The pretrain loading doesn't need "module."

https://github.com/facebookresearch/moco/blob/78b69cafae80bc74cd1a89ac3fb365dc20d157d3/main_lincls.py#L165,
https://github.com/facebookresearch/moco/blob/78b69cafae80bc74cd1a89ac3fb365dc20d157d3/main_lincls.py#L167,

I print out state_dict.keys()
`['queue', 'queue_ptr', 'encoder_q.conv1.weight', 'encoder_q.bn1.weight', 'encoder_q.bn1.bias', 'encoder_q.bn1.running_mean', 'encoder_q.bn1.running_var', 'encoder_q.bn1.num_batches_tracked', 'encoder_q.layer1.0.conv1.weight', 'encoder_q.layer1.0.bn1.weight', 'encoder_q.layer1.0.bn1.bias', 'encoder_q.layer1.0.bn1.running_mean', 'encoder_q.layer1.0.bn1.running_var', 'encoder_q.layer1.0.bn1.num_batches_tracked', 'encoder_q.layer1.0.conv2.weight', 'encoder_q.layer1.0.bn2.weight', 'encoder_q.layer1.0.bn2.bias', 'encoder_q.layer1.0.bn2.running_mean', 'encoder_q.layer1.0.bn2.running_var', 'encoder_q.layer1.0.bn2.num_batches_tracked', 'encoder_q.layer1.0.conv3.weight', 'encoder_q.layer1.0.bn3.weight', 'encoder_q.layer1.0.bn3.bias', 'encoder_q.layer1.0.bn3.running_mean', 'encoder_q.layer1.0.bn3.running_var', 'encoder_q.layer1.0.bn3.num_batches_tracked', 'encoder_q.layer1.0.downsample.0.weight', 'encoder_q.layer1.0.downsample.1.weight', 'encoder_q.layer1.0.downsample.1.bias', 'encoder_q.layer1.0.downsample.1.running_mean', 'encoder_q.layer1.0.downsample.1.running_var', 'encoder_q.layer1.0.downsample.1.num_batches_tracked', 'encoder_q.layer1.1.conv1.weight', 'encoder_q.layer1.1.bn1.weight', 'encoder_q.layer1.1.bn1.bias', 'encoder_q.layer1.1.bn1.running_mean', 'encoder_q.layer1.1.bn1.running_var', 'encoder_q.layer1.1.bn1.num_batches_tracked', 'encoder_q.layer1.1.conv2.weight', 'encoder_q.layer1.1.bn2.weight', 'encoder_q.layer1.1.bn2.bias', 'encoder_q.layer1.1.bn2.running_mean', 'encoder_q.layer1.1.bn2.running_var', 'encoder_q.layer1.1.bn2.num_batches_tracked', 'encoder_q.layer1.1.conv3.weight', 'encoder_q.layer1.1.bn3.weight', 'encoder_q.layer1.1.bn3.bias', 'encoder_q.layer1.1.bn3.running_mean', 'encoder_q.layer1.1.bn3.running_var', 'encoder_q.layer1.1.bn3.num_batches_tracked', 'encoder_q.layer1.2.conv1.weight', 'encoder_q.layer1.2.bn1.weight', 'encoder_q.layer1.2.bn1.bias', 'encoder_q.layer1.2.bn1.running_mean', 'encoder_q.layer1.2.bn1.running_var', 'encoder_q.layer1.2.bn1.num_batches_tracked', 'encoder_q.layer1.2.conv2.weight', 'encoder_q.layer1.2.bn2.weight', 'encoder_q.layer1.2.bn2.bias', 'encoder_q.layer1.2.bn2.running_mean', 'encoder_q.layer1.2.bn2.running_var', 'encoder_q.layer1.2.bn2.num_batches_tracked', 'encoder_q.layer1.2.conv3.weight', 'encoder_q.layer1.2.bn3.weight', 'encoder_q.layer1.2.bn3.bias', 'encoder_q.layer1.2.bn3.running_mean', 'encoder_q.layer1.2.bn3.running_var', 'encoder_q.layer1.2.bn3.num_batches_tracked', 'encoder_q.layer2.0.conv1.weight', 'encoder_q.layer2.0.bn1.weight', 'encoder_q.layer2.0.bn1.bias', 'encoder_q.layer2.0.bn1.running_mean', 'encoder_q.layer2.0.bn1.running_var', 'encoder_q.layer2.0.bn1.num_batches_tracked', 'encoder_q.layer2.0.conv2.weight', 'encoder_q.layer2.0.bn2.weight', 