missing keys in state_dict

[Anikily]

I am running upsnet_resnet50_coco_4gpu.yaml and i have download the resnet-50-caffe.pth and resnet-101-caffe.pth in model file,but i still get this worning:

UserWarning: missing keys in state_dict: "{'fcn_head.fcn_subnet.conv.1.0.conv.weight', 'fpn.fpn_p2.bias', 'fpn.fpn_p4.bias', 'rpn.cls_score.weight', 'mask_branch.mask_deconv1.0.weight', 'fcn_head.fcn_subnet.conv.0.0.conv_offset.weight', 'resnet_backbone.conv1.bn1.num_batches_tracked', 'resnet_backbone.res2.layers.2.bn3.num_batches_tracked', 'resnet_backbone.res4.layers.2.bn1.num_batches_tracked', 'mask_branch.mask_deconv1.0.bias', 'resnet_backbone.res5.layers.2.bn1.num_batches_tracked', 'rpn.conv_proposal.0.weight', 'mask_branch.mask_conv3.0.weight', 'rcnn.bbox_pred.bias', 'resnet_backbone.res3.layers.3.bn2.num_batches_tracked', 'resnet_backbone.res5.layers.2.bn2.num_batches_tracked', 'fcn_head.fcn_subnet.conv.2.0.conv_offset.weight', 'resnet_backbone.res3.layers.3.bn1.num_batches_tracked', 'resnet_backbone.res4.layers.1.bn3.num_batches_tracked', 'mask_branch.mask_conv1.0.bias', 'fcn_head.fcn_subnet.conv.2.0.conv_offset.bias', 'resnet_backbone.res3.layers.1.bn3.num_batches_tracked', 'resnet_backbone.res5.layers.1.bn2.num_batches_tracked', 'mask_branch.mask_conv4.0.weight', 'fcn_head.fcn_subnet.conv.1.0.conv_offset.bias', 'resnet_backbone.res4.layers.0.bn3.num_batches_tracked', 'fpn.fpn_p4.weight', 'fpn.fpn_gap.weight', 'fpn.fpn_p5.weight', 'resnet_backbone.res3.layers.1.bn1.num_batches_tracked', 'resnet_backbone.res5.layers.0.bn1.num_batches_tracked', 'fpn.fpn_p3.bias', 'mask_branch.mask_conv2.0.weight', 'resnet_backbone.res4.layers.1.bn1.num_batches_tracked', 'resnet_backbone.res4.layers.5.bn1.num_batches_tracked', 'resnet_backbone.res2.layers.0.bn2.num_batches_tracked', 'fcn_head.score.bias', 'fpn.fpn_p3.weight', 'resnet_backbone.res4.layers.3.bn1.num_batches_tracked', 'resnet_backbone.res2.layers.0.bn1.num_batches_tracked', 'resnet_backbone.res5.layers.0.downsample.1.num_batches_tracked', 'resnet_backbone.res4.layers.2.bn2.num_batches_tracked', 'resnet_backbone.res3.layers.0.bn3.num_batches_tracked', 'rcnn.bbox_pred.weight', 'resnet_backbone.res4.layers.0.bn2.num_batches_tracked', 'resnet_backbone.res4.layers.0.downsample.1.num_batches_tracked', 'resnet_backbone.res2.layers.2.bn1.num_batches_tracked', 'resnet_backbone.res3.layers.0.bn1.num_batches_tracked', 'fcn_head.fcn_subnet.conv.1.0.conv_offset.weight', 'fcn_head.fcn_subnet.conv.0.0.conv.weight', 'resnet_backbone.res2.layers.2.bn2.num_batches_tracked', 'fcn_head.fcn_subnet.conv.2.0.conv.bias', 'fcn_head.fcn_subnet.conv.0.0.conv_offset.bias', 'resnet_backbone.res4.layers.4.bn1.num_batches_tracked', 'fcn_head.fcn_subnet.conv.0.0.conv.bias', 'resnet_backbone.res2.layers.1.bn1.num_batches_tracked', 'rcnn.fc6.0.bias', 'fpn.fpn_p2_1x1.weight', 'resnet_backbone.res4.layers.4.bn3.num_batches_tracked', 'resnet_backbone.res5.layers.2.bn3.num_batches_tracked', 'fpn.fpn_p4_1x1.bias', 'rcnn.cls_score.weight', 'mask_branch.mask_conv1.0.weight', 'fpn.fpn_p5.bias', 'resnet_backbone.res4.layers.0.bn1.num_batches_tracked', 'rpn.bbox_pred.weight', 'rcnn.cls_score.bias', 'mask_branch.mask_score.bias', 'resnet_backbone.res3.layers.2.bn1.num_batches_tracked', 'rpn.conv_proposal.0.bias', 'rcnn.fc7.0.weight', 'resnet_backbone.res3.layers.0.downsample.1.num_batches_tracked', 'resnet_backbone.res5.layers.1.bn1.num_batches_tracked', 'mask_branch.mask_conv2.0.bias', 'fpn.fpn_p2_1x1.bias', 'fpn.fpn_p3_1x1.weight', 'resnet_backbone.res3.layers.1.bn2.num_batches_tracked', 'resnet_backbone.res3.layers.2.bn2.num_batches_tracked', 'resnet_backbone.res4.layers.5.bn3.num_batches_tracked', 'fpn.fpn_p5_1x1.weight', 'resnet_backbone.res5.layers.0.bn3.num_batches_tracked', 'fpn.fpn_p3_1x1.bias', 'rpn.cls_score.bias', 'resnet_backbone.res3.layers.3.bn3.num_batches_tracked', 'resnet_backbone.res4.layers.3.bn2.num_batches_tracked', 'fpn.fpn_p5_1x1.bias', 'rcnn.fc7.0.bias', 'resnet_backbone.res4.layers.2.bn3.num_batches_tracked', 'resnet_backbone.res2.layers.0.downsample.1.num_batches_tracked', 'fcn_head.fcn_subnet.conv.2.0.conv.weight', 'resnet_backbone.res2.layers.0.bn3.num_batches_tracked', 'resnet_backbone.res3.layers.0.bn2.num_batches_tracked', 'resnet_backbone.res4.layers.4.bn2.num_batches_tracked', 'fcn_head.score.weight', 'mask_branch.mask_score.weight', 'resnet_backbone.res2.layers.1.bn2.num_batches_tracked', 'rcnn.fc6.0.weight', 'fcn_head.fcn_subnet.conv.1.0.conv.bias', 'resnet_backbone.res5.layers.0.bn2.num_batches_tracked', 'fpn.fpn_p4_1x1.weight', 'rpn.bbox_pred.bias', 'mask_branch.mask_conv3.0.bias', 'resnet_backbone.res4.layers.5.bn2.num_batches_tracked', 'fpn.fpn_gap.bias', 'fpn.fpn_p2.weight', 'resnet_backbone.res2.layers.1.bn3.num_batches_tracked', 'mask_branch.mask_conv4.0.bias', 'resnet_backbone.res5.layers.1.bn3.num_batches_tracked', 'resnet_backbone.res4.layers.1.bn2.num_batches_tracked', 'resnet_backbone.res3.layers.2.bn3.num_batches_tracked', 'resnet_backbone.res4.layers.3.bn3.num_batches_tracked'}"

[PkuRainBow]

@Anikily So do you achieve any conclusion about this problem?

[Anikily]

@PkuRainBow doesn't matter.

[PkuRainBow]

Thanks!