Skip to content

Pull requests: JustVugg/colibri

New pull request New
26 Open 448 Closed
26 Open 448 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

metal: GPU attention for S>4 prefill (single command buffer)
#763 opened Aug 1, 2026 by RDouglasSharp Contributor Loading…
fix(cuda): dispatch grouped int4 (fmt=4) in the async expert-group path
#762 opened Aug 1, 2026 by kreuzzelg Loading…
1
docs: coli serve carries the dashboard too, and say so for headless hosts
#761 opened Aug 1, 2026 by terrizoaguimor Contributor Loading…
inkling: shared experts to the GPU, 1.75 to 2.46 tok/s on Apple Silicon metal Backend Metal/Apple performance Velocità / tok-s / ottimizzazioni
#757 opened Aug 1, 2026 by rgbkrk Contributor Loading…
1
Vulkan MoE GEMV backend for integrated/AMD GPUs (draft, complements #418) enhancement New feature or request vulkan Backend Vulkan/AMD
#729 opened Jul 31, 2026 by minne100 • Draft
1
feat(qwen36): CUDA VRAM expert tier — heat-based placement across GPUs via the shared CUDA backend cuda Backend CUDA/NVIDIA model-support Supporto a nuovi modelli
#713 opened Jul 30, 2026 by kreuzzelg • Draft
feat(qwen36): Qwen3.6-35B-A3B engine (CPU): hybrid Gated Attention + Gated DeltaNet + streaming MoE model-support Supporto a nuovi modelli
#712 opened Jul 30, 2026 by kreuzzelg Loading…
12
feat(win): fix silent CPU fallback, launcher suite, DirectStorage expert loads enhancement New feature or request needs-rebase Confligge, serve rebase dell'autore
#670 opened Jul 28, 2026 by khalilswdp Contributor Loading…
8 tasks done
3
feat(core): model-architecture seam, chat templates, text antiprompt enhancement New feature or request needs-rebase Confligge, serve rebase dell'autore
#667 opened Jul 28, 2026 by khalilswdp Contributor Loading…
5 tasks done
7
QLoRA training path: fine-tune GLM-5.2 (744B) in 64 GB RAM enhancement New feature or request feature Nuova funzionalità
#626 opened Jul 26, 2026 by pavolbauer Loading…
5 tasks done
5
WIP: MiniMax-M3 support — GQA + MSA block-sparse attention, o200k tokenizer, converter (follow-up to #418) model-support Supporto a nuovi modelli
#601 opened Jul 24, 2026 by steve-m Contributor Loading…
5 tasks
2
Metal fmt=4 grouped-int4 decode: attention + routed experts (#585) metal Backend Metal/Apple needs-rebase Confligge, serve rebase dell'autore
#587 opened Jul 24, 2026 by RDouglasSharp Contributor Loading…
27
Preserve model-declared EOS tokens in serve mode enhancement New feature or request needs-rebase Confligge, serve rebase dell'autore
#584 opened Jul 24, 2026 by saskw2010 • Draft
2
Fix stateful KV tail at the NGEN limit bug Difetto verificato nel codice
#567 opened Jul 23, 2026 by winklemad Contributor Loading…
3 of 5 tasks
4
add persistence controls for lower SSD writes enhancement New feature or request
#555 opened Jul 23, 2026 by Skater1808 Loading…
5 tasks
4
CPU: KV cache quantization — KV8 (fp8 e4m3) + KV_TQ (rotated-int4 / PolarQuant) enhancement New feature or request performance Velocità / tok-s / ottimizzazioni
#553 opened Jul 23, 2026 by NeuralNotwerk Contributor Loading…
7
feat: add distributed expert workers enhancement New feature or request
#551 opened Jul 23, 2026 by gauravsaini Loading…
1
feat: add dense MLP activation sharding enhancement New feature or request performance Velocità / tok-s / ottimizzazioni
#550 opened Jul 23, 2026 by gauravsaini Loading…
1
feat: add Qwen3-30B-A3B engine model-support Supporto a nuovi modelli needs-rebase Confligge, serve rebase dell'autore
#544 opened Jul 23, 2026 by opxyc Contributor Loading…
4 of 5 tasks
9
pilot: multi-worker PILOT_REAL prefetch (PILOT_WORKERS) — byte-identical base for the #441 hardware A/B performance Velocità / tok-s / ottimizzazioni
#480 opened Jul 21, 2026 by cdhdt Contributor • Draft
4 of 5 tasks
6
Port the coli CLI launcher to Go (dependency-free) — first step for #310 enhancement New feature or request
#454 opened Jul 20, 2026 by Magnet-js Contributor • Draft
2
nix: CUDA support enhancement New feature or request
#416 opened Jul 19, 2026 by attilaolah Contributor • Draft
5 tasks
4
KV cache quantization: fp8 (KV8) + 4-bit TurboQuant (KV_TQ) on CPU, CUDA, and Metal enhancement New feature or request performance Velocità / tok-s / ottimizzazioni
#399 opened Jul 18, 2026 by NeuralNotwerk Contributor Loading…
16
feat: Add NUMA-aware RAM-disk streaming discussion Proposta / discussione aperta, non un task enhancement New feature or request needs-rebase Confligge, serve rebase dell'autore performance Velocità / tok-s / ottimizzazioni
#377 opened Jul 17, 2026 by BColsey Loading…
12
Nearly double the hit-rate needs-rebase Confligge, serve rebase dell'autore performance Velocità / tok-s / ottimizzazioni
#223 opened Jul 14, 2026 by withinboredom Contributor Loading…
4 of 5 tasks
10
ProTip! Type g i on any issue or pull request to go back to the issue listing page.