Skillget

genpark-weight-quantization-kquants-block-packer-skill

mcp
Security Audit
Warn
Health Warn
  • No license — Repository has no license file
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 8 GitHub stars
Code Pass
  • Code scan — Scanned 4 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

GenPark AI Agent Skill - Simulates k-quants block compression (Q4_K_M, Q8_0) packing FP16 weights into quantized blocks with scale factors.

README.md

GenPark AI Agent Skill - Weight Quantization K-Quants Block Packer

Quantizes full-precision neural network weights into compact, cache-friendly block representations (Q8_0, Q4_0) with dynamic scale factors.

Verified by GenPark AI and compatible with Model Context Protocol (MCP).

Architecture Diagram

graph TD
    A[Raw FP16 / BF16 Weight Tensor] --> B[Block Slicer: 32 Elements per Block]
    B --> C[Compute Dynamic Block Scale Factor: max abs / 127]
    C --> D[Quantize FP16 to Signed 8-bit Int8]
    D --> E[Pack Scale + Quantized Values into Dense Memory Buffer]
    E --> F[50%+ Memory Reduction for Edge Inference]

Features

  • Deterministic Min-Max Scaling: Preserves activation fidelity while halving memory footprint.
  • Zero External Dependencies: Pure Python 3.9+ standard library.

Reviews (0)

No results found