Back to search
PyTorch implementation of TurboQuant (Google Research, 2026) for KV cache compression - 3.3× compression with only +6.1% perplexity degradation on GPT-2 Medium.
Stars
0
Forks
0
Watchers
0
Open Issues
0
Overall repository health assessment
No package.json found
This might not be a Node.js project
No contributors data available
1a21a44View on GitHub8260372View on GitHubTurboQuant KV cache demo — GPT-2 Medium, 4-bit Lloyd-Max MSE + QJL
4c53e80View on GitHub