'encoder_q.layer2.0.bn2.bias', 'encoder_q.layer2.0.bn2.running_mean', 'encoder_q.layer2.0.bn2.running_var', 'encoder_q.layer2.0.bn2.num_batches_tracked', 'encoder_q.layer2.0.conv3.weight', 'encoder_q.layer2.0.bn3.weight', 'encoder_q.layer2.0.bn3.bias', 'encoder_q.layer2.0.bn3.running_mean', 'encoder_q.layer2.0.bn3.running_var', 'encoder_q.layer2.0.bn3.num_batches_tracked', 'encoder_q.layer2.0.downsample.0.weight', 'encoder_q.layer2.0.downsample.1.weight', 'encoder_q.layer2.0.downsample.1.bias', 'encoder_q.layer2.0.downsample.1.running_mean', 'encoder_q.layer2.0.downsample.1.running_var', 'encoder_q.layer2.0.downsample.1.num_batches_tracked', 'encoder_q.layer2.1.conv1.weight', 'encoder_q.layer2.1.bn1.weight', 'encoder_q.layer2.1.bn1.bias', 'encoder_q.layer2.1.bn1.running_mean', 'encoder_q.layer2.1.bn1.running_var', 'encoder_q.layer2.1.bn1.num_batches_tracked', 'encoder_q.layer2.1.conv2.weight', 'encoder_q.layer2.1.bn2.weight', 'encoder_q.layer2.1.bn2.bias', 'encoder_q.layer2.1.bn2.running_mean', 'encoder_q.layer2.1.bn2.running_var', 'encoder_q.layer2.1.bn2.num_batches_tracked', 'encoder_q.layer2.1.conv3.weight', 'encoder_q.layer2.1.bn3.weight', 'encoder_q.layer2.1.bn3.bias', 'encoder_q.layer2.1.bn3.running_mean', 'encoder_q.layer2.1.bn3.running_var', 'encoder_q.layer2.1.bn3.num_batches_tracked', 'encoder_q.layer2.2.conv1.weight', 'encoder_q.layer2.2.bn1.weight', 'encoder_q.layer2.2.bn1.bias', 'encoder_q.layer2.2.bn1.running_mean', 'encoder_q.layer2.2.bn1.running_var', 'encoder_q.layer2.2.bn1.num_batches_tracked', 'encoder_q.layer2.2.conv2.weight', 'encoder_q.layer2.2.bn2.weight', 'encoder_q.layer2.2.bn2.bias', 'encoder_q.layer2.2.bn2.running_mean', 'encoder_q.layer2.2.bn2.running_var', 'encoder_q.layer2.2.bn2.num_batches_tracked', 'encoder_q.layer2.2.conv3.weight', 'encoder_q.layer2.2.bn3.weight', 'encoder_q.layer2.2.bn3.bias', 'encoder_q.layer2.2.bn3.running_mean', 'encoder_q.layer2.2.bn3.running_var', 'encoder_q.layer2.2.bn3.num_batches_tracked', 'encoder_q.layer2.3.conv1.weight',
'encoder_q.layer2.3.bn1.weight', 'encoder_q.layer2.3.bn1.bias', 'encoder_q.layer2.3.bn1.running_mean', 'encoder_q.layer2.3.bn1.running_var', 'encoder_q.layer2.3.bn1.num_batches_tracked', 'encoder_q.layer2.3.conv2.weight', 'encoder_q.layer2.3.bn2.weight', 'encoder_q.layer2.3.bn2.bias', 'encoder_q.layer2.3.bn2.running_mean', 'encoder_q.layer2.3.bn2.running_var', 'encoder_q.layer2.3.bn2.num_batches_tracked', 'encoder_q.layer2.3.conv3.weight', 'encoder_q.layer2.3.bn3.weight',
'encoder_q.layer2.3.bn3.bias', 'encoder_q.layer2.3.bn3.running_mean', 'encoder_q.layer2.3.bn3.running_var', 'encoder_q.layer2.3.bn3.num_batches_tracked', 'encoder_q.layer3.0.conv1.weight', 'encoder_q.layer3.0.bn1.weight', 'encoder_q.layer3.0.bn1.bias', 'encoder_q.layer3.0.bn1.running_mean', 'encoder_q.layer3.0.bn1.running_var', 'encoder_q.layer3.0.bn1.num_batches_tracked', 'encoder_q.layer3.0.conv2.weight', 'encoder_q.layer3.0.bn2.weight', 'encoder_q.layer3.0.bn2.bias', 'encoder_q.layer3.0.bn2.running_mean', 'encoder_q.layer3.0.bn2.running_var', 