pewdiepie-archdaemon a95ac7b47d Cookbook: auto-set KV cache to fp8 for DeepSeek V3/V4/R1 MoE families ...

These models OOM on --kv-cache-dtype auto (≈bf16) at any usable
context with current tensor-parallel layouts. _detectModelOptimizations
now seeds opts.kvCacheDtype='fp8' for them, and the serve panel's KV
Cache select picks that up as the default unless the user has a
saved override on this skill.

2026-06-14 08:57:29 +09:00

..

fonts

Odysseus v1.0

2026-05-31 23:58:26 +09:00

js

Cookbook: auto-set KV cache to fp8 for DeepSeek V3/V4/R1 MoE families

2026-06-14 08:57:29 +09:00

lib

Odysseus v1.0

2026-05-31 23:58:26 +09:00

app.js

Settings overhaul + UI polish pass

2026-06-10 15:15:13 +09:00

index.html

Settings: clamp logo SVGs to 18px chip + endpoint dropdown gets logo

2026-06-13 23:00:16 +09:00

login.html

fix(login): keep inputs >=16px on touch so iOS doesn't zoom on focus (#1632)

2026-06-03 14:23:24 +09:00

manifest.json

Odysseus v1.0

2026-05-31 23:58:26 +09:00

style.css

Settings: tighten endpoint logo+select gap + align fallback trash right

2026-06-13 23:04:27 +09:00

sw.js

fix(ui): modal drag + removed startDrag func (#2430)

2026-06-04 20:34:18 +02:00