Move ndarray library to its own repository - #301
Conversation
Sorry, something went wrong.
|
Sounds alright and looks fine, but did you want me to look at something in particular? Just one thing, we probably want to set the version to 0.4.0-SNAPSHOT instead of 0.3.1. |
Sorry, something went wrong.
Nothing in particular, was just seeking for a general feedback like this one, thanks. The version is 0.4.0-SNAPSHOT btw, it’s just the example in the README that shows the usage of the last released version. |
Sorry, something went wrong.
|
Since we're probably going to end up with more modules, maybe we could rename the "ndarray" one to "ndarray-core" or something? |
Sorry, something went wrong.
|
I think I still prefer the simplicity of org.tensorflow:ndarray for importing the main artifact of that library, then append a -something to any of its upcoming extensions. Ideally we would have done the same with tensorflow but org.tensorflow:tensorflow was already being used by TF 1.x |
Sorry, something went wrong.
|
Ok I’ll take this as a go, thanks! |
Sorry, something went wrong.
The Java NdArray library now have its own repository, to emphasis its independence from TensorFlow itself and allow it to have its own development/release cycle.
As part of the review of this PR, please take also a look at the new repository : https://github.com/tensorflow/java-ndarray and let me know if you see anything that should be changed there as well.
A note on this, I've moved the library as a sub-module since we might end up having other modules there as well, e.g. ndarray-kotlin. Also, I've picked tensorflow-java-ndarray as the name of the parent module so it matches the name of the repo (like tensorflow-java matches this one) but I guess this is also debatable.