A .pt file is a file extension commonly used in PyTorch to save models, tensors, or other data. In PyTorch, .pt and .pth are interchangeable extensions used to denote saved models or tensors.
Applications/Use Cases:
- Saving Models: After training a model, you can save its state_dict (the model’s parameters) to a
.ptfile for later use. - Saving Tensors: Individual tensors can be saved to
.ptfiles for later processing or analysis. - Saving Checkpoints: During training, you can save checkpoints to
.ptfiles to resume training from a specific point.
Saving A Model:
import torch
# Assuming 'geek' is your PyTorch model
torch.save(model.state_dict(), 'geek.pt')
Loading A Model:
import torch
# Assuming 'geek' is your PyTorch model
model.load_state_dict(torch.load('geek.pt'))
model.eval() # Set the model to evaluation mode