Luigi Antelmi

https://github.com/bhpfelix/Variational-Autoencoder-PyTorch/blob/135ad8cfff8f5d01a5bfd44fbfc53066150fa8e5/src/vanila_vae.py#L125 Is there a particular reason why the method ".exp_()" has been preferred to ".exp()" ?

https://github.com/LaRiffle/collateral-learning/blob/a8e40193e234e331fe49a5b0e1207b34464efa16/tutorials/Part%2001%20-%20Building%20a%202%20target%20features%20dataset.ipynb#L911 this function seems to be the bottolneck. if you use the fig2data from http://www.icare.univ-lille1.fr/tutorials/convert_a_matplotlib_figure the dataset generation should be faster

requires_grad=True for an input datapoint?

1

https://github.com/pytorch/tutorials/blob/f4ebb4d007792f5bc302affa7b360a9710e4a88b/advanced_source/super_resolution_with_onnxruntime.py#L144 It is obscure to me why there is the need to set the flag requires_grad to True for datapoint "x", which has no parameters to be learnt. Is it...

MatrixNormalDistribution is missing the CumulativeDistribution method. The MatrixNormalDistribution is a generalization of the MultiVariate Normal to the matrix case. Considering also that a MultivariateNormal distribution is also missing, there is...