Concated dataset to loader

[anthonysirico]

This is the result of this:

split_index = int(len(dataset)*split)
print(split_index)

ones = dataset.copy(idx=range(split_index))
zeros = dataset.copy(idx=range(split_index+1, len(dataset)))
print(ones[split_index-1].y)
print(zeros[0].y)

torch.manual_seed(12345) #set seed for reproducibility
ones.shuffle()
zeros.shuffle()

N = 10000
training_ones = ones.copy(idx=range(int(N*split)))
training_zeros = zeros.copy(idx=range(len(ones)))

training_dataset = torch.utils.data.ConcatDataset([training_ones, training_zeros])

from torch_geometric.data import DataLoader

train_loader = DataLoader(training_dataset, batch_size=64, shuffle=False)

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = gnn.GCN(hidden_channels=64, num_features=dataset.num_features, num_classes=dataset.num_classes).to(device)

optimizer = torch.optim.Adam(model.parameters(), lr=0.01)
criterion = torch.nn.CrossEntropyLoss()

def train():
    model.train()

    for data in train_loader:  # Iterate in batches over the training dataset.
        data = data.to(device)
        out = model(data.x, data.edge_index, data.batch)  # Perform a single forward pass.
        loss = criterion(out, data.y)  # Compute the loss.
        loss.backward()  # Derive gradients.
        optimizer.step()  # Update parameters based on gradients.
        optimizer.zero_grad()  # Clear gradients.

accuracy = []
epochs = 601
for epoch in tqdm(range(1, epochs), total=epochs):
    train()
    train_acc = test(train_loader)
    accuracy.append(train_acc)

Is it because the train_loader is a concated dataset?

[anthonysirico]

I figured it out, disregard.

[rusty1s]

Super, please share your findings such that others can benefit from it :)