yolo加载模型报错

测试可以加载多少个模型，使用yolov4-tiny循环加载，在加载了60多个模型之后报错

E RKNNAPI: __pthread_recv_msg,  recv(MsgHeader) fail, -9(ERROR_PIPE) < 108!
E RKNNAPI: __pthread_recv_msg,  recv(MsgHeader) fail, -9(ERROR_PIPE) < 108!
E RKNNAPI: __pthread_recv_msg,  recv(MsgHeader) fail, -9(ERROR_PIPE) < 108!
E RKNNAPI: __pthread_recv_msg,  recv(MsgHeader) fail, -9(ERROR_PIPE) < 108!
E RKNNAPI: __pthread_recv_msg,  recv(MsgHeader) fail, -9(ERROR_PIPE) < 108!
E RKNNAPI: __pthread_recv_msg,  recv(MsgHeader) fail, -9(ERROR_PIPE) < 108!
E RKNNAPI: __pthread_recv_msg,  recv(MsgHeader) fail, -9(ERROR_PIPE) < 108!
E RKNNAPI: __pthread_recv_msg,  recv(MsgHeader) fail, -9(ERROR_PIPE) < 108!
E RKNNAPI: __pthread_recv_msg,  recv(MsgHeader) fail, -9(ERROR_PIPE) < 108!
E RKNNAPI: rknn_init,  recv(MsgLoadAck) fail, -9(ERROR_PIPE) != 368!
det_rknn_init fail! ret=-3


这是什么原因

板子 toybrick 3399proD
系统是 debian10