Saving / Restoring Weights

mertmerci commented 4 years ago

Hello,

Are the weights automatically saved in eachh epoch, or how frequently they are saved during training ? How can I restore them if I need to stop the training at some point and continue later?

Thank you for your support.

Tramac commented 4 years ago

1.The weights are saved in each epoch if no_val=True. https://github.com/Tramac/Fast-SCNN-pytorch/blob/a197f5e734632cb80798c90e825dcfdafd8422d9/train.py#L150 2.resume the model. https://github.com/Tramac/Fast-SCNN-pytorch/blob/a197f5e734632cb80798c90e825dcfdafd8422d9/train.py#L94

mertmerci commented 4 years ago

Hello again,

I tried to restore the weights from the previous training with the starting command:

python train.py --model fast_scnn --dataset citys --resume ./weights/fast_scnn_citys.pth --epochs 600

but I get an error message, I directly paste the complete error below because it's very long.

Resuming training, loading ./weights/fast_scnn_citys.pth... Traceback (most recent call last): File "train.py", line 204, in <module> trainer = Trainer(args) File "train.py", line 110, in __init__ self.model.load_state_dict(torch.load(args.resume, map_location=lambda storage, loc: storage)) File "/home/gg392953/anaconda3/envs/myenv/lib/python3.7/site-packages/torch/nn/modules/module.py", line 777, in load_state_dict self.__class__.__name__, "\n\t".join(error_msgs))) RuntimeError: Error(s) in loading state_dict for FastSCNN: Missing key(s) in state_dict: "learning_to_downsample.conv.conv.0.weight", "learning_to_downsample.conv.conv.1.weight", "learning_to_downsample.conv.conv.1.bias", "learning_to_downsample.conv.conv.1.running_mean", "learning_to_downsample.conv.conv.1.running_var", "learning_to_downsample.dsconv1.conv.0.weight", "learning_to_downsample.dsconv1.conv.1.weight", "learning_to_downsample.dsconv1.conv.1.bias", "learning_to_downsample.dsconv1.conv.1.running_mean", "learning_to_downsample.dsconv1.conv.1.running_var", "learning_to_downsample.dsconv1.conv.3.weight", "learning_to_downsample.dsconv1.conv.4.weight", "learning_to_downsample.dsconv1.conv.4.bias", "learning_to_downsample.dsconv1.conv.4.running_mean", "learning_to_downsample.dsconv1.conv.4.running_var", "learning_to_downsample.dsconv2.conv.0.weight", "learning_to_downsample.dsconv2.conv.1.weight", "learning_to_downsample.dsconv2.conv.1.bias", "learning_to_downsample.dsconv2.conv.1.running_mean", "learning_to_downsample.dsconv2.conv.1.running_var", "learning_to_downsample.dsconv2.conv.3.weight", "learning_to_downsample.dsconv2.conv.4.weight", "learning_to_downsample.dsconv2.conv.4.bias", "learning_to_downsample.dsconv2.conv.4.running_mean", "learning_to_downsample.dsconv2.conv.4.running_var", "global_feature_extractor.bottleneck1.0.block.0.conv.0.weight", "global_feature_extractor.bottleneck1.0.block.0.conv.1.weight", "global_feature_extractor.bottleneck1.0.block.0.conv.1.bias", "global_feature_extractor.bottleneck1.0.block.0.conv.1.running_mean", "global_feature_extractor.bottleneck1.0.block.0.conv.1.running_var", "global_feature_extractor.bottleneck1.0.block.1.conv.0.weight", "global_feature_extractor.bottleneck1.0.block.1.conv.1.weight", "global_feature_extractor.bottleneck1.0.block.1.conv.1.bias", "global_feature_extractor.bottleneck1.0.block.1.conv.1.running_mean", "global_feature_extractor.bottleneck1.0.block.1.conv.1.running_var", "global_feature_extractor.bottleneck1.0.block.2.weight", "global_feature_extractor.bottleneck1.0.block.3.weight", "global_feature_extractor.bottleneck1.0.block.3.bias", "global_feature_extractor.bottleneck1.0.block.3.running_mean", "global_feature_extractor.bottleneck1.0.block.3.running_var", "global_feature_extractor.bottleneck1.1.block.0.conv.0.weight", "global_feature_extractor.bottleneck1.1.block.0.conv.1.weight", "global_feature_extractor.bottleneck1.1.block.0.conv.1.bias", "global_feature_extractor.bottleneck1.1.block.0.conv.1.running_mean