Getting very high validation loss for skewed dataset

vision
1

I have a very skewed dataset in which number of class samples is [74,859]. My training method is correct so I feel I am making an error is using the weighted random sampler and weights in cross entropy. Is my implementation correct? I used WeightedRandomSampler from train dataloader and SubsetRandomSampler for test dataloader.

splits = 10

kfold = KFold(n_splits=splits,shuffle=False)

for fold,(train_idx,test_idx) in enumerate(kfold.split(data)):

    print(f'FOLD {fold}')

    print('--------------------------------')

    class_sample_count = [74,859].
    weights = 1 / torch.Tensor(class_sample_count)
    weights = weights.double()
    train_sampler = torch.utils.data.sampler.WeightedRandomSampler(weights, 10)
    test_sampler = SubsetRandomSampler(test_idx)

    loaders = {
    'train': torch.utils.data.DataLoader(data, batch_size=16, sampler=train_sampler),
    'test': torch.utils.data.DataLoader(data, batch_size=16, sampler=test_sampler)
    }

    vgg16 = torchvision.models.vgg16(pretrained=True)
    for param in vgg16.features.parameters():
        param.requires_grad = False

    num_ftrs = vgg16.classifier[6].in_features
    vgg16.classifier[6] = nn.Linear(num_ftrs, 2)
    model = vgg16.to(device)
    criterion = nn.CrossEntropyLoss()
    optimizer = optim.Adam(model.parameters(), lr=0.001)
    train(3, loaders,   model, optimizer, criterion, train_on_gpu, 'model.pt')
    model.load_state_dict(torch.load('model.pt'))

Output:

Batch: 0	Epoch: 1 	Training Loss: 0.666256
Epoch: 1 	Training Loss: 0.666256 	Validation Loss: 11.949794 	Accuracy: 0.237942
Validation loss decreased (inf --> 11.949794).  Saving model ...
Batch: 0	Epoch: 2 	Training Loss: 0.000000
Epoch: 2 	Training Loss: 0.000000 	Validation Loss: 32.636715 	Accuracy: 0.237942
Batch: 0	Epoch: 3 	Training Loss: 0.000000
Epoch: 3 	Training Loss: 0.000000 	Validation Loss: 58.557861 	Accuracy: 0.237942

FOLD ACCURACY : 0.2379421221864952

Batch: 0	Epoch: 1 	Training Loss: 0.561094
Epoch: 1 	Training Loss: 0.561094 	Validation Loss: 22.119741 	Accuracy: 0.000000
Validation loss decreased (inf --> 22.119741).  Saving model ...
Batch: 0	Epoch: 2 	Training Loss: 0.000000
Epoch: 2 	Training Loss: 0.000000 	Validation Loss: 53.926662 	Accuracy: 0.000000
Batch: 0	Epoch: 3 	Training Loss: 0.000000
Epoch: 3 	Training Loss: 0.000000 	Validation Loss: 91.861961 	Accuracy: 0.000000

FOLD ACCURACY : 0.0

Is using WeightedRandomSampler and class weights in cross entropy toghether alright? Also should I use WeightedRandomSampler for testloader as well?