chore: update model paths

homes/x86_64-linux/evanreichard@lin-va-desktop/default.nix

@@ -23,6 +23,7 @@ in
       terminal = {
         bash = enabled;
         btop = enabled;
+        direnv = enabled;
         tmux = enabled;
       };
     };

systems/x86_64-linux/lin-va-desktop/default.nix

@@ -97,8 +97,9 @@ in
           cmd = ''
             ${pkgs.reichard.llama-cpp}/bin/llama-server \
               --port ''${PORT} \
-              -m /mnt/ssd/Models/Devstral-Small-2-24B-Instruct-2512-UD-Q4_K_XL.gguf \
+              -m /mnt/ssd/Models/Devstral/Devstral-Small-2-24B-Instruct-2512-UD-Q4_K_XL.gguf \
-              --chat-template-file /mnt/ssd/Models/Devstral-Small-2-24B-Instruct-2512-UD-Q4_K_XL_template.jinja \
+              --chat-template-file /mnt/ssd/Models/Devstral/Devstral-Small-2-24B-Instruct-2512-UD-Q4_K_XL_template.jinja \
+              --temp 0.15 \
               -c 98304 \
               -ctk q8_0 \
               -ctv q8_0 \
@@ -107,14 +108,30 @@ in
           '';
         };
 
+        # https://huggingface.co/mradermacher/gpt-oss-20b-heretic-v2-i1-GGUF/tree/main
+        #  --chat-template-kwargs '{\"reasoning_effort\":\"low\"}'
+        "gpt-oss-20b-thinking" = {
+          name = "GPT OSS (20B) - Thinking";
+          cmd = ''
+            ${pkgs.reichard.llama-cpp}/bin/llama-server \
+              --port ''${PORT} \
+              -m /mnt/ssd/Models/GPT-OSS/gpt-oss-20b-heretic-v2.i1-MXFP4_MOE.gguf \
+              -c 131072 \
+              --temp 1.0 \
+              --top-p 1.0 \
+              --top-k 40 \
+              -dev CUDA0
+          '';
+        };
+
         # https://huggingface.co/unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF/tree/main
         "qwen3-next-80b-instruct" = {
           name = "Qwen3 Next (80B) - Instruct";
           cmd = ''
             ${pkgs.reichard.llama-cpp}/bin/llama-server \
               --port ''${PORT} \
-              -m /mnt/ssd/Models/Qwen3-Next-80B-A3B-Instruct-UD-Q4_K_XL.gguf \
+              -m /mnt/ssd/Models/Qwen3/Qwen3-Next-80B-A3B-Instruct-UD-Q2_K_XL.gguf \
-              -c 131072 \
+              -c 262144 \
               --temp 0.7 \
               --min-p 0.0 \
               --top-p 0.8 \
@@ -122,10 +139,25 @@ in
               --repeat-penalty 1.05 \
               -ctk q8_0 \
               -ctv q8_0 \
-              -fit off \
+              -fit off
-              -ncmoe 15 \
-              -ts 77,23
           '';
+
+          # cmd = ''
+          #   ${pkgs.reichard.llama-cpp}/bin/llama-server \
+          #     --port ''${PORT} \
+          #     -m /mnt/ssd/Models/Qwen3/Qwen3-Next-80B-A3B-Instruct-UD-Q4_K_XL.gguf \
+          #     -c 131072 \
+          #     --temp 0.7 \
+          #     --min-p 0.0 \
+          #     --top-p 0.8 \
+          #     --top-k 20 \
+          #     --repeat-penalty 1.05 \
+          #     -ctk q8_0 \
+          #     -ctv q8_0 \
+          #     -fit off \
+          #     -ncmoe 15 \
+          #     -ts 77,23
+          # '';
         };
 
         # https://huggingface.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF/tree/main
@@ -134,7 +166,7 @@ in
           cmd = ''
             ${pkgs.reichard.llama-cpp}/bin/llama-server \
               --port ''${PORT} \
-              -m /mnt/ssd/Models/Qwen3-30B-A3B-Instruct-2507-Q4_K_M.gguf \
+              -m /mnt/ssd/Models/Qwen3/Qwen3-30B-A3B-Instruct-2507-Q4_K_M.gguf \
               -c 262144 \
               --temp 0.7 \
               --min-p 0.0 \
@@ -153,7 +185,7 @@ in
           cmd = ''
             ${pkgs.reichard.llama-cpp}/bin/llama-server \
               --port ''${PORT} \
-              -m /mnt/ssd/Models/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M.gguf \
+              -m /mnt/ssd/Models/Qwen3/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M.gguf \
               -c 262144 \
               --temp 0.7 \
               --min-p 0.0 \
@@ -172,7 +204,7 @@ in
           cmd = ''
             ${pkgs.reichard.llama-cpp}/bin/llama-server \
               --port ''${PORT} \
-              -m /mnt/ssd/Models/Qwen3-30B-A3B-Thinking-2507-UD-Q4_K_XL.gguf \
+              -m /mnt/ssd/Models/Qwen3/Qwen3-30B-A3B-Thinking-2507-UD-Q4_K_XL.gguf \
               -c 262144 \
               --temp 0.7 \
               --min-p 0.0 \
@@ -185,19 +217,17 @@ in
           '';
         };
 