", "global_feature_extractor.bottleneck1.1.block.0.conv.1.running_var", "global_feature_extractor.bottleneck1.1.block.1.conv.0.weight", "global_feature_extractor.bottleneck1.1.block.1.conv.1.weight", "global_feature_extractor.bottleneck1.1.block.1.conv.1.bias", "global_feature_extractor.bottleneck1.1.block.1.conv.1.running_mean", "global_feature_extractor.bottleneck1.1.block.1.conv.1.running_var", "global_feature_extractor.bottleneck1.1.block.2.weight", "global_feature_extractor.bottleneck1.1.block.3.weight", "global_feature_extractor.bottleneck1.1.block.3.bias", "global_feature_extractor.bottleneck1.1.block.3.running_mean", "global_feature_extractor.bottleneck1.1.block.3.running_var", "global_feature_extractor.bottleneck1.2.block.0.conv.0.weight", "global_feature_extractor.bottleneck1.2.block.0.conv.1.weight", "global_feature_extractor.bottleneck1.2.block.0.conv.1.bias", "global_feature_extractor.bottleneck1.2.block.0.conv.1.running_mean", "global_feature_extractor.bottleneck1.2.block.0.conv.1.running_var", "global_feature_extractor.bottleneck1.2.block.1.conv.0.weight", "global_feature_extractor.bottleneck1.2.block.1.conv.1.weight", "global_feature_extractor.bottleneck1.2.block.1.conv.1.bias", "global_feature_extractor.bottleneck1.2.block.1.conv.1.running_mean", "global_feature_extractor.bottleneck1.2.block.1.conv.1.running_var", "global_feature_extractor.bottleneck1.2.block.2.weight", "global_feature_extractor.bottleneck1.2.block.3.weight", "global_feature_extractor.bottleneck1.2.block.3.bias", "global_feature_extractor.bottleneck1.2.block.3.running_mean", "global_feature_extractor.bottleneck1.2.block.3.running_var", "global_feature_extractor.bottleneck2.0.block.0.conv.0.weight", "global_feature_extractor.bottleneck2.0.block.0.conv.1.weight", "global_feature_extractor.bottleneck2.0.block.0.conv.1.bias", "global_feature_extractor.bottleneck2.0.block.0.conv.1.running_mean", "global_feature_extractor.bottleneck2.0.block.0.conv.1.running_var", "global_feature_extractor.bottleneck2.0.block.1.conv.0.weight", "global_feature_extractor.bottleneck2.0.block.1.conv.1.weight", "global_feature_extractor.bottleneck2.0.block.1.conv.1.bias", "global_feature_extractor.bottleneck2.0.block.1.conv.1.running_mean", "global_feature_extractor.bottleneck2.0.block.1.conv.1.running_var", "global_feature_extractor.bottleneck2.0.block.2.weight", "global_feature_extractor.bottleneck2.0.block.3.weight", "global_feature_extractor.bottleneck2.0.block.3.bias", "global_feature_extractor.bottleneck2.0.block.3.running_mean", "global_feature_extractor.bottleneck2.0.block.3.running_var", "global_feature_extractor.bottleneck2.1.block.0.conv.0.weight", "global_feature_extractor.bottleneck2.1.block.0.conv.1.weight", "global_feature_extractor.bottleneck2.1.block.0.conv.1.bias", "global_feature_extractor.bottleneck2.1.block.0.conv.1.running_mean", "global_feature_extractor.bottleneck2.1.block.0.conv.1.running_var", "global_feature_extractor.bottleneck2.1.block.1.conv.0.weight", "global_feature_extractor.bottleneck2.1.block.1.conv.1.weight", "global_feature_extractor.bottleneck2.1.block.1.conv.1.bias", "global_feature_extractor.bottleneck2.1.block.1.conv.1.running_mean", "global_feature_extractor.bottleneck2.1.block.1.conv.1.running_var", "global_feature_extractor.bottleneck2.1.block.2.weight", "global_feature_extractor.bottleneck2.1.block.3.weight", "global_feature_extractor.bottleneck2.1.block.3.bias", "global_feature_extractor.bottleneck2.1.block.3.running_mean", "global_feature_extractor.bottleneck2.1.block.3.running_var", "global_feature_extractor.bottleneck2.2.block.0.conv.0.weight", "global_feature_extractor.bottleneck2.2.block.0.conv.1.weight", "global_feature_extractor.bottleneck2.2.block.0.conv.1.bias", "global_feature_extractor.bottleneck2.2.block.0.conv.1.running_mean", "global_feature_extractor.bottleneck2.2.block.0.conv.1.running_var", "global_feature_extractor.bottleneck2.2.block.1.conv