mirror of
https://github.com/onestardao/WFGY.git
synced 2026-04-30 04:29:57 +00:00
113 lines
4.1 KiB
Python
113 lines
4.1 KiB
Python
"""
|
||
╭──────────────────────────────────────────────────────────╮
|
||
│ WFGY SDK · Self-Healing Variance Gate for Any LLM │
|
||
│----------------------------------------------------------│
|
||
│ 💌 Contact : hello@onestardao.com / TG @PSBigBig │
|
||
│ 🌐 Docs : https://onestardao.com/papers │
|
||
│ 🐙 GitHub : https://github.com/onestardao/WFGY │
|
||
│ │
|
||
│ ★ Star WFGY 1.0 → Unlock 2.0 │
|
||
│ 10k ⭐ by **Aug 1st** = next-gen AI alchemy │
|
||
│ Your click = our quantum leap │
|
||
│ │
|
||
│ 🔍 Official PDF of WFGY 1.0 (Zenodo DOI): │
|
||
│ https://doi.org/10.5281/zenodo.15630969 │
|
||
│ (Hosted on Zenodo – trusted international archive) │
|
||
│ │
|
||
│ 🧬 WFGY BigBang Prompt Pack (v1.0): │
|
||
│ https://doi.org/10.5281/zenodo.15657016 │
|
||
│ (Prompts to trigger the gate; multilingual updates coming) │
|
||
│ │
|
||
│ 🧠 Hidden folder inside repo: /I_am_not_lizardman │
|
||
│ (X secret papers, wild prompts, and Einstein drama) │
|
||
│ │
|
||
│ ⚠ GPT-2 demo is just the appetizer. With bigger LLMs, │
|
||
│ WFGY activates variance-drop lasers and KL fireworks. │
|
||
│ │
|
||
│ 🎮 Bonus: Honest Hero RPG Channel → │
|
||
│ https://www.youtube.com/@OneStarDao │
|
||
╰──────────────────────────────────────────────────────────╯
|
||
"""
|
||
|
||
# bbmc.py
|
||
# Semantic Residue (BBMC) implementation
|
||
# License: MIT
|
||
|
||
from __future__ import annotations
|
||
import logging
|
||
from typing import Dict
|
||
import numpy as np
|
||
|
||
logger = logging.getLogger(__name__)
|
||
|
||
|
||
def _safe_normalise(vec: np.ndarray) -> np.ndarray:
|
||
"""Return a unit L2-normalised copy; if norm is 0 return original."""
|
||
norm = np.linalg.norm(vec)
|
||
return vec if norm == 0.0 else vec / norm
|
||
|
||
|
||
def compute_residue(
|
||
input_vec: np.ndarray,
|
||
ground_vec: np.ndarray,
|
||
m: float = 0.1,
|
||
c: float = 0.5,
|
||
*,
|
||
normalise: bool = True,
|
||
return_vector: bool = True
|
||
) -> Dict[str, np.ndarray | float]:
|
||
"""
|
||
Compute semantic residue B = I - G + m * c^2.
|
||
|
||
Parameters
|
||
----------
|
||
input_vec : np.ndarray
|
||
Input semantic vector I.
|
||
ground_vec : np.ndarray
|
||
Ground-truth semantic vector G.
|
||
m : float, optional
|
||
Matching coefficient.
|
||
c : float, optional
|
||
Context factor.
|
||
normalise : bool, optional
|
||
If True, I and G are L2-normalised before subtraction.
|
||
return_vector : bool, optional
|
||
If True, include full B_vec in the result.
|
||
|
||
Returns
|
||
-------
|
||
dict
|
||
{
|
||
"B_vec": np.ndarray,
|
||
"B_norm": float
|
||
}
|
||
"""
|
||
if input_vec.shape != ground_vec.shape:
|
||
raise ValueError("input_vec and ground_vec must share the same shape")
|
||
|
||
if normalise:
|
||
input_vec = _safe_normalise(input_vec)
|
||
ground_vec = _safe_normalise(ground_vec)
|
||
|
||
B_vec = input_vec - ground_vec + m * (c ** 2)
|
||
B_norm = float(np.linalg.norm(B_vec, ord=2))
|
||
|
||
out = {"B_norm": B_norm}
|
||
if return_vector:
|
||
out["B_vec"] = B_vec
|
||
|
||
logger.debug("BBMC ‖B‖ = %.6f", B_norm)
|
||
return out
|
||
|
||
|
||
# ------------------------- quick demo ------------------------- #
|
||
def run_demo() -> None:
|
||
rng = np.random.default_rng(42)
|
||
I = rng.normal(size=8)
|
||
G = I + rng.normal(scale=0.05, size=8) # small noise
|
||
res = compute_residue(I, G)
|
||
print(f"BBMC demo ‖B‖ = {res['B_norm']:.4f}")
|
||
|
||
|
||
if __name__ == "__main__":
|
||
run_demo()
|