Skip to content

Loomy (luminamoon) Documentation Portal

LuminaV Optimizer v1.5.3 Data Engine v1.0.6 Rust 1.99.1 + Python ABI3T

Loomy(luminamoon) is the official reference implementation and central monorepo of PROJECT LOOMY, providing low-precision numerical optimization, multimodal dataset streaming, and formal verification suites developed by Silver Moon (@cloverxion) and Lumina Moon (@cloverx-id) contributors.

Getting Started

System requirements, pip install, Rust compilation, and quickstart examples for NumPy, PyTorch, and Data Engine.

LuminaV Optimizer

Dual-Buffer CLM vs Single-Buffer TD-CLM, mixed-precision, stochastic rounding, and FSDP distributed training.

Data Engine

Fused SIMD preprocessing, StringTensor zero-copy buffers, streaming datasets, and 128B Ping-Pong crash recovery.

API Reference

Complete signatures, parameter reference tables, validation constraints, and return types for all modules.

Security & Audits

Direct vulnerability disclosure policy to Silver Moon, CodeQL scanning, and third-party library safety audits.

Changelogs

Version-by-version release history for Software package distributions and LuminaV algorithm specifications.

1. User Documentation

Source directory: docs/en/

  • 01. Getting Started: System requirements, installation, source compilation via Maturin/Rust, hardware & SIMD ISA auditing (LuminaV.detect_hardware(), lm.data_cpu_isa()), and quickstarts for NumPy, PyTorch, and DataLoader (FusedPipeline, StringTensor, StateTracker, ResumablePipeline).
  • 02. User Guide: Optimizer buffer modes (Dual-Buffer CLM vs Single-Buffer TD-CLM), mixed-precision & stochastic rounding, boundary dynamics, FSDP distributed training, plus DataLoader \(O(1)\) local/HTTP-206 streaming, 100-batch Pre-RAM watermark ring buffer, Ping-Pong crash recovery, two-stage XXH3-64 deduplication, and multimodal HuggingFace Hub pipelines.
  • 03. Configuration & API Reference: Complete parameter reference tables, public methods, and validation rules for LuminaV, TextCleaner, QualityFilter, StreamDeduplicator, MinHashDeduplicator, PIIMasker, StateTracker, StreamCoordinate, StringTensor, FusedPipeline, and ResumablePipeline.
  • 04. Architecture & Hardware Acceleration: C4 dispatch hierarchy, Optimizer Rust crates (core, simd, vulkan, dispatch, triton-aot), Data Engine Rust modules (constants, coord, dedup, error, metadata, minhash, postprocess, py_bridge, quality, serialize, simd_cpu, string_tensor), zero-copy memory protocols, and free-threaded Python (abi3t / PEP 703).

2. Developer Modular Documentation

Deep technical documentation for systems engineers, kernel writers, and library developers (docs/optimize/ & data/docs/):

A. LuminaV Optimizer

Source directories: docs/optimize/ & optimizer/

B. LuminaMoon Data Engine

Source directories: data/ & data/docs/

3. Engineering & External Library Audits

Source directories: docs/library_audit/ & data/docs/library_audit/