vidit09/domaingen

Va and Vb in the saved pth file

papz2000 opened this issue · 5 comments

Hello @vidit09,

In the saved .pth file of the model what are the weights of Va and Vb which are actually trained in the training phase?

Hello,
So Va and Vb are different ResNet blocks. Va- ResNet until block3 and Vb - ResNet Block4+attn layer. So the weights are not separately store but rather in the code , we have output of forward function as Va and Vb is computed through 1 and 2

But in the paper you have mentioned that resnet 1-3 blocks are Va and resnet 4+attn as vb
clip
and block 3,4 and attn are learned

so, from the pth file i am assuming
following keys of the dict are the keys for the weights of va:
backbone.enc.conv1.weight
backbone.enc.bn1.weight
backbone.enc.bn1.bias
backbone.enc.bn1.running_mean
backbone.enc.bn1.running_var
backbone.enc.bn1.num_batches_tracked
backbone.enc.conv2.weight
backbone.enc.bn2.weight
backbone.enc.bn2.bias
backbone.enc.bn2.running_mean
backbone.enc.bn2.running_var
backbone.enc.bn2.num_batches_tracked
backbone.enc.conv3.weight
backbone.enc.bn3.weight
backbone.enc.bn3.bias
backbone.enc.bn3.running_mean
backbone.enc.bn3.running_var
backbone.enc.bn3.num_batches_tracked
backbone.enc.layer1.0.conv1.weight
backbone.enc.layer1.0.bn1.weight
backbone.enc.layer1.0.bn1.bias
backbone.enc.layer1.0.bn1.running_mean
backbone.enc.layer1.0.bn1.running_var
backbone.enc.layer1.0.bn1.num_batches_tracked
backbone.enc.layer1.0.conv2.weight
backbone.enc.layer1.0.bn2.weight
backbone.enc.layer1.0.bn2.bias
backbone.enc.layer1.0.bn2.running_mean
backbone.enc.layer1.0.bn2.running_var
backbone.enc.layer1.0.bn2.num_batches_tracked
backbone.enc.layer1.0.conv3.weight
backbone.enc.layer1.0.bn3.weight
backbone.enc.layer1.0.bn3.bias
backbone.enc.layer1.0.bn3.running_mean
backbone.enc.layer1.0.bn3.running_var
backbone.enc.layer1.0.bn3.num_batches_tracked
backbone.enc.layer1.0.downsample.0.weight
backbone.enc.layer1.0.downsample.1.weight
backbone.enc.layer1.0.downsample.1.bias
backbone.enc.layer1.0.downsample.1.running_mean
backbone.enc.layer1.0.downsample.1.running_var
backbone.enc.layer1.0.downsample.1.num_batches_tracked
backbone.enc.layer1.1.conv1.weight
backbone.enc.layer1.1.bn1.weight
backbone.enc.layer1.1.bn1.bias
backbone.enc.layer1.1.bn1.running_mean
backbone.enc.layer1.1.bn1.running_var
backbone.enc.layer1.1.bn1.num_batches_tracked
backbone.enc.layer1.1.conv2.weight
backbone.enc.layer1.1.bn2.weight
backbone.enc.layer1.1.bn2.bias
backbone.enc.layer1.1.bn2.running_mean
backbone.enc.layer1.1.bn2.running_var
backbone.enc.layer1.1.bn2.num_batches_tracked
backbone.enc.layer1.1.conv3.weight
backbone.enc.layer1.1.bn3.weight
backbone.enc.layer1.1.bn3.bias
backbone.enc.layer1.1.bn3.running_mean
backbone.enc.layer1.1.bn3.running_var
backbone.enc.layer1.1.bn3.num_batches_tracked
backbone.enc.layer1.2.conv1.weight
backbone.enc.layer1.2.bn1.weight
backbone.enc.layer1.2.bn1.bias
backbone.enc.layer1.2.bn1.running_mean
backbone.enc.layer1.2.bn1.running_var
backbone.enc.layer1.2.bn1.num_batches_tracked
backbone.enc.layer1.2.conv2.weight
backbone.enc.layer1.2.bn2.weight
backbone.enc.layer1.2.bn2.bias
