Vincenzo Fornaro 7a5bdee8b5 Merge pull request #1715 from namespaceMarcello/test/glm53-harness-discovery ...

test(glm53): rename the harnesses out of the unittest glob, run the tiny oracles in CI

2026-09-23 20:52:26 +02:00

..

api-reference

docs: add Mintlify API reference

2026-08-24 17:00:05 -07:00

baselines

feat(tools): add three-engine serving baseline campaigns

2026-09-23 14:28:09 +08:00

experiments

bench(cuda): measure paired resident projection batching

2026-09-22 20:29:24 +08:00

media

readme, docs: the Brain and Profiling pages as they are in 1.12.0

2026-09-21 08:00:38 +02:00

api.md

Merge pull request #1720 from monotophic/api/seed-accepted

2026-09-23 20:49:50 +02:00

benchmarking.md

feat(tools): add three-engine serving baseline campaigns

2026-09-23 14:28:09 +08:00

benchmarks.md

Merge pull request #1508 from JustVugg/docs/gpu-datapoint-quality-check

2026-09-14 21:54:35 +02:00

brio.md

gateway: POST /v1/systemone, the request and reply of Jev, served by the brio channel

2026-09-21 23:34:23 +02:00

BUILD-cuda-glibc241.md

docs: add CUDA-backend build recipe for glibc >= 2.41

2026-07-26 10:08:15 -03:00

CACHE_ROUTE.md

qwen36: port CACHE_ROUTE with the VRAM tier as the first residency level

2026-09-17 21:32:07 -05:00

corpus-draft.md

docs(corpus-draft): state the scope of the +22% and cross-link #689

2026-07-29 19:41:51 +02:00

cuda.md

docs: clarify compressed tier output semantics

2026-07-31 05:43:46 +08:00

deepseek-v4.md

Merge pull request #1291 from ZacharyZcR/fix/1281-windows-release

2026-09-06 19:08:39 +02:00

deepseek-v4.zh-CN.md

docs(v4): CUDA tier — build (Windows DLLs, Linux CUDA=1), serve line, VRAM budget + per-card settings, performance, GPU coverage, environment reference, validation

2026-08-16 18:38:16 +02:00

deepseek-v41.md

fix(v41): size gateway expert cache from the RAM budget

2026-09-22 14:49:29 +08:00

docs.json

docs: brio mode

2026-09-20 14:07:17 +02:00

edge-runtime.md

release: v1.10.0 — Qwen3.8-Flash-Next, with tools and vision (#1289)

2026-08-31 09:20:21 +02:00

ENVIRONMENT.md

Merge dev into feat/glm53-assistant-continuation: keep the max_tokens ceiling paragraph next to the continuation one

2026-09-22 23:23:52 +02:00

FORMATS.md

docs(fp8): FORMATS.md names symbols, so stop promising line numbers

2026-09-22 23:15:52 -04:00

glm53-flash.md

test(glm53): rename the harnesses out of the unittest glob, run the tiny oracles in CI

2026-09-23 15:38:45 +02:00

grammar-draft.md

docs: the engine is colibri(.exe) everywhere — kill the obsolete rename step

2026-07-22 08:33:28 +02:00

index.mdx

docs: add Mintlify API reference

2026-08-24 17:00:05 -07:00

inkling.md

docs: restore Inkling literal tokens

2026-08-25 05:17:35 -07:00

int4-rans256-g0.md

docs(int4-rans256-g0): disambiguate expected_bytes identity vs extent (#1273)

2026-09-19 22:46:47 +08:00

kimi_k3_metal_tuning.md

docs: make pages MDX-compatible

2026-08-24 17:10:59 -07:00

kimi_k3.md

Merge pull request #1073 from bherald/fix/k3-prefill-cancellation

2026-08-18 15:49:46 +02:00

kimi_metal_gap_analysis.md

Metal Kimi K3: aligned KDA state/window bufs, wrap-once buffer cache, CPU-side MLA KV cache, one-shot banners + throttled stats

2026-08-19 16:54:44 +02:00

MAINTAINING-DOCS.md

Document the 79 undocumented environment variables

2026-08-03 09:49:43 +08:00

metal_implementation.md

docs: make pages MDX-compatible

2026-08-24 17:10:59 -07:00

METAL-M1ULTRA-FMT2-REPORT.md

docs(fmt2): M1 Ultra fmt=2 report + benchmarks row + fmt4->fmt2 converter

2026-08-04 20:56:18 -07:00

METAL-M5MAX-PERF-REPORT.md

docs: Metal M5 Max perf report — OMP hot-team spin (#77) throttles Apple GPU, NO_OMP+PIPE recovers to 2.24 tok/s (#116)

2026-07-13 20:47:19 +02:00

metal.md

metal: GPU attention for S>4 prefill (single command buffer, opt-in)

2026-08-02 12:50:49 +01:00

METAL.txt

Metal Kimi K3: aligned KDA state/window bufs, wrap-once buffer cache, CPU-side MLA KV cache, one-shot banners + throttled stats

2026-08-19 16:54:44 +02:00

multidisk.md

docs: add multi-disk setup and measurement guide

2026-09-20 21:05:09 +08:00

quickstart.md

make: find MacPorts libomp on macOS, in both Darwin probes

2026-09-17 18:55:37 -05:00

qwen36-cuda-tier.md

Merge dev into fix/windows-qwen36-cuda-dll: keep dev's qwen36 prerequisites plus the alias and .build-config

2026-09-22 22:28:50 +02:00

qwen36-phase01.md

feat(qwen36): add Qwen3.6-35B-A3B engine (CPU): hybrid Gated Attention + Gated DeltaNet + streaming MoE

2026-08-18 18:35:49 +02:00

qwen36-phase02.md

feat(qwen36): add Qwen3.6-35B-A3B engine (CPU): hybrid Gated Attention + Gated DeltaNet + streaming MoE

2026-08-18 18:35:49 +02:00

qwen36.md

ci(qwen36), docs: the mixed expert layout on the tiny fixture, and the #1370 numbers in "Which container?"

2026-09-22 20:45:13 +02:00

qwen38.md

Merge pull request #1618 from bherald/docs/qwen38-vision-overview-20260918

2026-09-22 22:30:30 +02:00

routing-telemetry.md

feat(glm53): add routing usage telemetry

2026-09-13 13:42:38 +02:00

segment-runtime.md

feat(qwen38): complete bounded prompt runtime

2026-08-27 13:23:33 -05:00

serve_protocol.md

docs(glm53): the engine, its knobs, and the frame the protocol grew

2026-08-28 18:00:55 +02:00

SETTINGS.md

perf(tune): measure safe RAM and cache caps

2026-08-24 02:29:52 +02:00

tuning.md

Merge pull request #906 from kritikagarg/feat/degrade-zero-miss-policy

2026-09-21 05:46:07 +02:00

vulkan.md

docs(vulkan): how to run the Khronos validation layers, and one layer defect

2026-09-17 20:15:21 -05:00

windows.md

build: refuse the 32-bit cl.exe before nvcc runs (#1405)

2026-09-10 02:46:26 +02:00