Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -2,36 +2,54 @@ from fastapi import FastAPI
|
|
2 |
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
3 |
import torch
|
4 |
import os
|
|
|
5 |
|
6 |
-
#
|
7 |
-
|
|
|
8 |
|
9 |
-
#
|
10 |
os.environ["HF_HOME"] = "/app/.cache/huggingface"
|
11 |
|
12 |
-
#
|
13 |
try:
|
14 |
-
|
15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
except Exception as e:
|
17 |
-
|
|
|
|
|
|
|
18 |
|
19 |
-
# 4. 接口定义
|
20 |
@app.post("/detect")
|
21 |
async def detect(code: str):
|
22 |
try:
|
23 |
-
#
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
inputs = tokenizer(code, return_tensors="pt", truncation=True)
|
28 |
with torch.no_grad():
|
29 |
outputs = model(**inputs)
|
30 |
-
|
|
|
|
|
31 |
return {
|
32 |
-
"label": model.config.id2label[
|
33 |
-
"score": outputs.logits.softmax(dim=-1).
|
34 |
}
|
35 |
|
36 |
except Exception as e:
|
37 |
-
return {"error": str(e)}
|
|
|
|
|
|
|
|
|
|
2 |
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
3 |
import torch
|
4 |
import os
|
5 |
+
import logging
|
6 |
|
7 |
+
# 初始化日志
|
8 |
+
logging.basicConfig(level=logging.INFO)
|
9 |
+
logger = logging.getLogger("CodeSecurityAPI")
|
10 |
|
11 |
+
# 强制设置缓存路径(解决权限问题)
|
12 |
os.environ["HF_HOME"] = "/app/.cache/huggingface"
|
13 |
|
14 |
+
# 加载模型
|
15 |
try:
|
16 |
+
logger.info("Loading model...")
|
17 |
+
model = AutoModelForSequenceClassification.from_pretrained(
|
18 |
+
"mrm8488/codebert-base-finetuned-detect-insecure-code",
|
19 |
+
cache_dir=os.getenv("HF_HOME")
|
20 |
+
)
|
21 |
+
tokenizer = AutoTokenizer.from_pretrained(
|
22 |
+
"mrm8488/codebert-base-finetuned-detect-insecure-code",
|
23 |
+
cache_dir=os.getenv("HF_HOME")
|
24 |
+
)
|
25 |
+
logger.info("Model loaded successfully")
|
26 |
except Exception as e:
|
27 |
+
logger.error(f"Model load failed: {str(e)}")
|
28 |
+
raise RuntimeError("模型加载失败,请检查网络连接或模型路径")
|
29 |
+
|
30 |
+
app = FastAPI()
|
31 |
|
|
|
32 |
@app.post("/detect")
|
33 |
async def detect(code: str):
|
34 |
try:
|
35 |
+
# 输入处理(限制长度)
|
36 |
+
code = code[:2000] # 截断超长输入
|
37 |
+
|
38 |
+
# 模型推理
|
39 |
+
inputs = tokenizer(code, return_tensors="pt", truncation=True, max_length=512)
|
40 |
with torch.no_grad():
|
41 |
outputs = model(**inputs)
|
42 |
+
|
43 |
+
# 解析结果
|
44 |
+
label_id = outputs.logits.argmax().item()
|
45 |
return {
|
46 |
+
"label": model.config.id2label[label_id],
|
47 |
+
"score": outputs.logits.softmax(dim=-1)[0][label_id].item()
|
48 |
}
|
49 |
|
50 |
except Exception as e:
|
51 |
+
return {"error": str(e)}
|
52 |
+
|
53 |
+
@app.get("/health")
|
54 |
+
async def health():
|
55 |
+
return {"status": "ok"}
|