I’m trying to load the weights of a Pytorch model but getting this error: _pickle.UnpicklingError: invalid load key, 'x1f'.
Here is the weights loading code:
<code>import os
import torch
import numpy as np
# from data_loader import VideoDataset
import timm
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print('Device being used:', device)
mname = os.path.join('./CDF2_0.pth')
checkpoints = torch.load(mname, map_location=device)
print("Checkpoint loaded successfully.")
model = timm.create_model('legacy_xception', pretrained=True, num_classes=2).to(device)
model.load_state_dict(checkpoints['state_dict'])
model.eval()
</code>
<code>import os
import torch
import numpy as np
# from data_loader import VideoDataset
import timm
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print('Device being used:', device)
mname = os.path.join('./CDF2_0.pth')
checkpoints = torch.load(mname, map_location=device)
print("Checkpoint loaded successfully.")
model = timm.create_model('legacy_xception', pretrained=True, num_classes=2).to(device)
model.load_state_dict(checkpoints['state_dict'])
model.eval()
</code>
import os
import torch
import numpy as np
# from data_loader import VideoDataset
import timm
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print('Device being used:', device)
mname = os.path.join('./CDF2_0.pth')
checkpoints = torch.load(mname, map_location=device)
print("Checkpoint loaded successfully.")
model = timm.create_model('legacy_xception', pretrained=True, num_classes=2).to(device)
model.load_state_dict(checkpoints['state_dict'])
model.eval()
I have tried with different Pytorch versions. I have tried to inspect the weights by changing the extension to .zip and opening with archive manager but can’t fix the issue. Here is a public link to the weights .pth file, I’m trying to load. Any help is highly appreciated!