backbone.enc.layer1.2.bn2.running_mean
backbone.enc.layer1.2.bn2.running_var
backbone.enc.layer1.2.bn2.num_batches_tracked
backbone.enc.layer1.2.conv3.weight
backbone.enc.layer1.2.bn3.weight
backbone.enc.layer1.2.bn3.bias
backbone.enc.layer1.2.bn3.running_mean
backbone.enc.layer1.2.bn3.running_var
backbone.enc.layer1.2.bn3.num_batches_tracked
backbone.enc.layer2.0.conv1.weight
backbone.enc.layer2.0.bn1.weight
backbone.enc.layer2.0.bn1.bias
backbone.enc.layer2.0.bn1.running_mean
backbone.enc.layer2.0.bn1.running_var
backbone.enc.layer2.0.bn1.num_batches_tracked
backbone.enc.layer2.0.conv2.weight
backbone.enc.layer2.0.bn2.weight
backbone.enc.layer2.0.bn2.bias
backbone.enc.layer2.0.bn2.running_mean
backbone.enc.layer2.0.bn2.running_var
backbone.enc.layer2.0.bn2.num_batches_tracked
backbone.enc.layer2.0.conv3.weight
backbone.enc.layer2.0.bn3.weight
backbone.enc.layer2.0.bn3.bias
backbone.enc.layer2.0.bn3.running_mean
backbone.enc.layer2.0.bn3.running_var
backbone.enc.layer2.0.bn3.num_batches_tracked
backbone.enc.layer2.0.downsample.0.weight
backbone.enc.layer2.0.downsample.1.weight
backbone.enc.layer2.0.downsample.1.bias
backbone.enc.layer2.0.downsample.1.running_mean
backbone.enc.layer2.0.downsample.1.running_var
backbone.enc.layer2.0.downsample.1.num_batches_tracked
backbone.enc.layer2.1.conv1.weight
backbone.enc.layer2.1.bn1.weight
backbone.enc.layer2.1.bn1.bias
backbone.enc.layer2.1.bn1.running_mean
backbone.enc.layer2.1.bn1.running_var
backbone.enc.layer2.1.bn1.num_batches_tracked
backbone.enc.layer2.1.conv2.weight
backbone.enc.layer2.1.bn2.weight
backbone.enc.layer2.1.bn2.bias
backbone.enc.layer2.1.bn2.running_mean
backbone.enc.layer2.1.bn2.running_var
backbone.enc.layer2.1.bn2.num_batches_tracked
backbone.enc.layer2.1.conv3.weight
backbone.enc.layer2.1.bn3.weight
backbone.enc.layer2.1.bn3.bias
backbone.enc.layer2.1.bn3.running_mean
backbone.enc.layer2.1.bn3.running_var
backbone.enc.layer2.1.bn3.num_batches_tracked
backbone.enc.layer2.2.conv1.weight
backbone.enc.layer2.2.bn1.weight
backbone.enc.layer2.2.bn1.bias
backbone.enc.layer2.2.bn1.running_mean
backbone.enc.layer2.2.bn1.running_var
backbone.enc.layer2.2.bn1.num_batches_tracked
backbone.enc.layer2.2.conv2.weight
backbone.enc.layer2.2.bn2.weight
backbone.enc.layer2.2.bn2.bias
backbone.enc.layer2.2.bn2.running_mean
backbone.enc.layer2.2.bn2.running_var
backbone.enc.layer2.2.bn2.num_batches_tracked
backbone.enc.layer2.2.conv3.weight
backbone.enc.layer2.2.bn3.weight
backbone.enc.layer2.2.bn3.bias
backbone.enc.layer2.2.bn3.running_mean
backbone.enc.layer2.2.bn3.running_var
backbone.enc.layer2.2.bn3.num_batches_tracked
backbone.enc.layer2.3.conv1.weight
backbone.enc.layer2.3.bn1.weight
backbone.enc.layer2.3.bn1.bias
backbone.enc.layer2.3.bn1.running_mean
backbone.enc.layer2.3.bn1.running_var
backbone.enc.layer2.3.bn1.num_batches_tracked
backbone.enc.layer2.3.conv2.weight
backbone.enc.layer2.3.bn2.weight
backbone.enc.layer2.3.bn2.bias
backbone.enc.layer2.3.bn2.running_mean
backbone.enc.layer2.3.bn2.running_var
backbone.enc.layer2.3.bn2.num_batches_tracked
backbone.enc.layer2.3.conv3.weight
backbone.enc.layer2.3.bn3.weight
