search

Lornatang / SRGAN-PyTorch

A simple and complete implementation of super-resolution paper.
Apache License 2.0
410 stars 105 forks source link

train_gan.py : Model weight file not found, state_dict Missing key(s), Unexpected key(s) #84

Closed syt06162 closed 1 year ago

syt06162 commented 1 year ago

python3 train_gan.py --config_path ./configs/train/SRGAN_X4.yaml

1. The first problem I encountered was

'FileNotFoundError: Model weight file not found'

By the way, this was solved by adding this to /configs/train/SRGAN_X4.yaml MODEL_WEIGHTS_PATH: "./results/pretrained_models/SRGAN_x4-ImageNet.pth.tar"

(Of course, I downloaded the weights guided by readme and put them in the /results/prepared_models location.)

Is this right?

2. However, there is a new problem.

RuntimeError: Error(s) in loading state_dict for _FeatureExtractor:
        Missing key(s) in state_dict: "features.0.weight", "features.0.bias", "features.2.weight", "features.2.bias", "features.5.weight", "features.5.bias", "features.7.weight", "features.7.bias", "features.10.weight", "features.10.bias", "features.12.weight", "features.12.bias", "features.14.weight", "features.14.bias", "features.16.weight", "features.16.bias", "features.19.weight", "features.19.bias", "features.21.weight", "features.21.bias", "features.23.weight", "features.23.bias", "features.25.weight", "features.25.bias", "features.28.weight", "features.28.bias", "features.30.weight", "features.30.bias", "features.32.weight", "features.32.bias", "features.34.weight", "features.34.bias", "classifier.0.weight", "classifier.0.bias", "classifier.3.weight", "classifier.3.bias", "classifier.6.weight", "classifier.6.bias". 
        Unexpected key(s) in state_dict: "conv1.0.weight", "conv1.0.bias", "conv1.1.weight", "trunk.0.rcb.0.weight", "trunk.0.rcb.1.weight", "trunk.0.rcb.1.bias", "trunk.0.rcb.1.running_mean", "trunk.0.rcb.1.running_var", "trunk.0.rcb.1.num_batches_tracked", "trunk.0.rcb.2.weight", "trunk.0.rcb.3.weight", "trunk.0.rcb.4.weight", "trunk.0.rcb.4.bias", "trunk.0.rcb.4.running_mean", "trunk.0.rcb.4.running_var", "trunk.0.rcb.4.num_batches_tracked", "trunk.1.rcb.0.weight", "trunk.1.rcb.1.weight", "trunk.1.rcb.1.bias", "trunk.1.rcb.1.running_mean", "trunk.1.rcb.1.running_var", "trunk.1.rcb.1.num_batches_tracked", "trunk.1.rcb.2.weight", "trunk.1.rcb.3.weight", "trunk.1.rcb.4.weight", "trunk.1.rcb.4.bias", "trunk.1.rcb.4.running_mean", "trunk.1.rcb.4.running_var", "trunk.1.rcb.4.num_batches_tracked", "trunk.2.rcb.0.weight", "trunk.2.rcb.1.weight", "trunk.2.rcb.1.bias", "trunk.2.rcb.1.running_mean", "trunk.2.rcb.1.running_var", "trunk.2.rcb.1.num_batches_tracked", "trunk.2.rcb.2.weight", "trunk.2.rcb.3.weight", "trunk.2.rcb.4.weight", "trunk.2.rcb.4.bias", "trunk.2.rcb.4.running_mean", "trunk.2.rcb.4.running_var", "trunk.2.rcb.4.num_batches_tracked", "trunk.3.rcb.0.weight", "trunk.3.rcb.1.weight", "trunk.3.rcb.1.bias", "trunk.3.rcb.1.running_mean", "trunk.3.rcb.1.running_var", "trunk.3.rcb.1.num_batches_tracked", "trunk.3.rcb.2.weight", "trunk.3.rcb.3.weight", "trunk.3.rcb.4.weight", "trunk.3.rcb.4.bias", "trunk.3.rcb.4.running_mean", "trunk.3.rcb.4.running_var", "trunk.3.rcb.4.num_batches_tracked", "trunk.4.rcb.0.weight", "trunk.4.rcb.1.weight", "trunk.4.rcb.1.bias", "trunk.4.rcb.1.running_mean", "trunk.4.rcb.1.running_var", "trunk.4.rcb.1.num_batches_tracked", "trunk.4.rcb.2.weight", "trunk.4.rcb.3.weight", "trunk.4.rcb.4.weight", "trunk.4.rcb.4.bias", "trunk.4.rcb.4.running_mean", "trunk.4.rcb.4.running_var", "trunk.4.rcb.4.num_batches_tracked", "trunk.5.rcb.0.weight", "trunk.5.rcb.1.weight", "trunk.5.rcb.1.bias", "trunk.5.rcb.1.running_mean", "trunk.5.rcb.1.running_var", "trunk.5.rcb.1.num_batches_tracked", "trunk.5.rcb.2.weight", "trunk.5.rcb.3.weight", "trunk.5.rcb.4.weight", "trunk.5.rcb.4.bias", "trunk.5.rcb.4.running_mean", "trunk.5.rcb.4.running_var", "trunk.5.rcb.4.num_batches_tracked", "trunk.6.rcb.0.weight", "trunk.6.rcb.1.weight", "trunk.6.rcb.1.bias", "trunk.6.rcb.1.running_mean", "trunk.6.rcb.1.running_var", "trunk.6.rcb.1.num_batches_tracked", "trunk.6.rcb.2.weight", "trunk.6.rcb.3.weight", "trunk.6.rcb.4.weight", "trunk.6.rcb.4.bias", "trunk.6.rcb.4.running_mean", "trunk.6.rcb.4.running_var", "trunk.6.rcb.4.num_batches_tracked", "trunk.7.rcb.0.weight", "trunk.7.rcb.1.weight", "trunk.7.rcb.1.bias", "trunk.7.rcb.1.running_mean", "trunk.7.rcb.1.running_var", "trunk.7.rcb.1.num_batches_tracked", "trunk.7.rcb.2.weight", "trunk.7.rcb.3.weight", "trunk.7.rcb.4.weight", "trunk.7.rcb.4.bias", "trunk.7.rcb.4.running_mean", "trunk.7.rcb.4.running_var", "trunk.7.rcb.4.num_batches_tracked", "trunk.8.rcb.0.weight", "trunk.8.rcb.1.weight", "trunk.8.rcb.1.bias", "trunk.8.rcb.1.running_mean", "trunk.8.rcb.1.running_var", "trunk.8.rcb.1.num_batches_tracked", "trunk.8.rcb.2.weight", "trunk.8.rcb.3.weight", "trunk.8.rcb.4.weight", "trunk.8.rcb.4.bias", "trunk.8.rcb.4.running_mean", "trunk.8.rcb.4.running_var", "trunk.8.rcb.4.num_batches_tracked", "trunk.9.rcb.0.weight", "trunk.9.rcb.1.weight", "trunk.9.rcb.1.bias", "trunk.9.rcb.1.running_mean", "trunk.9.rcb.1.running_var", "trunk.9.rcb.1.num_batches_tracked", "trunk.9.rcb.2.weight", "trunk.9.rcb.3.weight", "trunk.9.rcb.4.weight", "trunk.9.rcb.4.bias", "trunk.9.rcb.4.running_mean", "trunk.9.rcb.4.running_var", "trunk.9.rcb.4.num_batches_tracked", "trunk.10.rcb.0.weight", "trunk.10.rcb.1.weight", "trunk.10.rcb.1.bias", "trunk.10.rcb.1.running_mean", "trunk.10.rcb.1.running_var", "trunk