.0.weight", "global_feature_extractor.bottleneck2.2.block.1.conv.1.weight", "global_feature_extractor.bottleneck2.2.block.1.conv.1.bias", "global_feature_extractor.bottleneck2.2.block.1.conv.1.running_mean", "global_feature_extractor.bottleneck2.2.block.1.conv.1.running_var", "global_feature_extractor.bottleneck2.2.block.2.weight", "global_feature_extractor.bottleneck2.2.block.3.weight", "global_feature_extractor.bottleneck2.2.block.3.bias", "global_feature_extractor.bottleneck2.2.block.3.running_mean", "global_feature_extractor.bottleneck2.2.block.3.running_var", "global_feature_extractor.bottleneck3.0.block.0.conv.0.weight", "global_feature_extractor.bottleneck3.0.block.0.conv.1.weight", "global_feature_extractor.bottleneck3.0.block.0.conv.1.bias", "global_feature_extractor.bottleneck3.0.block.0.conv.1.running_mean", "global_feature_extractor.bottleneck3.0.block.0.conv.1.running_var", "global_feature_extractor.bottleneck3.0.block.1.conv.0.weight", "global_feature_extractor.bottleneck3.0.block.1.conv.1.weight", "global_feature_extractor.bottleneck3.0.block.1.conv.1.bias", "global_feature_extractor.bottleneck3.0.block.1.conv.1.running_mean", "global_feature_extractor.bottleneck3.0.block.1.conv.1.running_var", "global_feature_extractor.bottleneck3.0.block.2.weight", "global_feature_extractor.bottleneck3.0.block.3.weight", "global_feature_extractor.bottleneck3.0.block.3.bias", "global_feature_extractor.bottleneck3.0.block.3.running_mean", "global_feature_extractor.bottleneck3.0.block.3.running_var", "global_feature_extractor.bottleneck3.1.block.0.conv.0.weight", "global_feature_extractor.bottleneck3.1.block.0.conv.1.weight", "global_feature_extractor.bottleneck3.1.block.0.conv.1.bias", "global_feature_extractor.bottleneck3.1.block.0.conv.1.running_mean", "global_feature_extractor.bottleneck3.1.block.0.conv.1.running_var", "global_feature_extractor.bottleneck3.1.block.1.conv.0.weight", "global_feature_extractor.bottleneck3.1.block.1.conv.1.weight", "global_feature_extractor.bottleneck3.1.block.1.conv.1.bias", "global_feature_extractor.bottleneck3.1.block.1.conv.1.running_mean", "global_feature_extractor.bottleneck3.1.block.1.conv.1.running_var", "global_feature_extractor.bottleneck3.1.block.2.weight", "global_feature_extractor.bottleneck3.1.block.3.weight", "global_feature_extractor.bottleneck3.1.block.3.bias", "global_feature_extractor.bottleneck3.1.block.3.running_mean", "global_feature_extractor.bottleneck3.1.block.3.running_var", "global_feature_extractor.bottleneck3.2.block.0.conv.0.weight", "global_feature_extractor.bottleneck3.2.block.0.conv.1.weight", "global_feature_extractor.bottleneck3.2.block.0.conv.1.bias", "global_feature_extractor.bottleneck3.2.block.0.conv.1.running_mean", "global_feature_extractor.bottleneck3.2.block.0.conv.1.running_var", "global_feature_extractor.bottleneck3.2.block.1.conv.0.weight", "global_feature_extractor.bottleneck3.2.block.1.conv.1.weight", "global_feature_extractor.bottleneck3.2.block.1.conv.1.bias", "global_feature_extractor.bottleneck3.2.block.1.conv.1.running_mean", "global_feature_extractor.bottleneck3.2.block.1.conv.1.running_var", "global_feature_extractor.bottleneck3.2.block.2.weight", "global_feature_extractor.bottleneck3.2.block.3.weight", "global_feature_extractor.bottleneck3.2.block.3.bias", "global_feature_extractor.bottleneck3.2.block.3.running_mean", "global_feature_extractor.bottleneck3.2.block.3.running_var", "global_feature_extractor.ppm.conv1.conv.0.weight", "global_feature_extractor.ppm.conv1.conv.1.weight", "global_feature_extractor.ppm.conv1.conv.1.bias", "global_feature_extractor.ppm.conv1.conv.1.running_mean", "global_feature_extractor.ppm.conv1.conv.1.running_var", "global_feature_extractor.ppm.conv2.conv.0.weight", "global_feature_extractor.ppm.conv2.conv.1.weight", "global_feature_extractor.ppm.conv2.conv.1.bias", "global_feature_extractor.ppm.conv2.conv.1.running_mean", "global_feature_extractor.ppm.conv2.conv.1.running_var", "global_feature_extractor.ppm.conv3.conv.0.weight", "global_feature_extractor.ppm.conv3.conv.1.weight", "global_feature_extractor.ppm.conv3.conv.1.bias", "global_feature_extractor.ppm.conv3.conv.1.running_mean", "global_feature_extractor.ppm.conv3.conv.1.running_var", "global_feature_extractor.ppm.conv4.conv.0.weight", "global_feature_extractor.ppm.conv4.conv.1.weight", "global_feature_extractor.ppm.conv4.conv.1.bias", "global_feature_extractor.ppm.conv4.conv.1.running_mean", "global_feature_extractor.ppm.conv4.conv.1.running_var", "global_feature_extractor.ppm.out.conv.0.weight", "global_feature_extractor.ppm.out.conv.1.weight", "global_feature_extractor.ppm.out.conv.1.bias", "global_feature_extractor.ppm.out.conv.1.running_mean", "global_feature_extractor.ppm.out.conv.1.running_var", "feature_fusion.dwconv.conv.0.weight", "feature_fusion.dwconv.conv.1.weight", "feature_fusion.dwconv.conv.1.bias", "feature_fusion.dwconv.conv.1.running_mean", "feature_fusion.dwconv.conv.1.running_var", "feature_fusion.conv_lower_res.0.weight", "feature_fusion.conv_lower_res.0.bias", "feature_fusion.conv_lower_res.1.weight", "feature_fusion.conv_lower_res.1.bias", "feature_fusion.conv_lower_res.1.running_mean", "feature_fusion.conv_lower_res.1.running_var", "feature_fusion.conv_higher_res.0.weight", "feature_fusion.conv_higher_res.0.bias", "feature_fusion.conv_higher_res.1.weight", "feature_fusion.conv_higher_res.1.bias", "feature_fusion.conv_higher_res.1.running_mean", "feature_fusion.conv_higher_res.1.running_var", "classifier.dsconv1.conv.0.weight", "classifier.dsconv1.conv.1.weight", "classifier.dsconv1.conv.1.bias", "classifier.dsconv1.conv.1.running_mean", "classifier.dsconv1.conv.1.running_var", "classifier.dsconv1.conv.3.weight", "classifier.dsconv1.conv.4.weight", "classifier.dsconv1.conv.4.bias", "classifier.dsconv1.conv.4.running_mean", "classifier.dsconv1.conv.4.running_var", "classifier.dsconv2.conv.0.weight", "classifier.dsconv2.conv.1.weight", "classifier.dsconv2.conv.1.bias", "classifier.dsconv2.conv.1.running_mean", "classifier.dsconv2.conv.1.running_var", "classifier.dsconv2.conv.3.weight", "classifier.dsconv2.conv.4.weight", "classifier.dsconv2.conv.4.bias", "classifier.dsconv2.conv.4.running_mean", "classifier.dsconv2.conv.4.running_var", "classifier.conv.1.weight", "classifier.conv.1.bias". Unexpected key(s) in state_dict: "module.learning_to_downsample.conv.conv.0.weight", "module.learning_to_downsample.conv.conv.1.weight", "module.learning_to_downsample.conv.conv.1.bias", "module.learning_to_downsample.conv.conv.1.running_mean", "module.learning_to_downsample.conv.conv.1.running_var", "module.learning_to_downsample.conv.conv.1.num_batches_tracked", "module.learning_to_downsample.dsconv1.conv.0.weight", "module.learning_to_downsample.dsconv1.conv.1.weight", "module.learning_to_downsample.dsconv1.conv.1.bias", "module.learning_to_downsample.dsconv1.conv.1.running_mean", "module.learning_to_downsample.dsconv1.conv.1.running_var", "module.learning_to_downsample.dsconv1.conv.1.num_batches_tracked", "module.learning_to_downsample.dsconv1.conv.3.weight", "module.learning_to_downsample.dsconv1.conv.4.weight", "module.learning_to_downsample.dsconv1.conv.4.bias", "module.learning_to_downsample.dsconv1.conv.4.running_mean", "module.learning_to_downsample.dsconv1.conv.4.running_var", "module.learning_to_downsample.dsconv1.conv.4.num_batches_tracked", "module.learning_to_downsample.dsconv2.conv.0.weight", "module.learning_to_downsample.dsconv2.conv.1.weight", "module.learning_to_downsample.dsconv2.conv.1.bias", "module.learning_to_downsample.dsconv2.conv.1.running_mean", "module.learning_to_downsample.dsconv2.conv.1.running_var", "module.learning_to_downsample.dsconv2.conv.1.num_batches_tracked", "module.learning_to_downsample.dsconv2.conv.3.weight", "module.learning_to_downsample.dsconv2.conv.4.weight", "module.learning_to_downsample