Search results

Rotation-preconditioned quant for streamed experts (int2/int3 quality) worth considering?

Status: Open.

#81 In JustVugg/colibri;

· locallypwned opened on Jul 12, 2026

ARM64 (DGX Spark GB10): builds, runs, CUDA kernels pass — but generation is deterministically wrong (0/20 vs oracle)

Status: Open.

#76 In JustVugg/colibri;

· fishkiler opened on Jul 12, 2026

[Bug]: Too much ram used

Status: Open.

#74 In JustVugg/colibri;

· EugeoSynthesisThirtyTwo opened on Jul 12, 2026

[Feature]: Rocm Support

Status: Open.

#69 In JustVugg/colibri;

· cb88 opened on Jul 11, 2026

[Feature]: Language (Italian) hinders worlwide contribution

Status: Open.

#67 In JustVugg/colibri;

· maikelthedev opened on Jul 11, 2026

[Performance]: many runs with Intel 265K, 128GB DDR5, 9100 Pro 8-10GB/s direct (+980 Pro 6GB/s direct)

Status: Open.

#65 In JustVugg/colibri;

· zacaj opened on Jul 11, 2026

[Question] On the language choice and boarder implications

Status: Open.

#63 In JustVugg/colibri;

· nathan-hello opened on Jul 11, 2026

[Performance]: Apple M5 Max · 48 GB · macOS · 0.08–0.25 tok/s

Status: Open.

#57 In JustVugg/colibri;

· lucianlature opened on Jul 11, 2026

Idea: grammar-aware draft source — forced tokens in constrained JSON output are free, guaranteed-accepted drafts (composes with DRAFT/MTP; engages even with the int4 head from #8)

Status: Open.

#48 In JustVugg/colibri;

· fabio-rovai opened on Jul 11, 2026

Benchmark: Apple M3 Max · macOS · 128 GB · internal SSD (12.5 GB/s) — 0.35–0.45 tok/s; PILOT neutral; confirms #8; setup.sh libomp false-positive

Status: Open.

#47 In JustVugg/colibri;

· fabio-rovai opened on Jul 11, 2026

[Performance]: 2x L4 on 36c Haswell w/ 128GB RAM

Status: Open.

#41 In JustVugg/colibri;

· briantopping opened on Jul 11, 2026

[Performance]: Framework Desktop Ryzen AI Max+ 128GB - CPU Only

Status: Open.

#39 In JustVugg/colibri;

· tommy-bolger opened on Jul 10, 2026