baded69890
All checks were successful
Smoke Test / smoke (pull_request) Successful in 14s
- scripts/weekly_update.py: Auto-generates weekly update from git log, Gitea API (issues/PRs/blockers), and benchmark results. Supports --post to issue #76, --json for raw data, --since for date range. - scripts/weekly_update.sh: Shell wrapper for convenience. - docs/WEEKLY_TEMPLATE.md: Manual update template. - docs/PROJECT_STATUS.md: Added Weekly Progress Updates section with process (weekly cadence, benchmark-as-happens, blocker escalation). - tests/test_weekly_update.py: Validates script runs, JSON output, and handles edge cases.
TurboQuant
KV cache compression for local inference on M4 Max MacBook Pro.
What
TurboQuant (Google, ICLR 2026) is a three-stage KV cache compression method:
- PolarQuant — WHT rotation + polar coordinates + Lloyd-Max codebook (~4.2x compression)
- QJL — 1-bit quantized Johnson-Lindenstrauss residual correction
- TurboQuant — PolarQuant + QJL = ~3.5 bits/channel, zero accuracy loss
Why
Unlock 64K-128K context on qwen3.5:27b within 32GB unified memory. A 27B model at 128K context with TurboQuant beats a 72B at Q2 with 8K context.
Status
See issues for current progress.
Roles
- Strago: Build spec author
- Cid: Implementation, benchmarks, deployment
- Locke: Research support, upstream watch
- John: Quality review
- Frankie: Coordination
Source Repos
- TheTom/llama-cpp-turboquant — llama.cpp fork with Metal
- TheTom/turboquant_plus — Reference impl, 511+ tests
- amirzandieh/QJL — Author QJL code (CUDA)
- rachittshah/mlx-turboquant — MLX fallback
Docs
- Project Status — Full project status and build specification