-        # https://huggingface.co/mradermacher/gpt-oss-20b-heretic-v2-i1-GGUF/tree/main
+        # https://huggingface.co/unsloth/Nemotron-3-Nano-30B-A3B-GGUF/tree/main
-        #  --chat-template-kwargs '{\"reasoning_effort\":\"low\"}'
+        "nemotron-3-nano-30b-thinking" = {
-        "gpt-oss-20b-thinking" = {
+          name = "Nemotron 3 Nano (30B) - Thinking";
-          name = "GPT OSS (20B) - Thinking";
           cmd = ''
             ${pkgs.reichard.llama-cpp}/bin/llama-server \
               --port ''${PORT} \
-              -m /mnt/ssd/Models/gpt-oss-20b-heretic-v2.i1-MXFP4_MOE.gguf \
+              -m /mnt/ssd/Models/Nemotron/Nemotron-3-Nano-30B-A3B-UD-Q4_K_XL.gguf \
-              -c 131072 \
+              -c 1048576 \
-              --temp 1.0 \
+              --temp 1.1 \
-              --top-p 1.0 \
+              --top-p 0.95 \
-              --top-k 40 \
+              -fit off
-              -dev CUDA0
           '';
         };
 
@@ -208,7 +238,7 @@ in
             ${pkgs.reichard.llama-cpp}/bin/llama-server \
               --port ''${PORT} \
               -m /mnt/ssd/Models/Qwen3-VL-8B-Instruct-UD-Q4_K_XL.gguf \
-              --mmproj /mnt/ssd/Models/Qwen3-VL-8B-Instruct-UD-Q4_K_XL_mmproj-F16.gguf \
+              --mmproj /mnt/ssd/Models/Qwen3/Qwen3-VL-8B-Instruct-UD-Q4_K_XL_mmproj-F16.gguf \
               -c 65536 \
               --temp 0.7 \
               --min-p 0.0 \
@@ -225,7 +255,7 @@ in
           name = "Qwen2.5 Coder (7B) - Instruct";
           cmd = ''
             ${pkgs.reichard.llama-cpp}/bin/llama-server \
-              -m /mnt/ssd/Models/Qwen2.5-Coder-7B-Instruct-Q8_0.gguf \
+              -m /mnt/ssd/Models/Qwen2.5/Qwen2.5-Coder-7B-Instruct-Q8_0.gguf \
               --fim-qwen-7b-default \
               -c 131072 \
               --port ''${PORT} \
@@ -238,7 +268,7 @@ in
           name = "Qwen2.5 Coder (3B) - Instruct";
           cmd = ''
             ${pkgs.reichard.llama-cpp}/bin/llama-server \
-              -m /mnt/ssd/Models/Qwen2.5-Coder-3B-Instruct-Q8_0.gguf \
+              -m /mnt/ssd/Models/Qwen2.5/Qwen2.5-Coder-3B-Instruct-Q8_0.gguf \
               --fim-qwen-3b-default \
               --port ''${PORT} \
               -fit off \
@@ -246,58 +276,17 @@ in
           '';
         };
 
-        # https://huggingface.co/unsloth/SmolLM3-3B-128K-GGUF/tree/main
+        # https://huggingface.co/unsloth/Qwen3-4B-Instruct-2507-GGUF/tree/main
-        "smollm3-3b-instruct" = {
+        "qwen3-4b-2507-instruct" = {
-          name = "SmolLM3(3B) - Instruct";
+          name = "Qwen3 2507 (4B) - Instruct";
           cmd = ''
             ${pkgs.reichard.llama-cpp}/bin/llama-server \
               --port ''${PORT} \
-              -m /mnt/ssd/Models/SmolLM3-3B-128K-UD-Q4_K_XL.gguf \
+              -m /mnt/ssd/Models/Qwen3/Qwen3-4B-Instruct-2507-Q4_K_M.gguf \
               -c 98304 \
-              --temp 0.6 \
+              -fit off \
-              --top-p 0.95 \
+              -ctk q8_0 \
-              --reasoning-budget 0 \
+              -ctv q8_0 \
-              -dev CUDA0
-          '';
-        };
-
-        # https://huggingface.co/unsloth/ERNIE-4.5-21B-A3B-PT-GGUF/tree/main
-        "ernie4.5-21b-instruct" = {
-          name = "ERNIE4.5 (21B) - Instruct";
-          cmd = ''
-            ${pkgs.reichard.llama-cpp}/bin/llama-server \
-              --port ''${PORT} \
-              -m /mnt/ssd/Models/ERNIE-4.5-21B-A3B-PT-UD-Q4_K_XL.gguf \
-              -c 98304 \
-              --temp 0.7 \
-              --min-p 0.0 \
-              --top-p 0.8 \
-              --top-k 20
-          '';
-        };
-
-        # https://huggingface.co/mradermacher/OLMoE-1B-7B-0125-Instruct-GGUF/tree/main
-        "olmoe-7b-instruct" = {
-          name = "OLMoE (7B) - Instruct";
-          cmd = ''
-            ${pkgs.reichard.llama-cpp}/bin/llama-server \
-              --port ''${PORT} \
-              -m /mnt/ssd/Models/OLMoE-1B-7B-0125-Instruct.Q8_0.gguf \
-              -dev CUDA1
-          '';
-        };
-
-        # https://huggingface.co/gabriellarson/Phi-mini-MoE-instruct-GGUF/tree/main
-        "phi-mini-8b-instruct" = {
-          name = "Phi mini (8B) - Instruct";
-          cmd = ''
-            ${pkgs.reichard.llama-cpp}/bin/llama-server \
-              --port ''${PORT} \
-              -m /mnt/ssd/Models/Phi-mini-MoE-instruct-Q8_0.gguf \
-              --repeat-penalty 1.05 \
-              --temp 0.0 \
-              --top-p 1.0 \
-              --top-k 1 \
               -dev CUDA1
           '';
         };