NanoJev
A small, readable end-to-end training pipeline for a decision-head model on Qwen3 0.6B, trained on four games including Doom.
0 members list NanoJev, unchanged over the last 12 weeks, data as of 7 November 2026
Members who list NanoJev in their Stack, by week.
Members who list NanoJev
Be the first to add NanoJev
Nobody on Stackness runs NanoJev yet. Add it to your stack and you're the first member running it.
About NanoJev
NanoJev is a training framework for decision-head models that implements an end-to-end pipeline. It supports parallel decision-making with dynamic candidate selection, enabling developers to train models on various environments including game scenarios. The framework is built on Qwen3 0.6B and demonstrates its capabilities through training on multiple games such as Doom. Designed for accessibility and clarity, NanoJev provides a compact, readable implementation suitable for developers exploring reinforcement learning approaches.
Developed by TianyuCodings, NanoJev is distributed under the MIT license. The project is implemented in Python, making it accessible to developers familiar with common machine learning workflows. As an open-source tool, it serves as both a functional training system and a reference implementation for understanding decision-making architectures in smaller language models.
Checked 2 November 2026.
Nobody on Stackness runs NanoJev yet.
Sign up and add itPosts about NanoJev
- How do you run System One decision models locally? Ollaya, laya-mlx and the runtime slot
To run a System One model locally, pick a runtime too: Ollaya serves ten open model families, laya-mlx runs Laya on Apple silicon, and llama.cpp needs a shim.
- Open-source alternatives to Jev, and what each one replaces
The open-source alternatives to Jev are Laya, Kev, SemIf, Open-Jev and Nimble. They are four kinds of project, and here is what each one replaces.