ahuston-0/nix-dotfiles
1
0
Fork 0
Code Issues Pull Requests 1 Actions 1 Packages Projects Releases Wiki Activity

feature/fwupd #196

Merged
ahuston-0 merged 3 commits from feature/fwupd into main 2026-03-23 23:24:37 -04:00
Conversation 0 Commits 3 Files Changed 1 +8 -9
1 changed files with 8 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
systems/palatine-hill/ollama.nix
View File

@@ -12,19 +12,18 @@ in
package = pkgs.ollama;
syncModels = true;
loadModels = [
"deepseek-r1:latest"
"deepseek-r1:1.5b"
"deepseek-r1:32b"
"deepseek-r1:70b"
"qwen3"
#"qwen3-coder-next"
"qwen3-coder"
"qwen3-coder-next"
"lennyerik/zeta"
"qwen2.5-coder:1.5b-base"
"qwen2.5-coder:14b"
"qwen2.5-coder:32b"
"nomic-embed-text:latest"
"llama3.3:8b"
"llama4:scout"
"mistral:7b"
"deepseek-v3.2:cloud"
"minimax-m2.7:cloud"
"lfm2:24b"
"glm-4.7-flash"
@@ -34,16 +33,16 @@ in
FLASH_ATTENTION = "1";
OLLAMA_KV_CACHE_TYPE = "q8_0";
# Ollama memory configuration
OLLAMA_MAX_LOADED_MODELS = "2";
OLLAMA_MAX_QUEUE = "4";
OLLAMA_NUM_PARALLEL = "2";
OLLAMA_MAX_LOADED_MODELS = "3";
OLLAMA_MAX_QUEUE = "512";
OLLAMA_NUM_PARALLEL = "3";
# ROCm memory optimization
#HIP_VISIBLE_DEVICES = "0";
#ROCR_VISIBLE_DEVICES = "0";
# context length for agents
OLLAMA_CONTEXT_LENGTH = "64000";
OLLAMA_CONTEXT_LENGTH = "128000";
};
openFirewall = true;
host = "0.0.0.0"; # don't want to make this available via load-balancer yet, so making it available on the local network