rpn_box: 0.0000

[luyaozhi]

hi,
when i train with voc2007, i found th rpn_box always 0. like:
[session 1][epoch 1][iter 0/5011] loss: 10.1020, lr: 1.00e-03
fg/bg=(12/244), time cost: 1.131177
rpn_cls: 0.6953, rpn_box: 0.0000, rcnn_cls: 3.0396, rcnn_box 0.2489, rcnn_cls_2nd: 3.0223, rcnn_box_2n d 0.0840, rcnn_cls_3rd: 3.0056, rcnn_box_3rd 0.0062
[session 1][epoch 1][iter 100/5011] loss: 3.8306, lr: 1.00e-03
fg/bg=(10/246), time cost: 88.860887
rpn_cls: 0.0645, rpn_box: 0.0000, rcnn_cls: 0.6193, rcnn_box 0.3400, rcnn_cls_2nd: 0.4048, rcnn_box_2n d 0.0917, rcnn_cls_3rd: 0.2731, rcnn_box_3rd 0.0146
[session 1][epoch 1][iter 200/5011] loss: 1.9012, lr: 1.00e-03
fg/bg=(18/238), time cost: 90.126512
rpn_cls: 0.1469, rpn_box: 0.0000, rcnn_cls: 0.3413, rcnn_box 0.2015, rcnn_cls_2nd: 0.3152, rcnn_box_2n d 0.0820, rcnn_cls_3rd: 0.2835, rcnn_box_3rd 0.0214
[session 1][epoch 1][iter 300/5011] loss: 1.8332, lr: 1.00e-03
fg/bg=(15/241), time cost: 90.040235
rpn_cls: 0.1073, rpn_box: 0.0000, rcnn_cls: 0.7532, rcnn_box 0.4228, rcnn_cls_2nd: 0.4802, rcnn_box_2n d 0.1167, rcnn_cls_3rd: 0.3554, rcnn_box_3rd 0.0254
[session 1][epoch 1][iter 400/5011] loss: 2.1349, lr: 1.00e-03
fg/bg=(29/227), time cost: 90.676864
rpn_cls: 0.1254, rpn_box: 0.0000, rcnn_cls: 0.9470, rcnn_box 0.4488, rcnn_cls_2nd: 0.7405, rcnn_box_2n d 0.1417, rcnn_cls_3rd: 0.6263, rcnn_box_3rd 0.0342
[session 1][epoch 1][iter 500/5011] loss: 2.4244, lr: 1.00e-03
fg/bg=(22/234), time cost: 91.616999
rpn_cls: 0.2270, rpn_box: 0.0000, rcnn_cls: 0.8991, rcnn_box 0.3421, rcnn_cls_2nd: 0.5912, rcnn_box_2n d 0.0817, rcnn_cls_3rd: 0.5299, rcnn_box_3rd 0.0175
[session 1][epoch 1][iter 600/5011] loss: 2.4373, lr: 1.00e-03
fg/bg=(22/234), time cost: 89.696538
rpn_cls: 0.1168, rpn_box: 0.0000, rcnn_cls: 0.3554, rcnn_box 0.3495, rcnn_cls_2nd: 0.2947, rcnn_box_2n d 0.1586, rcnn_cls_3rd: 0.2536, rcnn_box_3rd 0.0350
[session 1][epoch 1][iter 700/5011] loss: 2.4608, lr: 1.00e-03
fg/bg=(24/232), time cost: 90.736187
rpn_cls: 0.1741, rpn_box: 0.0000, rcnn_cls: 0.8582, rcnn_box 0.4268, rcnn_cls_2nd: 0.6125, rcnn_box_2n d 0.0850, rcnn_cls_3rd: 0.5801, rcnn_box_3rd 0.0278
[session 1][epoch 1][iter 800/5011] loss: 2.4771, lr: 1.00e-03
fg/bg=(19/237), time cost: 90.459737
rpn_cls: 0.0907, rpn_box: 0.0000, rcnn_cls: 0.2473, rcnn_box 0.2492, rcnn_cls_2nd: 0.2476, rcnn_box_2n d 0.0739, rcnn_cls_3rd: 0.2737, rcnn_box_3rd 0.0222
[session 1][epoch 1][iter 900/5011] loss: 2.4687, lr: 1.00e-03
fg/bg=(18/238), time cost: 90.652463
rpn_cls: 0.1206, rpn_box: 0.0000, rcnn_cls: 0.5917, rcnn_box 0.2917, rcnn_cls_2n d: 0.5545, rcnn_box_2nd 0.1864, rcnn_cls_3rd: 0.3710, rcnn_box_3rd 0.0428

Can you help me with it?

[fengchuidehaoe]

in anchor_target_layer_fpn.py use:
positive_weights = 1.0 / num_examples.item()
negative_weights = 1.0 / num_examples.item()