预训练权重的加载问题

[mr-xiao-zuo]

加载预训练参数显示missing keys警告，这样是不是没有加载成功，如何正确加载预训练权重？ [2021-04-12 20:59:56,397-rk0-model_load.py# 48] load pretrained model from /home/dl/code/SiamGAT-main/tools/../pretrained_models/inception_v3.pth [2021-04-12 20:59:56,468-rk0-model_load.py# 42] remove prefix 'module.' [2021-04-12 20:59:56,470-rk0-model_load.py# 26] [Warning] missing keys: ['channel_reduce.0.weight', 'channel_reduce.1.weight', 'channel_reduce.0.bias', 'channel_reduce.1.running_var', 'channel_reduce.1.running_mean', 'channel_reduce.1.bias'] [2021-04-12 20:59:56,470-rk0-model_load.py# 27] missing keys:6 [2021-04-12 20:59:56,470-rk0-model_load.py# 30] [Warning] unused_pretrained_keys: {'Mixed_7c.branch3x3_2a.bn.running_var', 'Mixed_7c.branch3x3_2a.bn.bias', 'Mixed_7b.branch3x3_2a.bn.bias', 'Mixed_7c.branch1x1.bn.running_mean', 'Mixed_7b.branch3x3dbl_1.bn.bias', 'Mixed_7c.branch3x3dbl_3a.bn.running_var', 'Mixed_7c.branch1x1.bn.bias', 'Mixed_7a.branch7x7x3_4.bn.weight', 'Mixed_7b.branch3x3_2a.bn.running_var', 'AuxLogits.conv1.bn.running_mean', 'Mixed_7b.branch3x3dbl_3b.bn.running_mean', 'Mixed_7c.branch3x3_2b.conv.weight', 'Mixed_7c.branch3x3dbl_3a.bn.bias', 'Mixed_7c.branch3x3_2a.bn.weight', 'Mixed_7b.branch3x3dbl_1.bn.running_var', 'Mixed_7b.branch_pool.bn.running_mean', 'AuxLogits.conv0.bn.running_mean', 'Mixed_7b.branch_pool.conv.weight', 'Mixed_7b.branch1x1.bn.weight', 'Mixed_7c.branch3x3dbl_3b.bn.weight', 'Mixed_7a.branch7x7x3_3.conv.weight', 'Mixed_7a.branch7x7x3_1.bn.running_mean', 'Mixed_7b.branch3x3_2b.bn.bias', 'Mixed_7b.branch3x3dbl_2.bn.running_mean', 'AuxLogits.fc.bias', 'Mixed_7c.branch3x3_2a.conv.weight', 'Mixed_7b.branch3x3dbl_1.conv.weight', 'Mixed_7a.branch7x7x3_2.bn.running_var', 'Mixed_7a.branch7x7x3_4.bn.running_mean', 'Mixed_7c.branch_pool.bn.running_var', 'Mixed_7c.branch1x1.bn.running_var', 'Mixed_7b.branch3x3_2b.bn.running_mean', 'Mixed_7a.branch7x7x3_3.bn.weight', 'Mixed_7b.branch3x3dbl_1.bn.running_mean', 'Mixed_7c.branch3x3_2b.bn.running_var', 'Mixed_7c.branch3x3dbl_3b.bn.bias', 'Mixed_7b.branch3x3dbl_2.conv.weight', 'Mixed_7b.branch3x3dbl_2.bn.bias', 'AuxLogits.conv1.bn.weight', 'Mixed_7a.branch7x7x3_1.conv.weight', 'Mixed_7b.branch3x3_1.bn.weight', 'Mixed_7c.branch3x3dbl_3b.conv.weight', 'Mixed_7c.branch3x3dbl_3a.conv.weight', 'AuxLogits.fc.weight', 'Mixed_7c.branch3x3dbl_3a.bn.running_mean', 'Mixed_7b.branch3x3dbl_3b.conv.weight', 'Mixed_7c.branch3x3dbl_2.conv.weight', 'Mixed_7c.branch3x3dbl_1.bn.running_mean', 'Mixed_7c.branch3x3_1.conv.weight', 'Mixed_7c.branch3x3dbl_2.bn.running_mean', 'Mixed_7b.branch3x3_2b.bn.weight', 'AuxLogits.conv1.bn.bias', 'Mixed_7b.branch1x1.bn.running_mean', 'Mixed_7b.branch3x3_1.bn.running_mean', 'Mixed_7b.branch3x3dbl_3b.bn.bias', 'Mixed_7c.branch3x3dbl_1.bn.weight', 'Mixed_7a.branch3x3_2.bn.bias', 'Mixed_7a.branch3x3_1.bn.running_mean', 'Mixed_7a.branch7x7x3_2.bn.running_mean', 'Mixed_7b.branch_pool.bn.running_var', 'Mixed_7a.branch3x3_2.bn.running_var', 