efc1128fab
All checks were successful
Smoke Test / smoke (pull_request) Successful in 8s
Adds automated verification script for issue #125: - tests/verify_bounds_checking_m4max.sh — validates bounds guards present and compiles shader on M4 Max - docs/TESTING_BOUNDS_CHECKING.md — manual verification procedure Also includes the bounds checking changes from step35/57 branch: - kernel_fwht_128: data_len parameter + base/d bounds guards - kernel_turbo4_dequant: src_len, norms_len, dst_len + per-buffer guards - kernel_attention_turbo4: full buffer length guards (q, k_packed, k_norms, scores) Closes #125 Co-authored-by: step35-cli <step35-cli@timmy.foundation>
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