TypeError: cannot unpack non-iterable float object

vision
1

I have recently started on my machine learning and deep learning journey and I was following this DataQuest tutorial. Everything was going fine. Each cell was running just the way it should. Until i got to the training and even then the model trained for 3.5 hours then i got this error when it finished.

TypeError: cannot unpack non-iterable float object

This is the block of code that would gave me the error

for epoch in range(EPOCHS):
for batch, (images, labels, img_paths) in enumerate(train):
optimizer.zero_grad()

    images = images.to(device)
    pred = model(images.float())
    labels = labels.to(device)
    loss = loss_fn(pred, labels)

    loss.backward()
    optimizer.step()

loss, current = loss.item()
print(f"loss: {loss:>7f} [{epoch}]")

This is the entire Code on GitHub.

I would usually play around with the code a bit and see what i can do but 3.5 hours is a long time to wait to see if what i tried worked.

Any help would be appreciated.

Thank you

2

Hi, the issue lies in the line:
loss, current = loss.item()

loss.item() gives you a single float value. And it cannot be packed to two values, loss, current. The correct way would be:
loss = loss.item()

If you want a more in depth tutorial please refer to this example over here.

If you want to read more about loss.item() please refer this.

Regarding, it taking 3.5 hours. I thinking you are not using GPUs in google colab. Computations would be much faster in GPU.

1 Like
3

Just checked the video you linked as well, the author uses this as well over here

4

Thank you for your help and for the references for loss.item(). It is extremely appreciated. I used googles colab to code in because my computer isnt the greatest its an old hp prodesk 6200 and definitely doesnt have a nvidia gpu. I changed the runtime to gpu i just figured it took so long because of the size of the data set and the amount of epochs. In the video i was following he said it took him til the next day to finish training

1 Like