grenade

grenade pushed to main at helexa/cortex

2026-05-19 15:32:11 +00:00

12549c9aed fix(tp): import BackendStorage trait for CudaStorage methods

46527d7804 feat(tp): TP-aware Qwen3 dense model (Stage 7b-iii 2/2)

Compare 2 commits »

grenade pushed to main at helexa/cortex

2026-05-19 15:14:58 +00:00

8d3194f992 Stage 7b-iii (1/2): AllReduce CustomOp + ShardedVarBuilder-backed TP linears

grenade pushed to main at helexa/cortex

2026-05-19 14:49:46 +00:00

5436af9c73 fix(neuron/candle): dense Qwen3 returns rank-3 logits, double-squeeze

grenade pushed to main at helexa/cortex

2026-05-19 14:24:17 +00:00

8e882c0757 fix(neuron/tp): NcclError {e:?} + cudarc 0.19 deprecation cleanup

grenade pushed to main at helexa/cortex

2026-05-19 14:07:23 +00:00

93421f48e2 Stage 7b-ii: ColumnParallel + RowParallel sharded linear primitives

grenade pushed to main at helexa/cortex

2026-05-19 14:04:04 +00:00

05e15f3597 Stage 7b-i: dense safetensors Qwen3 load path

grenade pushed to main at helexa/cortex

2026-05-19 13:40:04 +00:00

da068ded6d Stage 7a-ii: real NCCL handshake behind the worker pool

grenade pushed to main at helexa/cortex

2026-05-19 12:53:04 +00:00

2a7ede0232 Stage 7a-i: TP worker lifecycle scaffolding

grenade pushed to main at helexa/cortex

2026-05-19 11:48:12 +00:00

18ae3c30ee post-validation cleanup: cuDNN runtime + repetition penalty

grenade pushed to main at helexa/cortex

2026-05-19 11:10:52 +00:00

1a0400131e fix(deploy): use dnf upgrade for stale installs, install only when absent

grenade pushed to main at helexa/cortex

2026-05-19 10:43:07 +00:00

1866b99a89 fix(validate-neuron): jq for JSON, say→stderr, sane max_tokens

60176e7c2e ci: monotonic prerelease versions + serialize CI on shared runner

Compare 2 commits »

grenade pushed to main at helexa/cortex

2026-05-19 10:16:43 +00:00

602e8e1471 fix(neuron/candle): source tokenizer.json from base repo when GGUF

grenade pushed to main at helexa/cortex

2026-05-19 10:13:40 +00:00

e9d0a75dd5 ci(prerelease): auto-build on every push to main

6cf87e328f chore(neuron): log load_model failures server-side with full chain

Compare 2 commits »

grenade pushed to main at gongfoo/gongfoo

2026-05-19 05:30:18 +00:00

96462c07b2 feat(controller): per-host label allow-list for placement

grenade pushed to main at helexa/cortex

2026-05-19 05:20:46 +00:00

f9f5fa41b6 fix(neuron): surface full anyhow chain + ensure $HOME exists at start

ed4d71db09 fix(validate-neuron): default to unsloth GGUF + capture curl errors

39010c779f add script/validate-neuron.sh — end-to-end candle harness smoke test

Compare 3 commits »

grenade pushed to main at helexa/cortex

2026-05-19 04:16:44 +00:00

57d7ef8d3c chore: revert dnf. runner user has no system privs

grenade pushed to main at helexa/cortex

2026-05-19 04:08:33 +00:00

0e9671dd7d fix(ci): drop sudo from dnf install (runner runs as root, no sudo)

e29c9e35f0 fix(ci): ensure rust toolchain present on cuda-13.0 runner

8a2334eacb deploy: dnf-native version check + lair.cafe repo bootstrap

aad314cdfa feat(neuron): graceful unload-on-shutdown via SIGTERM/SIGINT

Compare 4 commits »

grenade pushed to main at helexa/cortex

2026-05-18 14:56:43 +00:00

6779b7526a feat(neuron): load default_models on service activation

84f5662df1 feat(neuron): OpenAI-compatible SSE streaming chat completions

Compare 2 commits »

grenade pushed to main at helexa/cortex

2026-05-18 14:50:40 +00:00

249c9442e8 chore: track deployment script

5e17081fb4 ci(prerelease): drop redundant rustup install step

Compare 2 commits »

grenade pushed to main at helexa/cortex

2026-05-18 14:40:59 +00:00

03bed93fee add asset/manifest.yml describing fleet hosts and neuron flavours

4a5211d830 ci(prerelease): add ampere flavour alongside ada and blackwell

6d2dc5ff1a fix(ci): give fmt/clippy/test distinct CARGO_TARGET_DIR to avoid races

Compare 3 commits »