19 lines
935 B
Markdown
19 lines
935 B
Markdown
|
---
|
||
|
|
title: "nimphs"
|
||
|
|
weight: 20
|
||
|
|
---
|
||
|
|
|
||
|
|
# nimphs
|
||
|
|
---
|
||
|
|
the [nimphs](#) are a family of families of ml algorithms written in [nim](https://nim-lang.org), a low level systems language, with syntax similar to [python](https://python.org).
|
||
|
|
|
||
|
|
all of them use the [arraymancer](https://mratsim.github.io/Arraymancer/index.html) library, and all of them are in extreme pre-alpha, except for melite
|
||
|
|
|
||
|
|
## nyrids
|
||
|
|
[nyrids](#) are nlp models:
|
||
|
|
- [melite](#) is an implementation of a bigram model trained on public domain texts.
|
||
|
|
- [nimertes](#) is about translating the [transformer](https://arxiv.org/abs/1706.03762) architecture to [nim](https://nim-lang.org).
|
||
|
|
|
||
|
|
## oreads
|
||
|
|
[oreads](#) are reinforcement learning models and architectures. rather than being split like the [nyrids](#nyrids), they instead are an implementation library, akin to [stablebaselines](https://stable-baselines.readthedocs.io/en/master/) or [cleanrl](https://docs.cleanrl.dev/).
|