-
Notifications
You must be signed in to change notification settings - Fork 1.5k
Pull requests: JustVugg/colibri
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
convert(fp8->int4): --indir copies the full metadata set and resumes interrupted passes (#383)
#404
opened Jul 19, 2026 by
anrasi
Loading…
5 of 6 tasks
KV cache quantization: fp8 (KV8) + 4-bit TurboQuant (KV_TQ) on CPU, CUDA, and Metal
#399
opened Jul 18, 2026 by
NeuralNotwerk
Contributor
Loading…
tools: repair int4-converted MTP heads in place; warn on --mtp --ebits <8
#397
opened Jul 18, 2026 by
NeuralNotwerk
Contributor
Loading…
packaging: pyproject.toml (pip install) + .editorconfig + .clang-format
#396
opened Jul 18, 2026 by
ZacharyZcR
Contributor
Loading…
4 tasks done
Fix CUDA detection on Windows (coli/doctor.py always report CPU-only)
#395
opened Jul 18, 2026 by
Stonki13
Loading…
release: version infrastructure + GitHub Release workflow (v1.0.0)
#394
opened Jul 18, 2026 by
ZacharyZcR
Contributor
Loading…
3 of 4 tasks
plan: auto-tune heuristics — MTP/PIPE/NUMA decisions from bottleneck classification
#393
opened Jul 18, 2026 by
ZacharyZcR
Contributor
Loading…
3 tasks done
Metal: MTLResidencySet residency for expert slabs — removes the per-submit useResource walk (#379)
#392
opened Jul 18, 2026 by
monotophic
Contributor
Loading…
refactor: split glm.c → colibri.c + 4 header modules (−18%)
#391
opened Jul 18, 2026 by
ZacharyZcR
Contributor
Loading…
3 of 4 tasks
Metal: platform-aware expert-cache defaults via honest storage probe (#379)
#386
opened Jul 18, 2026 by
monotophic
Contributor
Loading…
feat: add local-cluster MoE, dense sharding, and WebGPU workers
#380
opened Jul 18, 2026 by
gauravsaini
Loading…
security: harden model-file parsing, server, and build (untrusted-mirror threat model)
#368
opened Jul 17, 2026 by
KingIcyCreamProjects
Contributor
Loading…
Windows: fix non-ASCII chat prompt corruption (ANSI codepage vs UTF-8)
#366
opened Jul 17, 2026 by
vologn4mor
Loading…
4 tasks done
win32: auto-enable the GPU in coli chat (stacked on #298)
#364
opened Jul 17, 2026 by
woolcoxm
Contributor
Loading…
win32: auto-enable the GPU in bare coli chat
#363
opened Jul 17, 2026 by
woolcoxm
Contributor
Loading…
feat(prefetch): async expert prefetcher v4 -- 94.3% cache hit rate, 2.8x speedup, zero RAM overhead
#362
opened Jul 17, 2026 by
EgonRuiter
Loading…
tests: efficiency suite — tiny-model regression gates + full-model optimization dossier (#359)
#360
opened Jul 17, 2026 by
woolcoxm
Contributor
Loading…
fix(test): make test_stops build on Windows (mkdtemp compat shim)
#352
opened Jul 17, 2026 by
woolcoxm
Contributor
Loading…
cuda: COLI_GROUP_ASYNC=1 — overlap the CPU expert rows with the GPU groups at decode (opt-in, +6-8%)
#342
opened Jul 17, 2026 by
ZacharyZcR
Contributor
Loading…
AMD GPU support: HIP/ROCm via single-source backend_gpu_compat.h (+ WMMA dispatch gate)
#339
opened Jul 17, 2026 by
noobdev-ph
Loading…
6 tasks done
GPU backend: failure-path hardening + tests (sticky-error fix, cached-tensor contract, fault-injection hook)
#338
opened Jul 17, 2026 by
noobdev-ph
Loading…
5 of 6 tasks
resource_plan: fix --auto-tier pinning decode to one core (#325)
#332
opened Jul 16, 2026 by
woolcoxm
Contributor
Loading…
Previous Next
ProTip!
no:milestone will show everything without a milestone.