.dsconv2.conv.4.bias", "module.learning_to_downsample.dsconv2.conv.4.running_mean", "module.learning_to_downsample.dsconv2.conv.4.running_var", "module.learning_to_downsample.dsconv2.conv.4.num_batches_tracked", "module.global_feature_extractor.bottleneck1.0.block.0.conv.0.weight", "module.global_feature_extractor.bottleneck1.0.block.0.conv.1.weight", "module.global_feature_extractor.bottleneck1.0.block.0.conv.1.bias", "module.global_feature_extractor.bottleneck1.0.block.0.conv.1.running_mean", "module.global_feature_extractor.bottleneck1.0.block.0.conv.1.running_var", "module.global_feature_extractor.bottleneck1.0.block.0.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck1.0.block.1.conv.0.weight", "module.global_feature_extractor.bottleneck1.0.block.1.conv.1.weight", "module.global_feature_extractor.bottleneck1.0.block.1.conv.1.bias", "module.global_feature_extractor.bottleneck1.0.block.1.conv.1.running_mean", "module.global_feature_extractor.bottleneck1.0.block.1.conv.1.running_var", "module.global_feature_extractor.bottleneck1.0.block.1.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck1.0.block.2.weight", "module.global_feature_extractor.bottleneck1.0.block.3.weight", "module.global_feature_extractor.bottleneck1.0.block.3.bias", "module.global_feature_extractor.bottleneck1.0.block.3.running_mean", "module.global_feature_extractor.bottleneck1.0.block.3.running_var", "module.global_feature_extractor.bottleneck1.0.block.3.num_batches_tracked", "module.global_feature_extractor.bottleneck1.1.block.0.conv.0.weight", "module.global_feature_extractor.bottleneck1.1.block.0.conv.1.weight", "module.global_feature_extractor.bottleneck1.1.block.0.conv.1.bias", "module.global_feature_extractor.bottleneck1.1.block.0.conv.1.running_mean", "module.global_feature_extractor.bottleneck1.1.block.0.conv.1.running_var", "module.global_feature_extractor.bottleneck1.1.block.0.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck1.1.block.1.conv.0.weight", "module.global_feature_extractor.bottleneck1.1.block.1.conv.1.weight", "module.global_feature_extractor.bottleneck1.1.block.1.conv.1.bias", "module.global_feature_extractor.bottleneck1.1.block.1.conv.1.running_mean", "module.global_feature_extractor.bottleneck1.1.block.1.conv.1.running_var", "module.global_feature_extractor.bottleneck1.1.block.1.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck1.1.block.2.weight", "module.global_feature_extractor.bottleneck1.1.block.3.weight", "module.global_feature_extractor.bottleneck1.1.block.3.bias", "module.global_feature_extractor.bottleneck1.1.block.3.running_mean", "module.global_feature_extractor.bottleneck1.1.block.3.running_var", "module.global_feature_extractor.bottleneck1.1.block.3.num_batches_tracked", "module.global_feature_extractor.bottleneck1.2.block.0.conv.0.weight", "module.global_feature_extractor.bottleneck1.2.block.0.conv.1.weight", "module.global_feature_extractor.bottleneck1.2.block.0.conv.1.bias", "module.global_feature_extractor.bottleneck1.2.block.0.conv.1.running_mean", "module.global_feature_extractor.bottleneck1.2.block.0.conv.1.running_var", "module.global_feature_extractor.bottleneck1.2.block.0.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck1.2.block.1.conv.0.weight", "module.global_feature_extractor.bottleneck1.2.block.1.conv.1.weight", "module.global_feature_extractor.bottleneck1.2.block.1.conv.1.bias", "module.global_feature_extractor.bottleneck1.2.block.1.conv.1.running_mean", "module.global_feature_extractor.bottleneck1.2.block.1.conv.1.running_var", "module.global_feature_extractor.bottleneck1.2.block.1.