Model Writing Issue with txt Format.

[Gaaaavin]

When running preprocess/auto_reorient.py with --model_type txt flag, the saved txt file will contain tensors.

After some investigation, I find out this is due to this line in the file.

In this line, the pt.xyz is a torch tensor due to previous pytorch operations. Directly applying the * operator would result in three individul tensors, which won't be written properly as numbers to the text file.

This can be simply solved by first converting the tensor into an numpy array, i.e. *pt.xyz.numpy().

I have opened a pull request #11 for this issue.