backbone.enc.layer2.3.bn3.bias
backbone.enc.layer2.3.bn3.running_mean
backbone.enc.layer2.3.bn3.running_var
backbone.enc.layer2.3.bn3.num_batches_tracked
backbone.enc.layer3.0.conv1.weight
backbone.enc.layer3.0.bn1.weight
backbone.enc.layer3.0.bn1.bias
backbone.enc.layer3.0.bn1.running_mean
backbone.enc.layer3.0.bn1.running_var
backbone.enc.layer3.0.bn1.num_batches_tracked
backbone.enc.layer3.0.conv2.weight
backbone.enc.layer3.0.bn2.weight
backbone.enc.layer3.0.bn2.bias
backbone.enc.layer3.0.bn2.running_mean
backbone.enc.layer3.0.bn2.running_var
backbone.enc.layer3.0.bn2.num_batches_tracked
backbone.enc.layer3.0.conv3.weight
backbone.enc.layer3.0.bn3.weight
backbone.enc.layer3.0.bn3.bias
backbone.enc.layer3.0.bn3.running_mean
backbone.enc.layer3.0.bn3.running_var
backbone.enc.layer3.0.bn3.num_batches_tracked
backbone.enc.layer3.0.downsample.0.weight
backbone.enc.layer3.0.downsample.1.weight
backbone.enc.layer3.0.downsample.1.bias
backbone.enc.layer3.0.downsample.1.running_mean
backbone.enc.layer3.0.downsample.1.running_var
backbone.enc.layer3.0.downsample.1.num_batches_tracked
backbone.enc.layer3.1.conv1.weight
backbone.enc.layer3.1.bn1.weight
backbone.enc.layer3.1.bn1.bias
backbone.enc.layer3.1.bn1.running_mean
backbone.enc.layer3.1.bn1.running_var
backbone.enc.layer3.1.bn1.num_batches_tracked
backbone.enc.layer3.1.conv2.weight
backbone.enc.layer3.1.bn2.weight
backbone.enc.layer3.1.bn2.bias
backbone.enc.layer3.1.bn2.running_mean
backbone.enc.layer3.1.bn2.running_var
backbone.enc.layer3.1.bn2.num_batches_tracked
backbone.enc.layer3.1.conv3.weight
backbone.enc.layer3.1.bn3.weight
backbone.enc.layer3.1.bn3.bias
backbone.enc.layer3.1.bn3.running_mean
backbone.enc.layer3.1.bn3.running_var
backbone.enc.layer3.1.bn3.num_batches_tracked
backbone.enc.layer3.2.conv1.weight
backbone.enc.layer3.2.bn1.weight
backbone.enc.layer3.2.bn1.bias
backbone.enc.layer3.2.bn1.running_mean
backbone.enc.layer3.2.bn1.running_var
backbone.enc.layer3.2.bn1.num_batches_tracked
backbone.enc.layer3.2.conv2.weight
backbone.enc.layer3.2.bn2.weight
backbone.enc.layer3.2.bn2.bias
backbone.enc.layer3.2.bn2.running_mean
backbone.enc.layer3.2.bn2.running_var
backbone.enc.layer3.2.bn2.num_batches_tracked
backbone.enc.layer3.2.conv3.weight
backbone.enc.layer3.2.bn3.weight
backbone.enc.layer3.2.bn3.bias
backbone.enc.layer3.2.bn3.running_mean
backbone.enc.layer3.2.bn3.running_var
backbone.enc.layer3.2.bn3.num_batches_tracked
backbone.enc.layer3.3.conv1.weight
backbone.enc.layer3.3.bn1.weight
backbone.enc.layer3.3.bn1.bias
backbone.enc.layer3.3.bn1.running_mean
backbone.enc.layer3.3.bn1.running_var
backbone.enc.layer3.3.bn1.num_batches_tracked
backbone.enc.layer3.3.conv2.weight
backbone.enc.layer3.3.bn2.weight
backbone.enc.layer3.3.bn2.bias
backbone.enc.layer3.3.bn2.running_mean
backbone.enc.layer3.3.bn2.running_var
backbone.enc.layer3.3.bn2.num_batches_tracked
backbone.enc.layer3.3.conv3.weight
backbone.enc.layer3.3.bn3.weight
backbone.enc.layer3.3.bn3.bias
backbone.enc.layer3.3.bn3.running_mean
backbone.enc.layer3.3.bn3.running_var
backbone.enc.layer3.3.bn3.num_batches_tracked
backbone.enc.layer3.4.conv1.weight