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck1.2.block.2.weight", "module.global_feature_extractor.bottleneck1.2.block.3.weight", "module.global_feature_extractor.bottleneck1.2.block.3.bias", "module.global_feature_extractor.bottleneck1.2.block.3.running_mean", "module.global_feature_extractor.bottleneck1.2.block.3.running_var", "module.global_feature_extractor.bottleneck1.2.block.3.num_batches_tracked", "module.global_feature_extractor.bottleneck2.0.block.0.conv.0.weight", "module.global_feature_extractor.bottleneck2.0.block.0.conv.1.weight", "module.global_feature_extractor.bottleneck2.0.block.0.conv.1.bias", "module.global_feature_extractor.bottleneck2.0.block.0.conv.1.running_mean", "module.global_feature_extractor.bottleneck2.0.block.0.conv.1.running_var", "module.global_feature_extractor.bottleneck2.0.block.0.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck2.0.block.1.conv.0.weight", "module.global_feature_extractor.bottleneck2.0.block.1.conv.1.weight", "module.global_feature_extractor.bottleneck2.0.block.1.conv.1.bias", "module.global_feature_extractor.bottleneck2.0.block.1.conv.1.running_mean", "module.global_feature_extractor.bottleneck2.0.block.1.conv.1.running_var", "module.global_feature_extractor.bottleneck2.0.block.1.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck2.0.block.2.weight", "module.global_feature_extractor.bottleneck2.0.block.3.weight", "module.global_feature_extractor.bottleneck2.0.block.3.bias", "module.global_feature_extractor.bottleneck2.0.block.3.running_mean", "module.global_feature_extractor.bottleneck2.0.block.3.running_var", "module.global_feature_extractor.bottleneck2.0.block.3.num_batches_tracked", "module.global_feature_extractor.bottleneck2.1.block.0.conv.0.weight", "module.global_feature_extractor.bottleneck2.1.block.0.conv.1.weight", "module.global_feature_extractor.bottleneck2.1.block.0.conv.1.bias", "module.global_feature_extractor.bottleneck2.1.block.0.conv.1.running_mean", "module.global_feature_extractor.bottleneck2.1.block.0.conv.1.running_var", "module.global_feature_extractor.bottleneck2.1.block.0.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck2.1.block.1.conv.0.weight", "module.global_feature_extractor.bottleneck2.1.block.1.conv.1.weight", "module.global_feature_extractor.bottleneck2.1.block.1.conv.1.bias", "module.global_feature_extractor.bottleneck2.1.block.1.conv.1.running_mean", "module.global_feature_extractor.bottleneck2.1.block.1.conv.1.running_var", "module.global_feature_extractor.bottleneck2.1.block.1.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck2.1.block.2.weight", "module.global_feature_extractor.bottleneck2.1.block.3.weight", "module.global_feature_extractor.bottleneck2.1.block.3.bias", "module.global_feature_extractor.bottleneck2.1.block.3.running_mean", "module.global_feature_extractor.bottleneck2.1.block.3.running_var", "module.global_feature_extractor.bottleneck2.1.block.3.num_batches_tracked", "module.global_feature_extractor.bottleneck2.2.block.0.conv.0.weight", "module.global_feature_extractor.bottleneck2.2.block.0.conv.1.weight", "module.global_feature_extractor.bottleneck2.2.block.0.conv.1.bias", "module.global_feature_extractor.bottleneck2.2.block.0.conv.1.running_mean", "module.global_feature_extractor.bottleneck2.2.block.0.conv.1.running_var", "module.global_feature_extractor.bottleneck2.2.block.0.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck2.2.block.1.conv.0.weight", "module.global_feature_extractor.bottleneck2.2.block.1.conv.1.weight", "module.global_feature_extractor.bottleneck2.2.block.1.conv.1.bias", "module.global_feature_extractor.bottleneck2.2.block.1.conv.1.running_mean", "module.global_feature_extractor.bottleneck2.2.block.1.conv.1.running_var", "module.global_feature_extractor.bottleneck2.2.block.1.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck2.2.block.2.weight", "module.global_feature_extractor.bottleneck2.2.block.3.weight", "module.global_feature_extractor.bottleneck2.2.block.3.bias", "module.global_feature_extractor.bottleneck2.2.block.3.running_mean", "module.global_feature_extractor.bottleneck2.2.block.3.running_var", "module.global_feature_extractor.bottleneck2.2.block.3.num_batches_tracked", "module.global_feature_extractor.bottleneck3.0.block.0.conv.0.weight", "module.global_feature_extractor.bottleneck3.0.block.0.conv.1.weight", "module.global_feature_extractor.bottleneck3.0.block.0.conv.1.bias", "module.global_feature_extractor.bottleneck3.0.block.0.conv.1.running_mean", "module.global_feature_extractor.bottleneck3.0.block.0.conv.1.running_var", "module.global_feature_extractor.bottleneck3.0.block.0.