'encoder_q.layer3.0.bn2.num_batches_tracked', 'encoder_q.layer3.0.conv3.weight', 'encoder_q.layer3.0.bn3.weight', 'encoder_q.layer3.0.bn3.bias', 'encoder_q.layer3.0.bn3.running_mean', 'encoder_q.layer3.0.bn3.running_var', 'encoder_q.layer3.0.bn3.num_batches_tracked', 'encoder_q.layer3.0.downsample.0.weight', 'encoder_q.layer3.0.downsample.1.weight', 'encoder_q.layer3.0.downsample.1.bias', 'encoder_q.layer3.0.downsample.1.running_mean', 'encoder_q.layer3.0.downsample.1.running_var', 'encoder_q.layer3.0.downsample.1.num_batches_tracked', 'encoder_q.layer3.1.conv1.weight', 'encoder_q.layer3.1.bn1.weight', 'encoder_q.layer3.1.bn1.bias', 'encoder_q.layer3.1.bn1.running_mean', 'encoder_q.layer3.1.bn1.running_var', 'encoder_q.layer3.1.bn1.num_batches_tracked', 'encoder_q.layer3.1.conv2.weight', 'encoder_q.layer3.1.bn2.weight', 'encoder_q.layer3.1.bn2.bias', 'encoder_q.layer3.1.bn2.running_mean', 'encoder_q.layer3.1.bn2.running_var', 'encoder_q.layer3.1.bn2.num_batches_tracked', 'encoder_q.layer3.1.conv3.weight', 'encoder_q.layer3.1.bn3.weight', 'encoder_q.layer3.1.bn3.bias', 'encoder_q.layer3.1.bn3.running_mean', 'encoder_q.layer3.1.bn3.running_var', 'encoder_q.layer3.1.bn3.num_batches_tracked', 'encoder_q.layer3.2.conv1.weight', 'encoder_q.layer3.2.bn1.weight', 'encoder_q.layer3.2.bn1.bias', 'encoder_q.layer3.2.bn1.running_mean', 'encoder_q.layer3.2.bn1.running_var', 'encoder_q.layer3.2.bn1.num_batches_tracked', 'encoder_q.layer3.2.conv2.weight', 'encoder_q.layer3.2.bn2.weight', 'encoder_q.layer3.2.bn2.bias', 'encoder_q.layer3.2.bn2.running_mean', 'encoder_q.layer3.2.bn2.running_var', 'encoder_q.layer3.2.bn2.num_batches_tracked', 'encoder_q.layer3.2.conv3.weight', 'encoder_q.layer3.2.bn3.weight', 'encoder_q.layer3.2.bn3.bias', 'encoder_q.layer3.2.bn3.running_mean', 'encoder_q.layer3.2.bn3.running_var', 'encoder_q.layer3.2.bn3.num_batches_tracked', 'encoder_q.layer3.3.conv1.weight', 'encoder_q.layer3.3.bn1.weight', 'encoder_q.layer3.3.bn1.bias', 'encoder_q.layer3.3.bn1.running_mean', 'encoder_q.layer3.3.bn1.running_var', 'encoder_q.layer3.3.bn1.num_batches_tracked', 'encoder_q.layer3.3.conv2.weight', 'encoder_q.layer3.3.bn2.weight', 'encoder_q.layer3.3.bn2.bias', 'encoder_q.layer3.3.bn2.running_mean', 'encoder_q.layer3.3.bn2.running_var', 'encoder_q.layer3.3.bn2.num_batches_tracked', 'encoder_q.layer3.3.conv3.weight', 'encoder_q.layer3.3.bn3.weight', 'encoder_q.layer3.3.bn3.bias', 'encoder_q.layer3.3.bn3.running_mean', 'encoder_q.layer3.3.bn3.running_var', 'encoder_q.layer3.3.bn3.num_batches_tracked', 'encoder_q.layer3.4.conv1.weight', 'encoder_q.layer3.4.bn1.weight', 'encoder_q.layer3.4.bn1.bias', 'encoder_q.layer3.4.bn1.running_mean', 'encoder_q.layer3.4.bn1.running_var', 'encoder_q.layer3.4.bn1.num_batches_tracked', 'encoder_q.layer3.4.conv2.weight', 'encoder_q.layer3.4.bn2.weight', 'encoder_q.layer3.4.bn2.bias', 'encoder_q.layer3.4.bn2.running_mean', 'encoder_q.layer3.4.bn2.running_var', 'encoder_q.layer3.4.bn2.num_batches_tracked', 'encoder_q.layer3.4.conv3.weight', 