paddle multi threads error

[HuuY]

bug描述 Describe the Bug

paddle cpu版本运行多线程时发生报错。

`import os
from time import sleep
import concurrent.futures
import traceback
from paddleocr import PaddleOCR

ocr = PaddleOCR(
ocr_version='PP-OCRv4',
text_detection_model_name="PP-OCRv4_mobile_det",
text_detection_model_dir="./PP-OCRv4_mobile_det",
text_recognition_model_name="PP-OCRv4_mobile_rec",
text_recognition_model_dir="./PP-OCRv4_mobile_rec",
use_doc_orientation_classify=False,
use_doc_unwarping=False,
use_textline_orientation=False,
device="cpu"
)

def predect_test(filepath):
result = ocr.predict(filepath)
print(f"{filepath} 预测结束")
for res in result:
res.print()

try:
thread_pool = concurrent.futures.ThreadPoolExecutor(max_workers=5)
for root, sub_dirs, files in os.walk("/data/hugo/paddle_test/images"):
for filename in files:
fixname = filename.split(".")[1]
if fixname=="jpg" or fixname=="jpeg" or fixname=="JPEG" or fixname=="JPG" or fixname=="wav" or fixname=="WAV" or fixname=="mp4" or fixname=="png":
bsname = os.path.join(root, filename)
print("提交文件预测任务:", bsname)
thread_pool.submit(predect_test,bsname)
thread_pool.shutdown(wait=True)
except Exception as e:
traceback.print_exc()`

测试情况：
在3.1.0 paddle cpu上运行报错。
在3.1.0 paddle gpu上运行报错。
在3.0.0 paddle gpu上运行正常。

paddlex 3.2.0
paddleocr 3.2.0

其他补充信息 Additional Supplementary Information

No response