-
-
Notifications
You must be signed in to change notification settings - Fork 925
Pull requests: TheTom/turboquant_plus
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
feat: implement offline channel calibration for outlier strategy
#94
opened Jun 3, 2026 by
andrea-gentilini
Loading…
Curated subset of #90 + K-side norm accounting fix
#91
opened May 9, 2026 by
TheTom
Owner
Loading…
2 tasks done
Accepted fixes & tests: V/K-norm size accounting, SeedSequence cleanup, QJL regression guard, rotation + ruff
#90
opened May 9, 2026 by
brosequist
Loading…
6 tasks
docs: ForgeAttention — fused 3-bit KV dequant inside Metal attention kernels
#76
opened Apr 10, 2026 by
user-23xyz
Loading…
TriAttention V3 hybrid recipe: two fixes for Qwen3.5 NIAH failure
#75
opened Apr 10, 2026 by
CG-8663
Loading…
fix: mock subprocess in test_missing_nvidia_smi_warning instead of relying on host hardware
#73
opened Apr 9, 2026 by
MLuqmanBR
Loading…
fix: correct KV cache memory stats for K/V metadata and fp16 baseline
#53
opened Mar 31, 2026 by
dipeshbabu
Loading…
feat: add MLX backend prototype for PolarQuant, TurboQuant, and KV cache compression
#52
opened Mar 31, 2026 by
dipeshbabu
Loading…
feat: TurboQuant enhancements — layer-adaptive, beta codebook, temporal decay
#37
opened Mar 26, 2026 by
MaTriXy
Loading…
1 of 3 tasks
RotorQuant: Clifford algebra vector quantization (88x fewer params)
#34
opened Mar 26, 2026 by
johndpope
Loading…
ProTip!
Find all pull requests that aren't related to any open issues with -linked:issue.