nnll

nnll (or null) is a comprehensive AI toolkit for managing and processing Diffusion and Large Language Models (LLMs). The project is divided into highly modular, ready-to-use components, and may appeal to researchers or developers working individual experiments or in the general field of large-scale machine learning model deployment.

We currently support 🧨Diffusers, 🤗Transformers, 🦙Llama inference, and refined methods for tasks such as extracting and classifying metadata, pipeline preparation, GPU configuration, consumer-grade system optimization, and a variety of generative AI preparations.

use

nnll-parse - Lookup model metadata headers and write out to .json file.
nnll-find - Scan .json files from -parse for a specific string pattern within layer contents.

nnll-toc - (run from root folder only) Recreate the project table of contents, populating preview and navigation links for IDEs

imports

import nnll_**.src

modules table of contents

Note

x/dev branch: updated ~=nightly.
main branch : stable, devoted to long-term support.

development setup

clone repo

git clone https://github.com/darkshapes/nnll.git

Next-->

create virtual environment

py -m venv .venv_nnll

Next-->

3 activate --> (windows powershell)

Set-ExecutionPolicy Bypass -Scope Process -Force; .venv_nnll\Scripts\Activate.ps1

3 activate --> ( linux | macos)

.venv_nnll/bin/activate

Next-->

4 install

cd nnll
pip install -e .

Done.