grenade

grenade pushed to main at helexa/helexa

2026-05-21 08:50:34 +00:00

09c945f81e feat(stage-8d-4): dispatch chunked_gated_delta_rule_recurrence at prefill

grenade pushed to main at helexa/helexa

2026-05-21 08:49:45 +00:00

05dc0bad18 feat(stage-8d-3): wire causal_conv1d_update/full CUDA kernels

grenade pushed to main at helexa/helexa

2026-05-21 08:44:18 +00:00

10c151efa5 feat(stage-8d-5): wire gated_delta_rule_recurrence kernel into tp_qwen3_5

grenade pushed to main at helexa/helexa

2026-05-21 08:39:35 +00:00

44ae927e38 feat(stage-8d-2): wire gated_delta_rule_recurrence kernel into qwen3_5

grenade pushed to main at helexa/helexa

2026-05-21 08:34:15 +00:00

1ebbe87651 feat(stage-8d-1): import mistralrs GDN CUDA kernels — build infra only

grenade pushed to main at helexa/helexa

2026-05-21 05:22:05 +00:00

70eb6af42b feat(tp): cancellation-safe inference + structured tracing

grenade pushed to main at helexa/helexa

2026-05-21 04:39:41 +00:00

d1a4aad91d fix(tp): always drain worker responses on leader failure

grenade pushed to main at helexa/helexa

2026-05-20 19:02:45 +00:00

95dc8745eb feat(stage-8c): TP-aware Qwen3-Next (tp_qwen3_5)

grenade pushed to main at helexa/helexa

2026-05-20 18:13:23 +00:00

495d3f7c05 fix(qwen3_5): promote beta to F32 alongside q/k/v in delta rule

grenade pushed to main at helexa/helexa

2026-05-20 13:48:19 +00:00

5c4c8e0eba fix(qwen3_5): tensor names are under model.language_model.*, not model.*

grenade pushed to main at helexa/helexa

2026-05-20 13:47:54 +00:00

a77f19686e fix(qwen3_5): tensor names are under model.language_model.*, not model.*

grenade pushed to main at grenade/moments

2026-05-20 13:37:11 +00:00

2821548e6e feat(ui): add avg-by-hour panel to dashboard stats

grenade pushed to main at helexa/helexa

2026-05-20 13:18:55 +00:00

07c44d5db1 fix(qwen3_5): nested rope_parameters + partial_rotary_factor=0.25

grenade pushed to main at helexa/helexa

2026-05-20 12:52:38 +00:00

e7eb3dab6a feat(stage-8c): full-attention layer + decoder + Model + ForCausalLM for qwen3_5

grenade pushed to main at helexa/helexa

2026-05-20 06:29:56 +00:00

180274548d feat(stage-8c): linear-attention layer (Qwen3-Next GatedDeltaNet)

grenade pushed to main at helexa/helexa

2026-05-20 05:58:04 +00:00

a70f317729 feat(stage-8c): scaffold qwen3_5 (Qwen3.6) — dispatch + stubs + TP gate

grenade pushed to main at helexa/helexa

2026-05-20 05:36:26 +00:00

c6022aa6b9 feat(stage-8b): Llama + Qwen3 MoE families on the candle harness

grenade pushed to main at helexa/helexa

2026-05-20 05:27:33 +00:00

9e31d8deca feat(stage-8a): pre-flight architecture check for dense model loads

grenade pushed to main at grenade/moments

2026-05-20 04:59:10 +00:00

72eeb547af chore(deploy): self-heal /tmp perms before staging

86411bb88e fix(worker): dedup gitea events from overlapping user and org feeds

Compare 2 commits »

grenade pushed to main at helexa/helexa

2026-05-20 04:52:54 +00:00

b400e8b704 feat(neuron): honour HF_HUB_CACHE / HF_HOME for the candle harness cache