'encoder_q.layer3.4.bn3.weight', 'encoder_q.layer3.4.bn3.bias', 'encoder_q.layer3.4.bn3.running_mean', 'encoder_q.layer3.4.bn3.running_var', 'encoder_q.layer3.4.bn3.num_batches_tracked', 'encoder_q.layer3.5.conv1.weight', 'encoder_q.layer3.5.bn1.weight', 'encoder_q.layer3.5.bn1.bias', 'encoder_q.layer3.5.bn1.running_mean', 'encoder_q.layer3.5.bn1.running_var', 'encoder_q.layer3.5.bn1.num_batches_tracked', 'encoder_q.layer3.5.conv2.weight', 'encoder_q.layer3.5.bn2.weight', 'encoder_q.layer3.5.bn2.bias', 'encoder_q.layer3.5.bn2.running_mean', 'encoder_q.layer3.5.bn2.running_var', 'encoder_q.layer3.5.bn2.num_batches_tracked', 'encoder_q.layer3.5.conv3.weight', 'encoder_q.layer3.5.bn3.weight', 'encoder_q.layer3.5.bn3.bias', 'encoder_q.layer3.5.bn3.running_mean', 'encoder_q.layer3.5.bn3.running_var', 'encoder_q.layer3.5.bn3.num_batches_tracked', 'encoder_q.layer3.6.conv1.weight', 'encoder_q.layer3.6.bn1.weight', 'encoder_q.layer3.6.bn1.bias', 'encoder_q.layer3.6.bn1.running_mean', 'encoder_q.layer3.6.bn1.running_var', 'encoder_q.layer3.6.bn1.num_batches_tracked', 'encoder_q.layer3.6.conv2.weight', 'encoder_q.layer3.6.bn2.weight', 'encoder_q.layer3.6.bn2.bias', 'encoder_q.layer3.6.bn2.running_mean', 'encoder_q.layer3.6.bn2.running_var', 'encoder_q.layer3.6.bn2.num_batches_tracked', 'encoder_q.layer3.6.conv3.weight', 'encoder_q.layer3.6.bn3.weight', 'encoder_q.layer3.6.bn3.bias', 'encoder_q.layer3.6.bn3.running_mean', 'encoder_q.layer3.6.bn3.running_var', 'encoder_q.layer3.6.bn3.num_batches_tracked', 'encoder_q.layer3.7.conv1.weight', 'encoder_q.layer3.7.bn1.weight', 'encoder_q.layer3.7.bn1.bias', 'encoder_q.layer3.7.bn1.running_mean', 'encoder_q.layer3.7.bn1.running_var', 'encoder_q.layer3.7.bn1.num_batches_tracked', 'encoder_q.layer3.7.conv2.weight', 'encoder_q.layer3.7.bn2.weight', 'encoder_q.layer3.7.bn2.bias', 'encoder_q.layer3.7.bn2.running_mean', 'encoder_q.layer3.7.bn2.running_var', 'encoder_q.layer3.7.bn2.num_batches_tracked', 'encoder_q.layer3.7.conv3.weight', 'encoder_q.layer3.7.bn3.weight', 'encoder_q.layer3.7.bn3.bias', 'encoder_q.layer3.7.bn3.running_mean', 'encoder_q.layer3.7.bn3.running_var', 'encoder_q.layer3.7.bn3.num_batches_tracked', 'encoder_q.layer3.8.conv1.weight', 'encoder_q.layer3.8.bn1.weight', 'encoder_q.layer3.8.bn1.bias', 'encoder_q.layer3.8.bn1.running_mean', 'encoder_q.layer3.8.bn1.running_var', 'encoder_q.layer3.8.bn1.num_batches_tracked', 'encoder_q.layer3.8.conv2.weight', 'encoder_q.layer3.8.bn2.weight', 'encoder_q.layer3.8.bn2.bias', 'encoder_q.layer3.8.bn2.running_mean', 'encoder_q.layer3.8.bn2.running_var', 'encoder_q.layer3.8.bn2.num_batches_tracked', 'encoder_q.layer3.8.conv3.weight', 'encoder_q.layer3.8.bn3.weight', 'encoder_q.layer3.8.bn3.bias', 'encoder_q.layer3.8.bn3.running_mean', 'encoder_q.layer3.8.bn3.running_var', 'encoder_q.layer3.8.bn3.num_batches_tracked', 'encoder_q.layer3.9.conv1.weight', 'encoder_q.layer3.9.bn1.weight', 'encoder_q.layer3.9.bn1.bias', 'encoder_q.layer3.9.bn1`

so remove the "module.", the loading will be correct

(env : torch 1.6, python 3.7)