Skillget

genpark-paged-attention-kv-cache-block-allocator-skill

mcp
Guvenlik Denetimi
Uyari
Health Uyari
  • 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 Gecti
  • Code scan — Scanned 4 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

GenPark AI Agent Skill - Simulates vLLM PagedAttention non-contiguous virtual memory block table allocation for dynamic KV-cache management.

README.md

GenPark AI Agent Skill - PagedAttention KV-Cache Block Allocator

Simulates PagedAttention non-contiguous virtual memory block table allocation for dynamic KV-cache management.

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

Architecture Diagram

graph TD
    A[Dynamic LLM Request Generation] --> B[PagedAttention Virtual Block Table]
    B --> C{Current Physical Block Full?}
    C -->|Yes| D[Pop New Block from Free Physical Pool]
    C -->|No| E[Write KV-Tensor to In-Block Slot]
    D --> E
    E --> F[Near-Zero Fragmentation VRAM Utilization]

Features

  • OS Paging Semantics: Maps logical token indices to physical block frames without requiring continuous memory.
  • Zero External Dependencies: Pure Python 3.9+ standard library.

Yorumlar (0)

Sonuc bulunamadi