'Mixed_7b.branch3x3_1.conv.weight', 'Mixed_7c.branch3x3dbl_3b.bn.running_mean', 'Mixed_7c.branch3x3dbl_3a.bn.weight', 'AuxLogits.conv1.bn.running_var', 'Mixed_7b.branch3x3_1.bn.running_var', 'Mixed_7c.branch3x3dbl_3b.bn.running_var', 'Mixed_7b.branch_pool.bn.weight', 'Mixed_7c.branch3x3_1.bn.running_mean', 'Mixed_7a.branch3x3_1.bn.running_var', 'Mixed_7c.branch3x3_2b.bn.running_mean', 'fc.bias', 'Mixed_7b.branch3x3dbl_2.bn.running_var', 'Mixed_7c.branch_pool.conv.weight', 'Mixed_7b.branch3x3_1.bn.bias', 'Mixed_7b.branch3x3dbl_2.bn.weight', 'Mixed_7b.branch3x3dbl_3b.bn.weight', 'Mixed_7b.branch1x1.conv.weight', 'Mixed_7a.branch7x7x3_3.bn.running_mean', 'Mixed_7a.branch3x3_2.bn.running_mean', 'AuxLogits.conv0.bn.weight', 'Mixed_7b.branch3x3_2b.conv.weight', 'Mixed_7c.branch_pool.bn.bias', 'Mixed_7b.branch3x3_2a.bn.running_mean', 'Mixed_7b.branch3x3dbl_3a.bn.weight', 'AuxLogits.conv0.bn.running_var', 'Mixed_7c.branch3x3dbl_1.conv.weight', 'Mixed_7a.branch3x3_1.bn.weight', 'Mixed_7c.branch1x1.bn.weight', 'Mixed_7c.branch3x3dbl_2.bn.running_var', 'Mixed_7c.branch_pool.bn.running_mean', 'Mixed_7c.branch3x3dbl_1.bn.running_var', 'Mixed_7c.branch3x3_1.bn.weight', 'Mixed_7a.branch7x7x3_3.bn.bias', 'AuxLogits.conv0.conv.weight', 'Mixed_7a.branch7x7x3_4.conv.weight', 'Mixed_7b.branch3x3dbl_3a.bn.bias', 'AuxLogits.conv1.conv.weight', 'Mixed_7a.branch3x3_2.bn.weight', 'AuxLogits.conv0.bn.bias', 'fc.weight', 'Mixed_7a.branch7x7x3_4.bn.running_var', 'Mixed_7a.branch3x3_1.bn.bias', 'Mixed_7b.branch3x3_2a.bn.weight', 'Mixed_7a.branch7x7x3_1.bn.bias', 'Mixed_7b.branch_pool.bn.bias', 'Mixed_7b.branch1x1.bn.running_var', 'Mixed_7a.branch7x7x3_3.bn.running_var', 'Mixed_7a.branch7x7x3_1.bn.weight', 'Mixed_7b.branch3x3dbl_3a.bn.running_mean', 'Mixed_7b.branch3x3_2a.conv.weight', 'Mixed_7c.branch3x3_2a.bn.running_mean', 'Mixed_7c.branch3x3_1.bn.bias', 'Mixed_7c.branch3x3dbl_2.bn.bias', 'Mixed_7a.branch7x7x3_4.bn.bias', 'Mixed_7c.branch3x3_2b.bn.weight', 'Mixed_7c.branch3x3dbl_2.bn.weight', 'Mixed_7a.branch7x7x3_2.bn.bias', 'Mixed_7b.branch1x1.bn.bias', 'Mixed_7c.branch3x3dbl_1.bn.bias', 'Mixed_7c.branch1x1.conv.weight', 'Mixed_7c.branch3x3_1.bn.running_var', 'Mixed_7c.branch_pool.bn.weight', 'Mixed_7b.branch3x3dbl_1.bn.weight', 'Mixed_7b.branch3x3dbl_3b.bn.running_var', 'Mixed_7b.branch3x3_2b.bn.running_var', 'Mixed_7b.branch3x3dbl_3a.conv.weight', 'Mixed_7a.branch7x7x3_2.conv.weight', 'Mixed_7a.branch7x7x3_2.bn.weight', 'Mixed_7a.branch3x3_2.conv.weight', 'Mixed_7c.branch3x3_2b.bn.bias', 'Mixed_7b.branch3x3dbl_3a.bn.running_var', 'Mixed_7a.branch7x7x3_1.bn.running_var', 'Mixed_7a.branch3x3_1.conv.weight'} [2021-04-12 20:59:56,470-rk0-model_load.py# 32] unused checkpoint keys:134 [2021-04-12 20:59:56,470-rk0-model_load.py# 33] used keys:350

[twotwo2]

已成功加载了预训练模型。为了适应目标跟踪的任务，我们修改了googlenet-v3的网络，修改内容包括移除了googlenet-v3的last stage，增加了用于降维的channel_reduce层。因此在加载预训练模型的时候，提示“缺乏channel_reduce的参数，没有用到googlenet-v3的部分参数”是正常的情况。