Upload expand_knowledge.py
Browse files- expand_knowledge.py +43 -0
expand_knowledge.py
ADDED
|
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# ======================================================================
|
| 2 |
+
# AXIS: Knowledge Expansion Tool (V1.2)
|
| 3 |
+
# This script converts raw text into AXIS-compatible Semantic Lattices.
|
| 4 |
+
# ======================================================================
|
| 5 |
+
|
| 6 |
+
import json
|
| 7 |
+
import torch
|
| 8 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
| 9 |
+
|
| 10 |
+
MODEL_ID = "kofdai/AXIS-Sovereign-Logic-Engine"
|
| 11 |
+
|
| 12 |
+
def extract_to_lattice(text):
|
| 13 |
+
print(f"🧐 [AXIS] 知識抽出中: {text[:30]}...")
|
| 14 |
+
|
| 15 |
+
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
|
| 16 |
+
model = AutoModelForCausalLM.from_pretrained(MODEL_ID, torch_dtype=torch.bfloat16, device_map="auto")
|
| 17 |
+
|
| 18 |
+
prompt = f"以下のテキストをAXIS立体十字形式(JSON)に変換せよ。論理矛盾があればconflictsに記載せよ。\n入力: {text}\nFormat: {{'nodes':[], 'edges':[], 'conflicts':[]}}"
|
| 19 |
+
|
| 20 |
+
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
|
| 21 |
+
with torch.no_grad():
|
| 22 |
+
outputs = model.generate(**inputs, max_new_tokens=512)
|
| 23 |
+
|
| 24 |
+
result = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 25 |
+
return result
|
| 26 |
+
|
| 27 |
+
if __name__ == "__main__":
|
| 28 |
+
# 拡張したい知識の例
|
| 29 |
+
raw_knowledge = [
|
| 30 |
+
"意味は計算の結果ではなく、計算が走るための初期構造である。",
|
| 31 |
+
"AXISの物理パージは、知能の純粋性を保つための儀式である。"
|
| 32 |
+
]
|
| 33 |
+
|
| 34 |
+
knowledge_base = []
|
| 35 |
+
for k in raw_knowledge:
|
| 36 |
+
lattice_piece = extract_to_lattice(k)
|
| 37 |
+
knowledge_base.append(lattice_piece)
|
| 38 |
+
|
| 39 |
+
# local_massive_data.json として保存
|
| 40 |
+
with open("local_massive_data.json", "w", encoding="utf-8") as f:
|
| 41 |
+
json.dump(knowledge_base, f, ensure_ascii=False, indent=4)
|
| 42 |
+
|
| 43 |
+
print("✅ [SUCCESS] 知識拡張完了。local_massive_data.json を生成しました。")
|