backbone.enc.layer3.4.bn1.weight
backbone.enc.layer3.4.bn1.bias
backbone.enc.layer3.4.bn1.running_mean
backbone.enc.layer3.4.bn1.running_var
backbone.enc.layer3.4.bn1.num_batches_tracked
backbone.enc.layer3.4.conv2.weight
backbone.enc.layer3.4.bn2.weight
backbone.enc.layer3.4.bn2.bias
backbone.enc.layer3.4.bn2.running_mean
backbone.enc.layer3.4.bn2.running_var
backbone.enc.layer3.4.bn2.num_batches_tracked
backbone.enc.layer3.4.conv3.weight
backbone.enc.layer3.4.bn3.weight
backbone.enc.layer3.4.bn3.bias
backbone.enc.layer3.4.bn3.running_mean
backbone.enc.layer3.4.bn3.running_var
backbone.enc.layer3.4.bn3.num_batches_tracked
backbone.enc.layer3.5.conv1.weight
backbone.enc.layer3.5.bn1.weight
backbone.enc.layer3.5.bn1.bias
backbone.enc.layer3.5.bn1.running_mean
backbone.enc.layer3.5.bn1.running_var
backbone.enc.layer3.5.bn1.num_batches_tracked
backbone.enc.layer3.5.conv2.weight
backbone.enc.layer3.5.bn2.weight
backbone.enc.layer3.5.bn2.bias
backbone.enc.layer3.5.bn2.running_mean
backbone.enc.layer3.5.bn2.running_var
backbone.enc.layer3.5.bn2.num_batches_tracked
backbone.enc.layer3.5.conv3.weight
backbone.enc.layer3.5.bn3.weight
backbone.enc.layer3.5.bn3.bias
backbone.enc.layer3.5.bn3.running_mean
backbone.enc.layer3.5.bn3.running_var
backbone.enc.layer3.5.bn3.num_batches_tracked
backbone.enc.layer3.6.conv1.weight
backbone.enc.layer3.6.bn1.weight
backbone.enc.layer3.6.bn1.bias
backbone.enc.layer3.6.bn1.running_mean
backbone.enc.layer3.6.bn1.running_var
backbone.enc.layer3.6.bn1.num_batches_tracked
backbone.enc.layer3.6.conv2.weight
backbone.enc.layer3.6.bn2.weight
backbone.enc.layer3.6.bn2.bias
backbone.enc.layer3.6.bn2.running_mean
backbone.enc.layer3.6.bn2.running_var
backbone.enc.layer3.6.bn2.num_batches_tracked
backbone.enc.layer3.6.conv3.weight
backbone.enc.layer3.6.bn3.weight
backbone.enc.layer3.6.bn3.bias
backbone.enc.layer3.6.bn3.running_mean
backbone.enc.layer3.6.bn3.running_var
backbone.enc.layer3.6.bn3.num_batches_tracked
backbone.enc.layer3.7.conv1.weight
backbone.enc.layer3.7.bn1.weight
backbone.enc.layer3.7.bn1.bias
backbone.enc.layer3.7.bn1.running_mean
backbone.enc.layer3.7.bn1.running_var
backbone.enc.layer3.7.bn1.num_batches_tracked
backbone.enc.layer3.7.conv2.weight
backbone.enc.layer3.7.bn2.weight
backbone.enc.layer3.7.bn2.bias
backbone.enc.layer3.7.bn2.running_mean
backbone.enc.layer3.7.bn2.running_var
backbone.enc.layer3.7.bn2.num_batches_tracked
backbone.enc.layer3.7.conv3.weight
backbone.enc.layer3.7.bn3.weight
backbone.enc.layer3.7.bn3.bias
backbone.enc.layer3.7.bn3.running_mean
backbone.enc.layer3.7.bn3.running_var
backbone.enc.layer3.7.bn3.num_batches_tracked
backbone.enc.layer3.8.conv1.weight
backbone.enc.layer3.8.bn1.weight
backbone.enc.layer3.8.bn1.bias
backbone.enc.layer3.8.bn1.running_mean
backbone.enc.layer3.8.bn1.running_var
backbone.enc.layer3.8.bn1.num_batches_tracked
backbone.enc.layer3.8.conv2.weight
backbone.enc.layer3.8.bn2.weight
backbone.enc.layer3.8.bn2.bias
backbone.enc.layer3.8.bn2.running_mean
backbone.enc.layer3.8.bn2.running_var
backbone.enc.layer3.8.bn2.num_batches_tracked
backbone.enc.layer3.8.conv3.weight
backbone.enc.layer3.8.bn3.weight
