Dixie #1

Merged
kyle merged 12 commits from dixie into master 2026-09-22 04:56:26 +00:00
2 changed files with 9 additions and 7 deletions
Showing only changes of commit 0543864f38 - Show all commits
+1 -2
View File
@@ -17,7 +17,7 @@ let
cfg = { cfg = {
"$schema" = "https://opencode.ai/config.json"; "$schema" = "https://opencode.ai/config.json";
enabled_providers = [ "llama.cpp" "zai-coding-plan" ]; enabled_providers = [ "llama.cpp" "zai-coding-plan" ];
model = "llama.cpp/laguna-s-2.1"; model = "llama.cpp/ornith-1.5-35b-a3b";
small_model = "llama.cpp/ornith-1.5-9b-uncensored"; small_model = "llama.cpp/ornith-1.5-9b-uncensored";
provider = { provider = {
"zai-coding-plan" = { "zai-coding-plan" = {
@@ -52,7 +52,6 @@ let
chunkTimeout = 7200000; chunkTimeout = 7200000;
}; };
models = { models = {
"laguna-s-2.1" = text "Laguna S 2.1";
"qwen3.6-35b-a3b-abliterated" = text "Qwen3.6 35B A3B"; "qwen3.6-35b-a3b-abliterated" = text "Qwen3.6 35B A3B";
"gemma4-26b-a4b-abliterated" = text "Gemma 4 26B A4B"; "gemma4-26b-a4b-abliterated" = text "Gemma 4 26B A4B";
"ornith-1.0-35b" = text "Ornith 1.0 35B"; "ornith-1.0-35b" = text "Ornith 1.0 35B";
+8 -5
View File
@@ -42,14 +42,17 @@ let
; DFlash GGUF is on disk (laguna-s-2.1-DFlash-BF16.gguf). nixpkgs llama.cpp 0.4.0 ; DFlash GGUF is on disk (laguna-s-2.1-DFlash-BF16.gguf). nixpkgs llama.cpp 0.4.0
; rejects it at load ("expected 76, got 69" tensors); Poolside's file needs their ; rejects it at load ("expected 76, got 69" tensors); Poolside's file needs their
; laguna-branch loader. Leave unwired so Laguna still starts. ; laguna-branch loader. Leave unwired so Laguna still starts.
[laguna-s-2.1] ; DISABLED 2026-09-19: resident beside Ornith 35B it left ~15 GB free, and a third
model = ${llamaModelsDir}/laguna-s-2.1-UD-Q4_K_S/Laguna-S-2.1-UD-Q4_K_S-00001-of-00003.gguf ; model load then hit the 104 GiB TTM cap and OOM-killed the desktop session.
; Weights stay on disk; uncomment to re-enable.
; [laguna-s-2.1]
; model = ${llamaModelsDir}/laguna-s-2.1-UD-Q4_K_S/Laguna-S-2.1-UD-Q4_K_S-00001-of-00003.gguf
; Two slots sharing one unified 131072-token KV pool: a second request (subagent, ; Two slots sharing one unified 131072-token KV pool: a second request (subagent,
; another client) runs alongside instead of queueing, and either one may still use ; another client) runs alongside instead of queueing, and either one may still use
; the full 131072 as long as their combined context fits. ; the full 131072 as long as their combined context fits.
parallel = 2 ; parallel = 2
kv-unified = true ; kv-unified = true
ctx-size = 131072 ; ctx-size = 131072
; Previous default model (a Qwen 3.5 derivative), kept for comparison runs. It was ; Previous default model (a Qwen 3.5 derivative), kept for comparison runs. It was
; fetched by the old -hf flag into the HF cache; dedup hides the cache's own entry. ; fetched by the old -hf flag into the HF cache; dedup hides the cache's own entry.