updated models
This commit is contained in:
+12
-6
@@ -10,7 +10,7 @@
|
||||
services.llama-swap = {
|
||||
enable = true;
|
||||
package = pkgs.unstable.llama-swap;
|
||||
port = 8085;
|
||||
port = 11111;
|
||||
openFirewall = true;
|
||||
|
||||
settings =
|
||||
@@ -36,15 +36,21 @@
|
||||
cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -ngl $\{ngl\} $\{kv_cache\} --hf-repo $\{hf_repo\}";
|
||||
aliases = [ "qwen3.5" ];
|
||||
ttl = 900;
|
||||
macros.hf_repo = "unsloth/Qwen3.5-35B-A3B-GGUF:Q2_K_L";
|
||||
macros.ngl = 40;
|
||||
macros = {
|
||||
hf_repo = "unsloth/Qwen3.5-35B-A3B-GGUF:Q2_K_XL";
|
||||
ngl = 40;
|
||||
};
|
||||
};
|
||||
"qwen3.5-9b" = {
|
||||
cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -ngl $\{ngl\} --hf-repo $\{hf_repo\}";
|
||||
ttl = 900;
|
||||
macros.hf_repo = "unsloth/Qwen3.5-9B-GGUF";
|
||||
};
|
||||
|
||||
"ministal-3-8b-reasonning" = {
|
||||
cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -ngl $\{ngl\} $\{kv_cache\} --hf-repo $\{hf_repo\}";
|
||||
cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -ngl $\{ngl\} --hf-repo $\{hf_repo\}";
|
||||
aliases = [ "ministral" ];
|
||||
ttl = 900;
|
||||
macros.hf_repo = "mistralai/Ministral-3-8B-Reasoning-2512-GGUF:Q4_K_M";
|
||||
macros.hf_repo = "mistralai/Ministral-3-8B-Reasoning-2512-GGUF";
|
||||
};
|
||||
};
|
||||
|
||||
|
||||
Reference in New Issue
Block a user