adriangl/nix-dotfiles-v2
1
1
Fork 0
Code Issues Pull Requests Releases Activity

update llama-swap config

Browse Source
This commit is contained in:
Adrian Gunnar Lauterer
2026-05-19 17:28:30 +02:00
parent 07712fc6fd
commit 78f4dec006
2 changed files with 7 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files
hosts/galadriel/hardware-configuration.nix
+4 -2
View File
@@ -22,8 +22,10 @@
"usbhid"
"sd_mod"
];
boot.kernelPackages = pkgs.linuxPackages_6_19; # migth need to bump down if zfs compat breaks.
boot.zfs.package = pkgs.zfs_2_4;
# unstable zfs and linux because we needed zfs 2.4.2 for linux 7.0
boot.kernelPackages = pkgs.unstable.linuxPackages_7_0; # migth need to bump down if zfs compat breaks.
boot.zfs.package = pkgs.unstable.zfs; #think this was the correct way
boot.initrd.kernelModules = [ ];
boot.kernelModules = [ "kvm-amd" ];
boot.extraModulePackages = [ ];
modules/llama-swap.nix
+3 -21
View File
@@ -51,13 +51,13 @@
models = {
"qwen3.5-35b-a3b" = {
"qwen3.6-35b-a3b" = {
cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -ngl $\{ngl\} $\{kv_cache\} $\{batch\} --hf-repo $\{hf_repo\} $\{image-tokens\} $\{qwen35-no-thinking\} ";
aliases = [ "qwen3.5" ];
aliases = [ "qwen3.6" ];
ttl = 1800;
macros = {
ctx = 49152;
hf_repo = "unsloth/Qwen3.5-35B-A3B-GGUF:UD-IQ3_XXS";
hf_repo = "unsloth/Qwen3.6-35B-A3B-GGUF:UD-IQ3_XXS";
};
};
"qwen3.5-9b" = {
@@ -68,24 +68,6 @@
ctx = 128000;
};
};
"qwen3.5-2b" = {
cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} $\{kv_cache\} $\{batch\} $\{image-tokens\} -ngl $\{ngl\} --hf-repo $\{hf_repo\} ";
ttl = 900;
macros = {
hf_repo = "unsloth/Qwen3.5-2B-GGUF:UD-Q8_K_XL";
ctx = 200000;
};
};
"glm4.7-flash" = {
cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -fa 0 -ngl $\{ngl\} --hf-repo $\{hf_repo\}";
ttl = 900;
aliases = [ "coder" ];
macros = {
hf_repo = "unsloth/GLM-4.7-Flash-REAP-23B-A3B-GGUF"; # Reap should allow more context in gpu mem
ctx = 32000;
};
};
"gemma4" = {
cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -fa 0 -ngl $\{ngl\} --hf-repo $\{hf_repo\}";