RPNAcc_FG keep staying 0.0000

I have trained a mobilenet_0_25 model using wilderface dataset and train.py on detection/retinaface/train.py, during traing process, the Loss and RPNAcc_BG keep changing but the RPNAcc_FG always staying at 0.0000 after many epochs.

This is my config:
INFO:root:Called with argument: Namespace(begin_epoch=0, dataset='retinaface', dataset_path='data/retinaface', end_epoch=100, frequent=20, image_set='train', kvstore='device', lr=0.01, lr_step='30,60,80', network='mnet', no_flip=False, no_shuffle=False, prefix='./model', pretrained='model/mobilenet_0_25', pretrained_epoch=0, root_path='data', wd=0.0005, work_load_list=None)
gpu num: 1
INFO:root:{'BBOX_MASK_THRESH': 0,
'CASCADE': 0,
'CASCADE_BBOX_STRIDES': [64, 32, 16, 8, 4],
'CASCADE_CLS_STRIDES': [64, 32, 16, 8, 4],
'CASCADE_MODE': 1,
'COLOR_JITTERING': 0.125,
'COLOR_MODE': 1,
'CONTEXT_FILTER_RATIO': 1,
'DENSE_ANCHOR': False,
'FACE_LANDMARK': True,
'FIXED_PARAMS': ['^stage1', '^.*upsampling'],
'HEAD_BOX': False,
'HEAD_FILTER_NUM': 64,
'HEAD_MODULE': 'SSH',
'IMAGE_STRIDE': 0,
'LANDMARK_LR_MULT': 2.5,
'LAYER_FIX': True,
'LR_MODE': 0,
'MIXUP': 0.0,
'MORE_SMALL_BOX': True,
'NET_MODE': 2,
'NUM_ANCHORS': 2,
'NUM_CLASSES': 2,
'NUM_CPU': 4,
'ORIGIN_SCALE': False,
'PIXEL_MEANS': array([0., 0., 0.]),
'PIXEL_SCALE': 1.0,
'PIXEL_STDS': array([1., 1., 1.]),
'PRE_SCALES': [(1200, 1600)],
'RANDOM_FEAT_STRIDE': False,
'RPN_ANCHOR_CFG': {'16': {'ALLOWED_BORDER': 9999,
'BASE_SIZE': 16,
'NUM_ANCHORS': 2,
'RATIOS': [1.0],
'SCALES': [8, 4]},
'32': {'ALLOWED_BORDER': 9999,
'BASE_SIZE': 16,
'NUM_ANCHORS': 2,
'RATIOS': [1.0],
'SCALES': [32, 16]},
'8': {'ALLOWED_BORDER': 9999,
'BASE_SIZE': 16,
'NUM_ANCHORS': 2,
'RATIOS': [1.0],
'SCALES': [2, 1]}},
'RPN_FEAT_STRIDE': [32, 16, 8],
'SCALES': [(640, 640)],
'SHARE_WEIGHT_BBOX': False,
'SHARE_WEIGHT_LANDMARK': False,
'TEST': {'BATCH_IMAGES': 1,
'CXX_PROPOSAL': True,
'HAS_RPN': False,
'IOU_THRESH': 0.5,
'NMS': 0.3,
'RPN_NMS_THRESH': 0.3,
'RPN_POST_NMS_TOP_N': 3000,
'RPN_PRE_NMS_TOP_N': 1000,
'SCORE_THRESH': 0.05},
'TRAIN': {'ASPECT_GROUPING': False,
'BATCH_IMAGES': 32,
'BBOX_STDS': [1.0, 1.0, 1.0, 1.0],
'CASCADE_OVERLAP': [0.4, 0.5],
'END2END': True,
'IMAGE_ALIGN': 0,
'LANDMARK_STD': 1.0,
'MIN_BOX_SIZE': 0,
'OHEM_MODE': 1,
'RPN_BATCH_SIZE': 256,
'RPN_CLOBBER_POSITIVES': False,
'RPN_ENABLE_OHEM': 2,
'RPN_FG_FRACTION': 0.25,
'RPN_FORCE_POSITIVE': False,
'RPN_NEGATIVE_OVERLAP': 0.3,
'RPN_POSITIVE_OVERLAP': 0.5},
'USE_3D': False,
'USE_BLUR': False,
'USE_CROP': True,
'USE_DCN': 0,
'USE_FPN': True,
'USE_MAXOUT': 0,
'USE_OCCLUSION': False,
'dataset': 'retinaface',
'max_feat_channel': 8888,
'network': 'mnet'}
origin image size 12880
retinaface_train gt roidb loaded from data/cache/retinaface_train_train_gt_roidb.pkl
roidb size 12876
INFO:root:retinaface_train append flipped images to roidb
flipped roidb size 25752
INFO:root:loading model/mobilenet_0_25,0
[03:42:31] ../src/nnvm/legacy_json_util.cc:208: Loading symbol saved by previous version v1.3.0. Attempting to upgrade...
[03:42:31] ../src/nnvm/legacy_json_util.cc:216: Symbol successfully upgraded!

This is the traning log of epoch 10
INFO:root:Saved checkpoint to "./model-0000.params"
INFO:root:Epoch[10] Train-RPNAcc_s32=0.974243
INFO:root:Epoch[10] Train-RPNAcc_s32_BG=1.000000
INFO:root:Epoch[10] Train-RPNAcc_s32_FG=0.000000
INFO:root:Epoch[10] Train-RPNL1Loss_s32=0.394602
INFO:root:Epoch[10] Train-RPNLandMarkL1Loss_s32=0.902790
INFO:root:Epoch[10] Train-RPNAcc_s16=0.970903
INFO:root:Epoch[10] Train-RPNAcc_s16_BG=1.000000
INFO:root:Epoch[10] Train-RPNAcc_s16_FG=0.000000
INFO:root:Epoch[10] Train-RPNL1Loss_s16=0.283116
INFO:root:Epoch[10] Train-RPNLandMarkL1Loss_s16=0.572992
INFO:root:Epoch[10] Train-RPNAcc_s8=0.983558
INFO:root:Epoch[10] Train-RPNAcc_s8_BG=1.000000
INFO:root:Epoch[10] Train-RPNAcc_s8_FG=0.000000
INFO:root:Epoch[10] Train-RPNL1Loss_s8=0.282509
INFO:root:Epoch[10] Train-RPNLandMarkL1Loss_s8=0.621004
INFO:root:Epoch[10] Time cost=989.293

After about 10 epochs, the RPNAcc_s32_FG, RPNAcc_s16_FG and RPN_Acc_s8_FG keep staying at 0.
Please help me to fix it.