PyBlaze 2.3.2

Update requirements.txt

PyBlaze 2.3.1

• Bump sklearn dependency version constraint

PyBlaze 2.3.0

• Improve the Implementation of xnn.MaskedAutoregressiveTransform1d
• Add xnn.FlipTransform1d

PyBlaze 2.2.6

• Setting gpu=True now chooses the first GPU instead of the ones with the highest amount of free memory

PyBlaze 2.2.5

• Allow TransformedGmmLoss to learn the GMM means
• Remove useless generate_random_gmm function

PyBlaze 2.2.4

• Replace deprecated torch.cuda.memory_cached by torch.cuda.memory_allocated when passing gpu=True to multiple functions

PyBlaze 2.2.3

New Modules

• LeakyReLUTransform
• PReLUTransform

Release 2.2.2

• Bugfix Release

Enable Density Estimation via GMMs

2.2.1

Add docs

Simplify Engine Design

• It is now possible to override the to_device method used to transfer model inputs and outputs between devices (fixes #4)
• The parallel implementation of the predict method has been removed as it introduced needless challenges