22.02.2024 Views

Daniel Voigt Godoy - Deep Learning with PyTorch Step-by-Step A Beginner’s Guide-leanpub

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

soon, so please bear with me for now.

Notebook Cell 2.7

%run -i model_training/v5.py

What does the model look like after training another 200 epochs?

print(model.state_dict())

Output

OrderedDict([('0.weight', tensor([[1.9448]], device='cuda:0')),

('0.bias', tensor([1.0295], device='cuda:0'))])

Well, it didn’t change at all, which is no surprise: The original model had converged

already; that is, the loss was at a minimum. These extra epochs served an

educational purpose only; they did not improve the model. But, since we are at it,

let’s check the evolution of the losses, before and after checkpointing:

Figure 2.8 - Losses, before and after resuming training

Clearly, the loss was already at a minimum before the checkpoint, so nothing has

changed!

It turns out, the model we saved to disk was a fully trained model, so we can use it

for…

Saving and Loading Models | 167

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!