Downloading trained Torch7 models (.t7) to PyTorch

Question

I use the Torch7 library to implement neural networks. Basically, I rely on pre-prepared models. In Lua, I use the torch.load function to load a model saved as a torch.t7 file. I am interested in switching to PyTorch ( http://pytorch.org ) and I read the docs. I could not find any information on the loading mechanisms of the pre-prepared model. The only relevant information I could find is this page: http://pytorch.org/docs/torch.html

But the torch.load function described on the page loads the file saved with pickle. If anyone has more information about loading .t7 models into PyTorch, please share it here.

Answer 1

The correct function load_lua:

from torch.utils.serialization import load_lua

x = load_lua('x.t7')