scores):
results.append({
"index": i + j,
"score": score,
"vulnerability_level": get_vulnerability_level(score),
"code_preview": batch[j][:100] + "..." if len(batch[j]) > 100 else batch[j]
})
return jsonify({"results": results})
except Exception as e:
return jsonify({"error": f"Batch prediction error: {str(e)}"}), 500
@app.route("/predict_get", methods=['GET'])
def predict_get():
try:
if model is None or tokenizer is None:
return jsonify({"error": "Model not loaded properly"}), 500
code = request.args.get("code")
if not code:
return jsonify({"error": "Missing 'code' URL parameter"}), 400
score = predict_vulnerability(code)
return jsonify({
"score": score,
"vulnerability_level": get_vulnerability_level(score),
"code_preview": code[:200] + "..." if len(code) > 200 else code
})
except Exception as e:
return jsonify({"error": f"Prediction error: {str(e)}"}), 500
def predict_vulnerability(code):
dynamic_length = min(max(len(code.split()) * 2, 128), 512)
inputs = tokenizer(
code,
truncation=True,
padding='max_length',
max_length=dynamic_length,
return_tensors='pt'
)
inputs = {k: v.to(device) for k, v in inputs.items()}
with torch.no_grad():
with torch.cuda.amp.autocast() if device.type == 'cuda' else torch.no_grad():
outputs = model(**inputs)
if hasattr(outputs, 'logits'):
score = torch.sigmoid(outputs.logits).cpu().item()
else:
score = torch.sigmoid(outputs[0]).cpu().item()
return round(score, 4)
def predict_vulnerability_batch(codes):
max_len = max([len(code.split()) * 2 for code in codes])
dynamic_length = min(max(max_len, 128), 512)
inputs = tokenizer(
codes,
truncation=True,
padding='max_length',
max_length=dynamic_length,
return_tensors='pt'
)
inputs = {k: v.to(device) for k, v in inputs.items()}
with torch.no_grad():
with torch.cuda.amp.autocast() if device.type == 'cuda' else torch.no_grad():
outputs = model(**inputs)
if hasattr(outputs, 'logits'):
scores = torch.sigmoid(outputs.logits).cpu().numpy()
else:
scores = torch.sigmoid(outputs[0]).cpu().numpy()
return [round(float(score), 4) for score in scores.flatten()]
def get_vulnerability_level(score):
if score < 0.3:
return "Low"
elif score < 0.7:
return "Medium"
else:
return "High"
@app.route("/health", methods=['GET'])
def health_check():
return jsonify({
"status": "healthy",
"model_loaded": model is not None,
"tokenizer_loaded": tokenizer is not None,
"device": str(device) if device else "unknown"
})
if __name__ == "__main__":
app.run(host="0.0.0.0", port=7860, debug=False, threaded=True) |