backbone.enc.layer3.8.bn3.bias
backbone.enc.layer3.8.bn3.running_mean
backbone.enc.layer3.8.bn3.running_var
backbone.enc.layer3.8.bn3.num_batches_tracked
backbone.enc.layer3.9.conv1.weight
backbone.enc.layer3.9.bn1.weight
backbone.enc.layer3.9.bn1.bias
backbone.enc.layer3.9.bn1.running_mean
backbone.enc.layer3.9.bn1.running_var
backbone.enc.layer3.9.bn1.num_batches_tracked
backbone.enc.layer3.9.conv2.weight
backbone.enc.layer3.9.bn2.weight
backbone.enc.layer3.9.bn2.bias
backbone.enc.layer3.9.bn2.running_mean
backbone.enc.layer3.9.bn2.running_var
backbone.enc.layer3.9.bn2.num_batches_tracked
backbone.enc.layer3.9.conv3.weight
backbone.enc.layer3.9.bn3.weight
backbone.enc.layer3.9.bn3.bias
backbone.enc.layer3.9.bn3.running_mean
backbone.enc.layer3.9.bn3.running_var
backbone.enc.layer3.9.bn3.num_batches_tracked
backbone.enc.layer3.10.conv1.weight
backbone.enc.layer3.10.bn1.weight
backbone.enc.layer3.10.bn1.bias
backbone.enc.layer3.10.bn1.running_mean
backbone.enc.layer3.10.bn1.running_var
backbone.enc.layer3.10.bn1.num_batches_tracked
backbone.enc.layer3.10.conv2.weight
backbone.enc.layer3.10.bn2.weight
backbone.enc.layer3.10.bn2.bias
backbone.enc.layer3.10.bn2.running_mean
backbone.enc.layer3.10.bn2.running_var
backbone.enc.layer3.10.bn2.num_batches_tracked
backbone.enc.layer3.10.conv3.weight
backbone.enc.layer3.10.bn3.weight
backbone.enc.layer3.10.bn3.bias
backbone.enc.layer3.10.bn3.running_mean
backbone.enc.layer3.10.bn3.running_var
backbone.enc.layer3.10.bn3.num_batches_tracked
backbone.enc.layer3.11.conv1.weight
backbone.enc.layer3.11.bn1.weight
backbone.enc.layer3.11.bn1.bias
backbone.enc.layer3.11.bn1.running_mean
backbone.enc.layer3.11.bn1.running_var
backbone.enc.layer3.11.bn1.num_batches_tracked
backbone.enc.layer3.11.conv2.weight
backbone.enc.layer3.11.bn2.weight
backbone.enc.layer3.11.bn2.bias
backbone.enc.layer3.11.bn2.running_mean
backbone.enc.layer3.11.bn2.running_var
backbone.enc.layer3.11.bn2.num_batches_tracked
backbone.enc.layer3.11.conv3.weight
backbone.enc.layer3.11.bn3.weight
backbone.enc.layer3.11.bn3.bias
backbone.enc.layer3.11.bn3.running_mean
backbone.enc.layer3.11.bn3.running_var
backbone.enc.layer3.11.bn3.num_batches_tracked
backbone.enc.layer3.12.conv1.weight
backbone.enc.layer3.12.bn1.weight
backbone.enc.layer3.12.bn1.bias
backbone.enc.layer3.12.bn1.running_mean
backbone.enc.layer3.12.bn1.running_var
backbone.enc.layer3.12.bn1.num_batches_tracked
backbone.enc.layer3.12.conv2.weight
backbone.enc.layer3.12.bn2.weight
backbone.enc.layer3.12.bn2.bias
backbone.enc.layer3.12.bn2.running_mean
backbone.enc.layer3.12.bn2.running_var
backbone.enc.layer3.12.bn2.num_batches_tracked
backbone.enc.layer3.12.conv3.weight
backbone.enc.layer3.12.bn3.weight
backbone.enc.layer3.12.bn3.bias
backbone.enc.layer3.12.bn3.running_mean
backbone.enc.layer3.12.bn3.running_var
backbone.enc.layer3.12.bn3.num_batches_tracked
backbone.enc.layer3.13.conv1.weight
backbone.enc.layer3.13.bn1.weight
backbone.enc.layer3.13.bn1.bias
backbone.enc.layer3.13.bn1.running_mean
backbone.enc.layer3.13.bn1.running_var
backbone.enc.layer3.13.bn1.num_batches_tracked
