Julia programming language

[anhinga_anhinga]

Julia is an unusual language. It is based around the idea of "eating your cake and having it too, again and again". Flexible and very fast at the same time, friendly readable syntax and Lisp-strength macros and multiple dispatch, etc

Comments

[russhatter]

Hi, thanks for these references. I have just found needs to look through them.

Julia is not good for me:
a) it is either good or excelent for advanced users of MathCAD - I am not, I am a software developer, not an engineer
b) there are plenty of work in data preparation before applying neural calculation: tokenization and another ... "goods" of this nature. I use not a standard data and formulations, so I need to make all this stuff by my hands without helpers, and it is good (really, not so ugly) if I can do it in Python

Trax: tried to install it - it is too heavy. It is collection of all NLP models (including torch! it is not TF at all!). It is heavy and possibly good, but I do need all these models! What I need is something lightweight and good to solve "easy" tasks, more "primitive" than NLP. If there are more than ten persons who are working on automatical translation - why do I need to join them? I would prefer to do something unique... I hate olimpiades...

JAX: looks VERY good for my purposes. Thanks!

[anhinga_anhinga]

That's cool! I am still torn between these two ecosystems, and between Julia Flux and JAX in particular

[anhinga_anhinga]

Нашёлся новый весьма информативный обзор про современное состояние JAX и про то, как его используют в DeepMind:

https://deepmind.com/blog/article/using-jax-to-accelerate-our-research

[anhinga_anhinga]

So, what seems to be interesting is that they don't include tensor-oriented TRAX ( https://github.com/google/trax ) in their list of useful tools, but emphasize pytree protocol:

https://jax.readthedocs.io/en/latest/pytrees.html

https://jax.readthedocs.io/en/latest/jax.tree_util.html

***

Let me include a link to this educational post for people who want to migrate from PyTorch to JAX:

https://sjmielke.com/jax-purify.htm

"Luckily, JAX is not just comfortable differentiating with respect to scalars, vectors, and matrices, but also with respect to a number of tree-like data structures that it calls pytrees-and they include python dicts."