Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦
C
Commits on main
qwen38: planner support (supports_accelerator) and the GPU section of the docs
768fc6c kreuzzelg committed 2 months ago
qwen38: CUDA VRAM expert tier (fp8 streaming mode), engine test, FP8 fixture shards
a5d177b kreuzzelg committed 2 months ago
tools(qwen38): --fp8-experts writes the tiny fixture in the release's expert layout
aa87ff5 kreuzzelg committed 2 months ago
Merge pull request #1477 from JustVugg/feat/expert-ffn-kernel
e92c139 Vincenzo Fornaro committed 2 months ago
expert_ffn.h: one routed-expert kernel for the MoE engines, qwen36 first
275a1ee JustVugg committed 2 months ago
Merge pull request #1473 from JustVugg/docs/discord-invite-main
f028d26 Vincenzo Fornaro committed 2 months ago
docs: link the GLM-5.3 int4 container from the model section (#1403)
e982635 HoneyTyagii committed 2 months ago
docs: replace the expired Discord invite (all 4 READMEs + site)
25db80d JustVugg committed 2 months ago
chore(glm53): standardize user-facing diagnostics in English
58498af trigger2k20 committed 2 months ago
Merge pull request #1471 from JustVugg/docs/discord-invite
e5e23de Vincenzo Fornaro committed 2 months ago
docs: replace the expired Discord invite (all 4 READMEs + site)
cad8f06 JustVugg committed 2 months ago
Merge pull request #1468 from JustVugg/fix/v4-moe-says-why
7d3e30c Vincenzo Fornaro committed 2 months ago
fix(deepseek-v4): the MoE step says why it failed, instead of "failed in MoE"
849f43f JustVugg committed 2 months ago
Merge pull request #1465 from JustVugg/release-1.11.0
e96f336 Vincenzo Fornaro committed 2 months ago
Merge pull request #1456 from trigger2k20/fix/glm53-resource-planning
b6ffc1a Vincenzo Fornaro committed 2 months ago
Merge pull request #1457 from trigger2k20/feat/glm53-routing-telemetry
771f995 Vincenzo Fornaro committed 2 months ago
Merge pull request #1463 from JustVugg/fix/glm53-mat-initializers
8b2cf38 Vincenzo Fornaro committed 2 months ago
fix(glm53): initialise the two Mat fields #1462 added, so #1458's cleanup holds
6f03a0f JustVugg committed 2 months ago
Merge pull request #1458 from trigger2k20/fix/clang-compiler-warnings
248e130 Vincenzo Fornaro committed 2 months ago
fix(glm53): correct RAM sizing and Apple unified-memory planning
58ed660 trigger2k20 committed 2 months ago
Merge pull request #1462 from trigger2k20/feat/glm53-metal-clean
4468c30 Vincenzo Fornaro committed 2 months ago