backbone.enc.layer3.13.conv2.weight
backbone.enc.layer3.13.bn2.weight
backbone.enc.layer3.13.bn2.bias
backbone.enc.layer3.13.bn2.running_mean
backbone.enc.layer3.13.bn2.running_var
backbone.enc.layer3.13.bn2.num_batches_tracked
backbone.enc.layer3.13.conv3.weight
backbone.enc.layer3.13.bn3.weight
backbone.enc.layer3.13.bn3.bias
backbone.enc.layer3.13.bn3.running_mean
backbone.enc.layer3.13.bn3.running_var
backbone.enc.layer3.13.bn3.num_batches_tracked
backbone.enc.layer3.14.conv1.weight
backbone.enc.layer3.14.bn1.weight
backbone.enc.layer3.14.bn1.bias
backbone.enc.layer3.14.bn1.running_mean
backbone.enc.layer3.14.bn1.running_var
backbone.enc.layer3.14.bn1.num_batches_tracked
backbone.enc.layer3.14.conv2.weight
backbone.enc.layer3.14.bn2.weight
backbone.enc.layer3.14.bn2.bias
backbone.enc.layer3.14.bn2.running_mean
backbone.enc.layer3.14.bn2.running_var
backbone.enc.layer3.14.bn2.num_batches_tracked
backbone.enc.layer3.14.conv3.weight
backbone.enc.layer3.14.bn3.weight
backbone.enc.layer3.14.bn3.bias
backbone.enc.layer3.14.bn3.running_mean
backbone.enc.layer3.14.bn3.running_var
backbone.enc.layer3.14.bn3.num_batches_tracked
backbone.enc.layer3.15.conv1.weight
backbone.enc.layer3.15.bn1.weight
backbone.enc.layer3.15.bn1.bias
backbone.enc.layer3.15.bn1.running_mean
backbone.enc.layer3.15.bn1.running_var
backbone.enc.layer3.15.bn1.num_batches_tracked
backbone.enc.layer3.15.conv2.weight
backbone.enc.layer3.15.bn2.weight
backbone.enc.layer3.15.bn2.bias
backbone.enc.layer3.15.bn2.running_mean
backbone.enc.layer3.15.bn2.running_var
backbone.enc.layer3.15.bn2.num_batches_tracked
backbone.enc.layer3.15.conv3.weight
backbone.enc.layer3.15.bn3.weight
backbone.enc.layer3.15.bn3.bias
backbone.enc.layer3.15.bn3.running_mean
backbone.enc.layer3.15.bn3.running_var
backbone.enc.layer3.15.bn3.num_batches_tracked
backbone.enc.layer3.16.conv1.weight
backbone.enc.layer3.16.bn1.weight
backbone.enc.layer3.16.bn1.bias
backbone.enc.layer3.16.bn1.running_mean
backbone.enc.layer3.16.bn1.running_var
backbone.enc.layer3.16.bn1.num_batches_tracked
backbone.enc.layer3.16.conv2.weight
backbone.enc.layer3.16.bn2.weight
backbone.enc.layer3.16.bn2.bias
backbone.enc.layer3.16.bn2.running_mean
backbone.enc.layer3.16.bn2.running_var
backbone.enc.layer3.16.bn2.num_batches_tracked
backbone.enc.layer3.16.conv3.weight
backbone.enc.layer3.16.bn3.weight
backbone.enc.layer3.16.bn3.bias
backbone.enc.layer3.16.bn3.running_mean
backbone.enc.layer3.16.bn3.running_var
backbone.enc.layer3.16.bn3.num_batches_tracked
backbone.enc.layer3.17.conv1.weight
backbone.enc.layer3.17.bn1.weight
backbone.enc.layer3.17.bn1.bias
backbone.enc.layer3.17.bn1.running_mean
backbone.enc.layer3.17.bn1.running_var
backbone.enc.layer3.17.bn1.num_batches_tracked
backbone.enc.layer3.17.conv2.weight
backbone.enc.layer3.17.bn2.weight
backbone.enc.layer3.17.bn2.bias
backbone.enc.layer3.17.bn2.running_mean
backbone.enc.layer3.17.bn2.running_var
backbone.enc.layer3.17.bn2.num_batches_tracked
backbone.enc.layer3.17.conv3.weight
backbone.enc.layer3.17.bn3.weight
backbone.enc.layer3.17.bn3.bias
backbone.enc.layer3.17.bn3.running_mean
backbone.enc.layer3.17.bn3.running_var
