jakeveo05/tcm-divination-training

TCM & Divination Training Dataset v2

Comprehensive training dataset for Bazi, Tử Vi (Zi Wei Dou Shu), TCM, and divination domains.

Dataset Summary

Metric	Value
Total Unique Samples	162,384
File Size	651 MB
Languages	Vietnamese, English, Chinese
Last Updated	2026-01-11

Data Sources

Source	Unique Samples	Description
bazi_books	74,533	Extracted from Bazi/Tử Vi books (OCR)
gpt_training_ready	48,551	GPT-generated Q&A pairs
divination_combined	20,763	Horoscope, Tarot, Occult
chinese_knowledge	11,780	TCM, Kinh Dịch, Đông Y
ziwei_qa	6,557	Chinese fortune telling Q&A
sft_fortune	120	ZWDS interpretations
final_alpaca	80	Bazi specific

Files

merged_all_training.json (Primary)

• Format: Alpaca (instruction, input, output, source)
• Samples: 162,384
• Use: Instruction fine-tuning (SFT)

chinese_knowledge_pretrain.jsonl

• Format: JSONL with text chunks
• Use: Continued pre-training

chinese_knowledge_qa.json

• Format: Alpaca Q&A pairs
• Use: Knowledge Q&A

Domains Covered

• Bazi (八字) - Chinese astrology / Four Pillars
• Zi Wei Dou Shu (紫微斗数) - Purple Star Astrology
• TCM (中医) - Traditional Chinese Medicine
• Kinh Dịch (易经) - I Ching / Book of Changes
• Phong Thủy (风水) - Feng Shui
• Horoscope - Western astrology
• Tarot - Card readings
• Đông Y Việt Nam - Vietnamese traditional medicine

Usage

from datasets import load_dataset

# Load full dataset
ds = load_dataset("jakeveo05/tcm-divination-training", data_files="data/merged_all_training.json")

# Filter by source
bazi_data = [s for s in ds['train'] if s['source'] == 'bazi_books']
ziwei_data = [s for s in ds['train'] if s['source'] == 'ziwei_qa']

Training Recommendation

• Model: Qwen2.5-3B or larger
• Method: QLoRA 4-bit
• Steps: 1000-2000 (adjust based on loss)
• Learning Rate: 2e-4

License

CC-BY-4.0 - Educational and research use.

Changelog

• v2 (2025-01-11): Added Ziwei Q&A, Bazi books, GPT training data - 162K samples
• v1 (2025-01-11): Initial release - 53K samples