-
Notifications
You must be signed in to change notification settings - Fork 3.7k
Pull requests: JustVugg/colibri
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
fix(serve): an unpaired surrogate escape answered 500 "engine failed"
#1589
opened Sep 17, 2026 by
kevin9327
Loading…
fix(qwen36): refuse loudly on a failed encode-buffer realloc
#1588
opened Sep 17, 2026 by
alter
Loading…
3 of 6 tasks
fix(serve): a non-object json_schema answered 500 "engine failed"
#1587
opened Sep 17, 2026 by
kevin9327
Loading…
fix(coli): --gpu on DeepSeek V4 refused an install with only the DeepGEMM DLL
#1586
opened Sep 17, 2026 by
kevin9327
Loading…
fix(planner): auto-tune advised colibri.c-only knobs for every engine
#1585
opened Sep 17, 2026 by
kevin9327
Loading…
fix(autotune): coli tune swept GLM-only CUDA and NUMA knobs on every engine
#1584
opened Sep 17, 2026 by
kevin9327
Loading…
glm53: replay a bare tool-call turn as the model wrote it (#1576)
#1583
opened Sep 17, 2026 by
JustVugg
Owner
Loading…
coli: --auto-tier applies the plan's VRAM tier on sibling engines, and says why when it cannot (#1581)
#1582
opened Sep 17, 2026 by
JustVugg
Owner
Loading…
fix(qwen36): the pre-init device-count probe must load the backend DLL (#1577)
#1579
opened Sep 17, 2026 by
ZhiyangK
Loading…
dense Qwen3 family: engine, int4-gs64 converter, tiny-fixture oracle, registry entry
#1578
opened Sep 16, 2026 by
Xore
Loading…
docs: clarify documentation and typo fixes are welcome
#1575
opened Sep 16, 2026 by
cyberspace-cs
Loading…
perf(olmoe): O(1) LRU victim selection via intrusive recency lists (#1050)
#1571
opened Sep 16, 2026 by
Petsku01
Contributor
Loading…
tests: skip the cuda-test recipe checks when make is absent
#1569
opened Sep 16, 2026 by
SurefireStudios
Loading…
5 of 6 tasks
feat(web): add Indonesian dashboard translation
#1568
opened Sep 16, 2026 by
Yoruxyv
Loading…
10 tasks done
glm53 vulkan: clamp SiLU*Up in the shared gate_up shader
#1566
opened Sep 16, 2026 by
gyanu2507
Loading…
qt expert tier + expert-cache RAM ceilings (glm53 & qwen36), A3 soak-verified
#1564
opened Sep 16, 2026 by
Xore
Loading…
qwen36: --down-bits, the mixed expert layout (int4 gate/up, int8 down) -- the knob behind the #1370 numbers
#1559
opened Sep 15, 2026 by
kreuzzelg
Contributor
Loading…
fix(qwen36): a truncated multibyte tail made the tokenizer read past the prompt
#1557
opened Sep 15, 2026 by
Ti-03
Loading…
3 of 6 tasks
colibri: add CONSIST=1 prefill/decode self-consistency check
#1555
opened Sep 15, 2026 by
phreakocious
Loading…
3 of 6 tasks
qwen38: the int8 trunk carries one scale per 64 weights (grouped fmt 1 on the GPU), perplexity-neutral
#1552
opened Sep 15, 2026 by
kreuzzelg
Contributor
Loading…
fix(st): publish the lazy shard-mapping table without a data race
#1549
opened Sep 15, 2026 by
basil-k-aji-dev
Loading…
Previous Next
ProTip!
Mix and match filters to narrow down what you’re looking for.