backbone.enc.layer3.17.bn3.num_batches_tracked
backbone.enc.layer3.18.conv1.weight
backbone.enc.layer3.18.bn1.weight
backbone.enc.layer3.18.bn1.bias
backbone.enc.layer3.18.bn1.running_mean
backbone.enc.layer3.18.bn1.running_var
backbone.enc.layer3.18.bn1.num_batches_tracked
backbone.enc.layer3.18.conv2.weight
backbone.enc.layer3.18.bn2.weight
backbone.enc.layer3.18.bn2.bias
backbone.enc.layer3.18.bn2.running_mean
backbone.enc.layer3.18.bn2.running_var
backbone.enc.layer3.18.bn2.num_batches_tracked
backbone.enc.layer3.18.conv3.weight
backbone.enc.layer3.18.bn3.weight
backbone.enc.layer3.18.bn3.bias
backbone.enc.layer3.18.bn3.running_mean
backbone.enc.layer3.18.bn3.running_var
backbone.enc.layer3.18.bn3.num_batches_tracked
backbone.enc.layer3.19.conv1.weight
backbone.enc.layer3.19.bn1.weight
backbone.enc.layer3.19.bn1.bias
backbone.enc.layer3.19.bn1.running_mean
backbone.enc.layer3.19.bn1.running_var
backbone.enc.layer3.19.bn1.num_batches_tracked
backbone.enc.layer3.19.conv2.weight
backbone.enc.layer3.19.bn2.weight
backbone.enc.layer3.19.bn2.bias
backbone.enc.layer3.19.bn2.running_mean
backbone.enc.layer3.19.bn2.running_var
backbone.enc.layer3.19.bn2.num_batches_tracked
backbone.enc.layer3.19.conv3.weight
backbone.enc.layer3.19.bn3.weight
backbone.enc.layer3.19.bn3.bias
backbone.enc.layer3.19.bn3.running_mean
backbone.enc.layer3.19.bn3.running_var
backbone.enc.layer3.19.bn3.num_batches_tracked
backbone.enc.layer3.20.conv1.weight
backbone.enc.layer3.20.bn1.weight
backbone.enc.layer3.20.bn1.bias
backbone.enc.layer3.20.bn1.running_mean
backbone.enc.layer3.20.bn1.running_var
backbone.enc.layer3.20.bn1.num_batches_tracked
backbone.enc.layer3.20.conv2.weight
backbone.enc.layer3.20.bn2.weight
backbone.enc.layer3.20.bn2.bias
backbone.enc.layer3.20.bn2.running_mean
backbone.enc.layer3.20.bn2.running_var
backbone.enc.layer3.20.bn2.num_batches_tracked
backbone.enc.layer3.20.conv3.weight
backbone.enc.layer3.20.bn3.weight
backbone.enc.layer3.20.bn3.bias
backbone.enc.layer3.20.bn3.running_mean
backbone.enc.layer3.20.bn3.running_var
backbone.enc.layer3.20.bn3.num_batches_tracked
backbone.enc.layer3.21.conv1.weight
backbone.enc.layer3.21.bn1.weight
backbone.enc.layer3.21.bn1.bias
backbone.enc.layer3.21.bn1.running_mean
backbone.enc.layer3.21.bn1.running_var
backbone.enc.layer3.21.bn1.num_batches_tracked
backbone.enc.layer3.21.conv2.weight
backbone.enc.layer3.21.bn2.weight
backbone.enc.layer3.21.bn2.bias
backbone.enc.layer3.21.bn2.running_mean
backbone.enc.layer3.21.bn2.running_var
backbone.enc.layer3.21.bn2.num_batches_tracked
backbone.enc.layer3.21.conv3.weight
backbone.enc.layer3.21.bn3.weight
backbone.enc.layer3.21.bn3.bias
backbone.enc.layer3.21.bn3.running_mean
backbone.enc.layer3.21.bn3.running_var
backbone.enc.layer3.21.bn3.num_batches_tracked
backbone.enc.layer3.22.conv1.weight
backbone.enc.layer3.22.bn1.weight
backbone.enc.layer3.22.bn1.bias
backbone.enc.layer3.22.bn1.running_mean
backbone.enc.layer3.22.bn1.running_var
backbone.enc.layer3.22.bn1.num_batches_tracked
backbone.enc.layer3.22.conv2.weight
backbone.enc.layer3.22.bn2.weight
backbone.enc.layer3.22.bn2.bias
backbone.enc.layer3.22.bn2.running_mean