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck3.0.block.1.conv.0.weight", "module.global_feature_extractor.bottleneck3.0.block.1.conv.1.weight", "module.global_feature_extractor.bottleneck3.0.block.1.conv.1.bias", "module.global_feature_extractor.bottleneck3.0.block.1.conv.1.running_mean", "module.global_feature_extractor.bottleneck3.0.block.1.conv.1.running_var", "module.global_feature_extractor.bottleneck3.0.block.1.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck3.0.block.2.weight", "module.global_feature_extractor.bottleneck3.0.block.3.weight", "module.global_feature_extractor.bottleneck3.0.block.3.bias", "module.global_feature_extractor.bottleneck3.0.block.3.running_mean", "module.global_feature_extractor.bottleneck3.0.block.3.running_var", "module.global_feature_extractor.bottleneck3.0.block.3.num_batches_tracked", "module.global_feature_extractor.bottleneck3.1.block.0.conv.0.weight", "module.global_feature_extractor.bottleneck3.1.block.0.conv.1.weight", "module.global_feature_extractor.bottleneck3.1.block.0.conv.1.bias", "module.global_feature_extractor.bottleneck3.1.block.0.conv.1.running_mean", "module.global_feature_extractor.bottleneck3.1.block.0.conv.1.running_var", "module.global_feature_extractor.bottleneck3.1.block.0.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck3.1.block.1.conv.0.weight", "module.global_feature_extractor.bottleneck3.1.block.1.conv.1.weight", "module.global_feature_extractor.bottleneck3.1.block.1.conv.1.bias", "module.global_feature_extractor.bottleneck3.1.block.1.conv.1.running_mean", "module.global_feature_extractor.bottleneck3.1.block.1.conv.1.running_var", "module.global_feature_extractor.bottleneck3.1.block.1.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck3.1.block.2.weight", "module.global_feature_extractor.bottleneck3.1.block.3.weight", "module.global_feature_extractor.bottleneck3.1.block.3.bias", "module.global_feature_extractor.bottleneck3.1.block.3.running_mean", "module.global_feature_extractor.bottleneck3.1.block.3.running_var", "module.global_feature_extractor.bottleneck3.1.block.3.num_batches_tracked", "module.global_feature_extractor.bottleneck3.2.block.0.conv.0.weight", "module.global_feature_extractor.bottleneck3.2.block.0.conv.1.weight", "module.global_feature_extractor.bottleneck3.2.block.0.conv.1.bias", "module.global_feature_extractor.bottleneck3.2.block.0.conv.1.running_mean", "module.global_feature_extractor.bottleneck3.2.block.0.conv.1.running_var", "module.global_feature_extractor.bottleneck3.2.block.0.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck3.2.block.1.conv.0.weight", "module.global_feature_extractor.bottleneck3.2.block.1.conv.1.weight", "module.global_feature_extractor.bottleneck3.2.block.1.conv.1.bias", "module.global_feature_extractor.bottleneck3.2.block.1.conv.1.running_mean", "module.global_feature_extractor.bottleneck3.2.block.1.conv.1.running_var", "module.global_feature_extractor.bottleneck3.2.block.1.conv.1.num_batches_tracked", "module.global_feature_extractor.bottleneck3.2.block.2.weight", "module.global_feature_extractor.bottleneck3.2.block.3.weight", "module.global_feature_extractor.bottleneck3.2.block.3.bias", "module.global_feature_extractor.bottleneck3.2.block.3.running_mean", "module.global_feature_extractor.bottleneck3.2.block.3.running_var", "module.global_feature_extractor.bottleneck3.2.block.3.num_batches_tracked", "module.global_feature_extractor.ppm.conv1.conv.0.weight", "module.global_feature_extractor.ppm.conv1.conv.1.weight", "module.global_feature_extractor.ppm