reproducibility

[Henrique Weber]

Hi!

How should I proceed to make the example of the pose estimation reproducible? https://colab.research.google.com/github/BachiLi/redner/blob/master/tutorials/pose_estimation.ipynb

For now, every time I run it, I get slightly different results.
I added the seed argument on the render_albedo call and also set torch.manual_seed(0) at the beginning of the code, but still get different results at every run.

Thanks!