kvcache-ai/ktransformers

Skip(New discovery — enrichment pending)

Why this repo matters

Enables efficient LLM inference and fine-tuning on consumer hardware via CPU-GPU heterogeneous computing optimizations.
Provides day-zero support for latest models like MiniMax-M3, GLM-5.2, and DeepSeek-V4-Flash.
Integrates unified workflows for RL-DPO fine-tuning and expert scheduling using standard tools like LLaMA-Factory.

Key Metrics

Stars: 19.5k
Forks: 1.6k
Open Issues: 511
Stars (7d delta): 0
Stars (30d delta): 68

Scores & Metadata

Trend Velocity Score: 0.00
Opportunity Score: 25.00
Confidence Score: 100.00
Language: Python
First Seen: 2 months ago

Star History

19.5k stars+362 over 13 snapshots8/2/2026 – 9/13/2026

Links & Sources

Related Repositories

codecrafters-io/build-your-own-x

547.5kMarkdown

freeCodeCamp/freeCodeCamp

freeCodeCamp.org's open-source codebase and curriculum. Learn math, programming, and computer science for free.

456.9kTypeScript

affaan-m/ECC

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

269kJavaScript

multica-ai/andrej-karpathy-skills

A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.

217.5k