tool: automatically download llama.cpp and model files; run chat completions server (win, cuda)

DeepSeek-V2-Chat.q2_k.gguf/download_and_run.bat

@@ -0,0 +1,32 @@
+@echo off
+setlocal
+
+:: Create and change to the directory
+mkdir DeepSeek-V2-Chat.Q2_K.gguf
+cd DeepSeek-V2-Chat.Q2_K.gguf
+
+:: Download the GGUF files
+for %%i in (1 2 3 4 5) do (
+    powershell -command "Invoke-WebRequest -Uri 'https://huggingface.co/leafspark/DeepSeek-V2-Chat-GGUF/resolve/main/DeepSeek-V2-Chat.q2_k.gguf/DeepSeek-V2-Chat.Q2_K-0000%%i-of-00005.gguf?download=true' -OutFile DeepSeek-V2-Chat.Q2_K-0000%%i-of-00005.gguf"
+)
+
+:: Download the llama.cpp binaries
+powershell -command "Invoke-WebRequest -Uri 'https://github.com/ggerganov/llama.cpp/releases/download/b2961/llama-b2961-bin-win-avx2-x64.zip' -OutFile llama-b2961-bin-win-avx2-x64.zip"
+powershell -command "Expand-Archive -Path llama-b2961-bin-win-avx2-x64.zip -DestinationPath ."
+
+:: Check for GPU and download additional files if needed
+powershell -Command "if (Get-WmiObject -Query 'select * from Win32_VideoController where Description like "NVIDIA%"') {
+    Invoke-WebRequest -Uri 'https://github.com/ggerganov/llama.cpp/releases/download/b2961/cudart-llama-bin-win-cu11.7.1-x64.zip' -OutFile cudart-llama-bin-win-cu11.7.1-x64.zip
+    Invoke-WebRequest -Uri 'https://github.com/ggerganov/llama.cpp/releases/download/b2961/llama-b2961-bin-win-cuda-cu11.7.1-x64.zip' -OutFile llama-b2961-bin-win-cuda-cu11.7.1-x64.zip
+    Expand-Archive -Path cudart-llama-bin-win-cu11.7.1-x64.zip -DestinationPath .
+    Expand-Archive -Path llama-b2961-bin-win-cuda-cu11.7.1-x64.zip -DestinationPath .
+}"
+
+:: Execute the server command
+server.exe ^
+  -m DeepSeek-V2-Chat.q2_k.gguf ^
+  -c 4096 ^
+  -i ^
+  --mlock
+
+endlocal