backbone.enc.layer3.22.bn2.running_var
backbone.enc.layer3.22.bn2.num_batches_tracked
backbone.enc.layer3.22.conv3.weight
backbone.enc.layer3.22.bn3.weight
backbone.enc.layer3.22.bn3.bias
backbone.enc.layer3.22.bn3.running_mean
backbone.enc.layer3.22.bn3.running_var
backbone.enc.layer3.22.bn3.num_batches_tracked


And Vb are:
backbone.enc.layer4.0.conv1.weight
backbone.enc.layer4.0.bn1.weight
backbone.enc.layer4.0.bn1.bias
backbone.enc.layer4.0.bn1.running_mean
backbone.enc.layer4.0.bn1.running_var
backbone.enc.layer4.0.bn1.num_batches_tracked
backbone.enc.layer4.0.conv2.weight
backbone.enc.layer4.0.bn2.weight
backbone.enc.layer4.0.bn2.bias
backbone.enc.layer4.0.bn2.running_mean
backbone.enc.layer4.0.bn2.running_var
backbone.enc.layer4.0.bn2.num_batches_tracked
backbone.enc.layer4.0.conv3.weight
backbone.enc.layer4.0.bn3.weight
backbone.enc.layer4.0.bn3.bias
backbone.enc.layer4.0.bn3.running_mean
backbone.enc.layer4.0.bn3.running_var
backbone.enc.layer4.0.bn3.num_batches_tracked
backbone.enc.layer4.0.downsample.0.weight
backbone.enc.layer4.0.downsample.1.weight
backbone.enc.layer4.0.downsample.1.bias
backbone.enc.layer4.0.downsample.1.running_mean
backbone.enc.layer4.0.downsample.1.running_var
backbone.enc.layer4.0.downsample.1.num_batches_tracked
backbone.enc.layer4.1.conv1.weight
backbone.enc.layer4.1.bn1.weight
backbone.enc.layer4.1.bn1.bias
backbone.enc.layer4.1.bn1.running_mean
backbone.enc.layer4.1.bn1.running_var
backbone.enc.layer4.1.bn1.num_batches_tracked
backbone.enc.layer4.1.conv2.weight
backbone.enc.layer4.1.bn2.weight
backbone.enc.layer4.1.bn2.bias
backbone.enc.layer4.1.bn2.running_mean
backbone.enc.layer4.1.bn2.running_var
backbone.enc.layer4.1.bn2.num_batches_tracked
backbone.enc.layer4.1.conv3.weight
backbone.enc.layer4.1.bn3.weight
backbone.enc.layer4.1.bn3.bias
backbone.enc.layer4.1.bn3.running_mean
backbone.enc.layer4.1.bn3.running_var
backbone.enc.layer4.1.bn3.num_batches_tracked
backbone.enc.layer4.2.conv1.weight
backbone.enc.layer4.2.bn1.weight
backbone.enc.layer4.2.bn1.bias
backbone.enc.layer4.2.bn1.running_mean
backbone.enc.layer4.2.bn1.running_var
backbone.enc.layer4.2.bn1.num_batches_tracked
backbone.enc.layer4.2.conv2.weight
backbone.enc.layer4.2.bn2.weight
backbone.enc.layer4.2.bn2.bias
backbone.enc.layer4.2.bn2.running_mean
backbone.enc.layer4.2.bn2.running_var
backbone.enc.layer4.2.bn2.num_batches_tracked
backbone.enc.layer4.2.conv3.weight
backbone.enc.layer4.2.bn3.weight
backbone.enc.layer4.2.bn3.bias
backbone.enc.layer4.2.bn3.running_mean
backbone.enc.layer4.2.bn3.running_var
backbone.enc.layer4.2.bn3.num_batches_tracked
backbone.enc.attnpool.positional_embedding
backbone.enc.attnpool.k_proj.weight
backbone.enc.attnpool.k_proj.bias
backbone.enc.attnpool.q_proj.weight
backbone.enc.attnpool.q_proj.bias
backbone.enc.attnpool.v_proj.weight
backbone.enc.attnpool.v_proj.bias
backbone.enc.attnpool.c_proj.weight
backbone.enc.attnpool.c_proj.bias

is this correct or am i missing something?

Thanks for your time and consideration.

Yes, you are correct. Va until 3 and Vb until 4 + attn layers

Okay, Thank you soo much for the response