Technical documentationDaoTi V53

If it runs, it is engineering

Every command and link below points at a file that exists in the repository. You can reproduce all of it.

Chapter I

Quick start

git clone https://github.com/zhibaiYingChuan/DaoTi.git
cd DaoTi
pip install -r requirements.txt

# Interactive demo (recommended)
python app/app.py

# Or a command-line demo
python scripts/demo_real_text.py

Do not want to install anything locally? Colab Notebook runs the full inference in about 5 minutes on a free CPU.

Chapter III

Model facts

5,059,040
Base parameters
8145
Vocabulary size
256
Sequence length
CPU
Training and inference

The base emits structured semantic representations: an encoding-space vector, a Luoshu state vector, and a 64-dimensional prototype vector (hexagram space). Three reasoning methods — Zhouyi, Meihua Yishu and Liuyao — can be switched at inference time.

Chapter IV

Boundaries and license

  • “Gauge field”, “degenerate ground state” and “Goldstone mode” are engineering analogies used to describe observed optimisation behaviour, not physical conclusions.
  • Code is distributed under Apache 2.0. Model weights are under the DaoTi Research License v1.0, which forbids reverse engineering and redistribution.
  • Full training code, data recipes and detailed architecture design are not public and require separate authorisation.
  • The papers are working documents by an independent researcher and have not been peer reviewed.