.conv1.conv.1.bias", "module.global_feature_extractor.ppm.conv1.conv.1.running_mean", "module.global_feature_extractor.ppm.conv1.conv.1.running_var", "module.global_feature_extractor.ppm.conv1.conv.1.num_batches_tracked", "module.global_feature_extractor.ppm.conv2.conv.0.weight", "module.global_feature_extractor.ppm.conv2.conv.1.weight", "module.global_feature_extractor.ppm.conv2.conv.1.bias", "module.global_feature_extractor.ppm.conv2.conv.1.running_mean", "module.global_feature_extractor.ppm.conv2.conv.1.running_var", "module.global_feature_extractor.ppm.conv2.conv.1.num_batches_tracked", "module.global_feature_extractor.ppm.conv3.conv.0.weight", "module.global_feature_extractor.ppm.conv3.conv.1.weight", "module.global_feature_extractor.ppm.conv3.conv.1.bias", "module.global_feature_extractor.ppm.conv3.conv.1.running_mean", "module.global_feature_extractor.ppm.conv3.conv.1.running_var", "module.global_feature_extractor.ppm.conv3.conv.1.num_batches_tracked", "module.global_feature_extractor.ppm.conv4.conv.0.weight", "module.global_feature_extractor.ppm.conv4.conv.1.weight", "module.global_feature_extractor.ppm.conv4.conv.1.bias", "module.global_feature_extractor.ppm.conv4.conv.1.running_mean", "module.global_feature_extractor.ppm.conv4.conv.1.running_var", "module.global_feature_extractor.ppm.conv4.conv.1.num_batches_tracked", "module.global_feature_extractor.ppm.out.conv.0.weight", "module.global_feature_extractor.ppm.out.conv.1.weight", "module.global_feature_extractor.ppm.out.conv.1.bias", "module.global_feature_extractor.ppm.out.conv.1.running_mean", "module.global_feature_extractor.ppm.out.conv.1.running_var", "module.global_feature_extractor.ppm.out.conv.1.num_batches_tracked", "module.feature_fusion.dwconv.conv.0.weight", "module.feature_fusion.dwconv.conv.1.weight", "module.feature_fusion.dwconv.conv.1.bias", "module.feature_fusion.dwconv.conv.1.running_mean", "module.feature_fusion.dwconv.conv.1.running_var", "module.feature_fusion.dwconv.conv.1.num_batches_tracked", "module.feature_fusion.conv_lower_res.0.weight", "module.feature_fusion.conv_lower_res.0.bias", "module.feature_fusion.conv_lower_res.1.weight", "module.feature_fusion.conv_lower_res.1.bias", "module.feature_fusion.conv_lower_res.1.running_mean", "module.feature_fusion.conv_lower_res.1.running_var", "module.feature_fusion.conv_lower_res.1.num_batches_tracked", "module.feature_fusion.conv_higher_res.0.weight", "module.feature_fusion.conv_higher_res.0.bias", "module.feature_fusion.conv_higher_res.1.weight", "module.feature_fusion.conv_higher_res.1.bias", "module.feature_fusion.conv_higher_res.1.running_mean", "module.feature_fusion.conv_higher_res.1.running_var", "module.feature_fusion.conv_higher_res.1.num_batches_tracked", "module.classifier.dsconv1.conv.0.weight", "module.classifier.dsconv1.conv.1.weight", "module.classifier.dsconv1.conv.1.bias", "module.classifier.dsconv1.conv.1.running_mean", "module.classifier.dsconv1.conv.1.running_var", "module.classifier.dsconv1.conv.1.num_batches_tracked", "module.classifier.dsconv1.conv.3.weight", "module.classifier.dsconv1.conv.4.weight", "module.classifier.dsconv1.conv.4.bias", "module.classifier.dsconv1.conv.4.running_mean", "module.classifier.dsconv1.conv.4.running_var", "module.classifier.dsconv1.conv.4.num_batches_tracked", "module.classifier.dsconv2.conv.0.weight", "module.classifier.dsconv2.conv.1.weight", "module.classifier.dsconv2.conv.1.bias", "module.classifier.dsconv2.conv.1.running_mean", "module.classifier.dsconv2.conv.1.running_var", "module.classifier.dsconv2.conv.1.num_batches_tracked", "module.classifier.dsconv2.conv.3.weight", "module.classifier.dsconv2.conv.4.weight", "module.classifier.dsconv2.conv.4.bias", "module.classifier.dsconv2.conv.4.running_mean", "module.classifier.dsconv2.conv.4.running_var", "module.classifier.dsconv2.conv.4.num_batches_tracked", "module.classifier.conv.1.weight", "module.classifier.conv.1.bias".