File size: 3,989 Bytes
e9e9f83 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 | XERO — VOVINA ZEDEC PRO
=======================
Copyright (c) 2026 Michael Laurence Curzi — ZEDEC AI / 36N9 Genetics LLC.
The XERO framework code in this repository is licensed under the MIT License
(Attribution to Michael Laurence Curzi Required). See LICENSE.
This product includes, references, or interoperates with third-party works.
Those works remain under their own licenses and are attributed here. None of the
third-party *code* below is bundled in this package — it is downloaded on demand
by the setup wizard (see setup_wizard.py / requirements*.txt) so that this
package carries no third-party copyrighted code.
--------------------------------------------------------------------------------
1. OUTER-CORE MODEL (downloaded on demand; NOT included)
--------------------------------------------------------------------------------
* Qwen2.5-3B-Instruct
- Copyright (c) Alibaba Cloud.
- License: Apache License 2.0.
- Role: XERO's "outer core" LLM. XERO wraps and fine-tunes it (LoRA); the base
weights are not re-licensed by this project. Apache-2.0 terms and the
upstream NOTICE apply to those weights.
- Source: https://huggingface.co/Qwen/Qwen2.5-3B-Instruct
--------------------------------------------------------------------------------
2. THIRD-PARTY PYTHON LIBRARIES (downloaded on demand; NOT included)
--------------------------------------------------------------------------------
* PyTorch (torch) — Copyright (c) Meta Platforms, Inc. — BSD-3-Clause
* Transformers — Copyright (c) Hugging Face, Inc. — Apache-2.0
* PEFT (LoRA) — Copyright (c) Hugging Face, Inc. — Apache-2.0
* Datasets — Copyright (c) Hugging Face, Inc. — Apache-2.0
* Accelerate — Copyright (c) Hugging Face, Inc. — Apache-2.0
* NumPy — Copyright (c) NumPy Developers — BSD-3-Clause
Full license texts are distributed with each package when installed via pip;
see requirements.txt and training/requirements-train.txt.
--------------------------------------------------------------------------------
3. TRAINING DATA — PROVENANCE & LICENSES
--------------------------------------------------------------------------------
* The "VOVINA 500" books (data/vovina_books.jsonl)
- Copyright (c) Michael Laurence Curzi. The author's OWN works, included by the
author for training his own model. Not third-party content.
* Harvested free-to-use corpus (data/fresh_corpus.jsonl, learned_knowledge.jsonl)
- Assembled by the polymath harvester from public, free-to-use sources, each
retaining its own license/terms. Principal sources include:
- Project Gutenberg — public domain (per-work; US).
- Wikipedia / Wikimedia — CC BY-SA 4.0.
- Open Library / Internet Archive metadata — open.
- arXiv — per-paper licenses / open metadata.
- NCBI PubMed, GenBank, Europe PMC — public-domain / open-access records.
- This corpus is provided for research/training. Downstream users are
responsible for honoring each source's terms for any redistribution.
--------------------------------------------------------------------------------
4. CONCEPTUAL INSPIRATION (metaphor only; NO third-party code or text included)
--------------------------------------------------------------------------------
XERO's architecture borrows *names and principles* (time crystal, Casimir/zero-
point, holographic projection) as engineering metaphors implemented in original
mathematics. No proprietary documents are bundled. The harmonic-chemistry law is
the author's own published work ("Making Chemistry with Sound", Curzi 2022).
--------------------------------------------------------------------------------
Attribution requirement: any derivative work, publication, or deployment of the
XERO framework must credit "Michael Laurence Curzi — ZEDEC AI / 36N9 Genetics
LLC". See LICENSE for full terms.
|