Hugging Face's logo

Datasets:
kalpesh77
/
vedic-neural-geometry

Tasks:
Feature Extraction
Graph Machine Learning
Text Retrieval
Modalities:
Text
Languages:
Marathi
Sanskrit
English
Size:
1K<n<10K
Tags:
vedic
knowledge-graph
graph-neural-networks
geometric-deep-learning
rag
multilingual
License:
Dataset card Data Studio Files
xet
vedic-neural-geometry / README.md
kalpesh77's picture
kalpesh77
Update README.md
7f8ef58 verified
preview code
|
Raw
History Blame
4.34 kB
metadata 
license: apache-2.0
language:
  - mr
  - sa
  - en
task_categories:
  - feature-extraction
  - graph-ml
  - text-retrieval
  - question-answering
pretty_name: Vedic Neural Geometry
size_categories:
  - 1K<n<10K
tags:
  - vedic
  - knowledge-graph
  - graph-neural-networks
  - geometric-deep-learning
  - rag
  - multilingual
  - sanskrit
  - marathi
  - hinduism
  - philosophy
configs:
  - config_name: default
    data_files:
      - split: train
        path: संग्रह/csv/**/*.csv

"""

🕉️ Vedic Neural Geometry

वैदिक ज्ञान आणि आधुनिक Neural Networks, Knowledge Graphs, Geometric Embeddings आणि Hybrid RAG यांचा संगम.

Dataset License Version

📊 Current Statistics (v1.4)

Component Value
Nodes {n_nodes:,}
Edges {n_edges:,}
Connected Components {n_comps} ✅
Core Chain 5/5 ✅
RAG Embeddings 384-dim multilingual
GNN Embeddings 128-dim (GCN)
Core Geometric Nodes 8
Geometric Matrices 3D/8D/16D/32D/64D (108×7×N)

🎯 Architecture

Core Geometric Chain 


Bindu_Brahma (0,0,0)
↓ MANIFESTS_AS
मल्टीवर्स_फील्ड
↓ CONTAINS_LAYER
Septa_Avaran (7 Layers)
↓ PROJECTS_ONTO
अंगुली_ग्रिड_१०८ (108 Divisions)
↓ PROJECTS_INTO
श्रीयंत्र (9 Triangles)

Feature Vector (1110 dims) 


x = [सत्त्व, रज, तम, geometric_features(1107)]

🚀 Quick Start

RAG Query 

from sentence_transformers import SentenceTransformer
import faiss, pandas as pd
from huggingface_hub import hf_hub_download

REPO = "kalpesh77/vedic-neural-geometry"
meta = pd.read_csv(hf_hub_download(REPO, "rag/embeddings/node_metadata.csv", repo_type="dataset"))
index = faiss.read_index(hf_hub_download(REPO, "rag/vector_store/nodes.faiss", repo_type="dataset"))

model = SentenceTransformer('intfloat/multilingual-e5-small')
q = model.encode(["शिव कोण आहे?"])
faiss.normalize_L2(q)
scores, idx = index.search(q.astype('float32'), 5)
print(meta.iloc[idx[0]][['display_name', 'प्रकार', 'text']])

🗂️ Structure

vedic-neural-geometry/
├── core/
│   ├── hierarchy.yaml          ← Single Source of Truth
│   ├── bindu_space.py
│   ├── septa_avaran.py
│   ├── multiverse_field.py
│   ├── kalachakra_clock.py
│   └── angular_grid.py
├── gnn/
│   ├── graphs/vedic_graph.gpickle
│   ├── features/{{pyg_data, simple_gcn, node_to_idx}}
│   └── matrices/mean_3d_108x7x3.npy ... mean_64d_108x7x64.npy
├── rag/
│   ├── embeddings/node_embeddings.npy, node_metadata.csv
│   ├── vector_store/nodes.faiss
│   └── aliases.json
└── scripts/validate_geometry_hierarchy.py

📈 Changelog

v1.4 (Current) — Full Connectivity

· ✅ 417 components → 1 (fully connected) · ✅ 416 isolated + 30 strategic + 43 final bridges · ✅ Core chain 100% complete · ✅ GNN retrained (80 epochs, loss 0.98) · ✅ {n_edges:,} edges

v1.3 — Node Merging

· 23 duplicate nodes merged · Canonical attributes restored

v1.2 — Core Geometric Nodes

· 9 Core nodes added + Manual overrides

v1.1 — RAG Improvements

· Query rewriting (aliases) + Core boosting (2x)

🔑 Core Nodes

  1. Bindu_Brahma — केंद्रबिंदू (0,0,0)
  2. मल्टीवर्स_फील्ड — Multiverse Field
  3. Septa_Avaran — 7 Layers
  4. अंगुली_ग्रिड_१०८ — 108 Grid
  5. श्रीयंत्र — 9 Triangles
  6. ॐ — प्रणव नाद
  7. कालचक्र — काल जिओमेट्री
  8. हिरण्यगर्भ — ब्रह्मांड बीज

📜 License

Apache-2.0

📖 Citation

@dataset{{vedic_neural_geometry_2024,
  author = {{Kalpesh}},
  title = {{Vedic Neural Geometry}},
  year = {{2024}},
  url = {{https://huggingface.co/datasets/kalpesh77/vedic-neural-geometry}}
}}