I was trying to understand the significance of this. There is a blog post here[0]. (I had a brief glance, please correct me if I'm wrong) It seems to be a framework allowing neural networks (and layers of neural networks) to be composed together more easily.

It's not completely clear to me what the advantages are without digging in, I assume to simplify/standardise the interface between the layers?

[0] https://devblogs.nvidia.com/neural-modules-for-speech-langua...

it has “neural type” system which check what can be connected to what. Also, comes with several modules for speech recognition and NLP

> A “Neural Module” is a block of code that computes a set of outputs from a set of inputs.

I struggle to find any clarity in this description. You can omit most of it:

> [...] code [...] computes [...] output [...] from [...] input [...]

Which is true, and trivial. This important points for understanding (what kind of input? what kind of output? why sets? what's 'Neural' about functions?) is all missing.

Further in the text:

> A Neural Module’s inputs/outputs have a Neural Type, that describes the semantics, the axis order, and the dimensions of the input/output tensor. This typing allows Neural Modules to be safely chained together to build applications, as in the ASR example below.

Sounds like a typed Pytorch module then?

> inputs/outputs have a Neural Type, that describes the semantics, the axis order, and the dimensions of the input/output tensor

Does that really answer any of my questions? What's 'Neural' about the type?

Sounds kind of like what's described in Neural Networks, Types, and Functional Programming: http://colah.github.io/posts/2015-09-NN-Types-FP/

Is that mentioned in the readme? Is this readme worth reading if you don't already know what the project is about?