.10.rcb.1.num_batches_tracked", "trunk.10.rcb.2.weight", "trunk.10.rcb.3.weight", "trunk.10.rcb.4.weight", "trunk.10.rcb.4.bias", "trunk.10.rcb.4.running_mean", "trunk.10.rcb.4.running_var", "trunk.10.rcb.4.num_batches_tracked", "trunk.11.rcb.0.weight", "trunk.11.rcb.1.weight", "trunk.11.rcb.1.bias", "trunk.11.rcb.1.running_mean", "trunk.11.rcb.1.running_var", "trunk.11.rcb.1.num_batches_tracked", "trunk.11.rcb.2.weight", "trunk.11.rcb.3.weight", "trunk.11.rcb.4.weight", "trunk.11.rcb.4.bias", "trunk.11.rcb.4.running_mean", "trunk.11.rcb.4.running_var", "trunk.11.rcb.4.num_batches_tracked", "trunk.12.rcb.0.weight", "trunk.12.rcb.1.weight", "trunk.12.rcb.1.bias", "trunk.12.rcb.1.running_mean", "trunk.12.rcb.1.running_var", "trunk.12.rcb.1.num_batches_tracked", "trunk.12.rcb.2.weight", "trunk.12.rcb.3.weight", "trunk.12.rcb.4.weight", "trunk.12.rcb.4.bias", "trunk.12.rcb.4.running_mean", "trunk.12.rcb.4.running_var", "trunk.12.rcb.4.num_batches_tracked", "trunk.13.rcb.0.weight", "trunk.13.rcb.1.weight", "trunk.13.rcb.1.bias", "trunk.13.rcb.1.running_mean", "trunk.13.rcb.1.running_var", "trunk.13.rcb.1.num_batches_tracked", "trunk.13.rcb.2.weight", "trunk.13.rcb.3.weight", "trunk.13.rcb.4.weight", "trunk.13.rcb.4.bias", "trunk.13.rcb.4.running_mean", "trunk.13.rcb.4.running_var", "trunk.13.rcb.4.num_batches_tracked", "trunk.14.rcb.0.weight", "trunk.14.rcb.1.weight", "trunk.14.rcb.1.bias", "trunk.14.rcb.1.running_mean", "trunk.14.rcb.1.running_var", "trunk.14.rcb.1.num_batches_tracked", "trunk.14.rcb.2.weight", "trunk.14.rcb.3.weight", "trunk.14.rcb.4.weight", "trunk.14.rcb.4.bias", "trunk.14.rcb.4.running_mean", "trunk.14.rcb.4.running_var", "trunk.14.rcb.4.num_batches_tracked", "trunk.15.rcb.0.weight", "trunk.15.rcb.1.weight", "trunk.15.rcb.1.bias", "trunk.15.rcb.1.running_mean", "trunk.15.rcb.1.running_var", "trunk.15.rcb.1.num_batches_tracked", "trunk.15.rcb.2.weight", "trunk.15.rcb.3.weight", "trunk.15.rcb.4.weight", "trunk.15.rcb.4.bias", "trunk.15.rcb.4.running_mean", "trunk.15.rcb.4.running_var", "trunk.15.rcb.4.num_batches_tracked", "conv2.0.weight", "conv2.1.weight", "conv2.1.bias", "conv2.1.running_mean", "conv2.1.running_var", "conv2.1.num_batches_tracked", "upsampling.0.upsample_block.0.weight", "upsampling.0.upsample_block.0.bias", "upsampling.0.upsample_block.2.weight", "upsampling.1.upsample_block.0.weight", "upsampling.1.upsample_block.0.bias", "upsampling.1.upsample_block.2.weight", "conv3.weight", "conv3.bias".

This seems to be something wrong with loading the weight values. As the readme guided me, I trained SRresnet first and trained SRgan. Are all the weights of SRgan uploaded on google drive correct?

Lornatang commented 1 year ago

Now all model weights have been updated