@echo off
chcp 65001 > nul
title Gemma4-26B-A4B-Uncensored @ 128K 极速响应版

:: ========== Colors ==========
set "ESC=["
set "RED=%ESC%91m"
set "GREEN=%ESC%92m"
set "YELLOW=%ESC%93m"
set "BLUE=%ESC%94m"
set "PURPLE=%ESC%95m"
set "CYAN=%ESC%96m"
set "WHITE=%ESC%97m"
set "RESET=%ESC%0m"

:: ========== Paths ==========
set SERVER=D:\Llama Server\llama.cpp\bin\llama-server.exe
set MODEL=D:\Llama Server\llama.cpp\Gemma4-26B-A4B-Uncensored\Gemma4-26B-A4B-Uncensored-HauhauCS-Balanced-Q4_K_M.gguf
set MMPROJ=D:\Llama Server\llama.cpp\Gemma4-26B-A4B-Uncensored\mmproj-Gemma4-26B-A4B-Uncensored-HauhauCS-Balanced-f16.gguf
set SLOTS=D:\Llama Server\llama.cpp\slots

:: ========== File Checks ==========
if not exist "%MODEL%" (echo %RED%[X] Model file missing%RESET% & pause & exit /b 1)
if not exist "%MMPROJ%" (echo %RED%[X] mmproj file missing%RESET% & pause & exit /b 1)
if not exist "%SERVER%" (echo %RED%[X] Server binary missing%RESET% & pause & exit /b 1)
if not exist "%SLOTS%" mkdir "%SLOTS%" 2>nul

cls
echo.
echo %PURPLE% +---------------------------------------------------+%RESET%
echo %PURPLE% ^|%RESET% %GREEN%Gemma4-26B-A4B-Uncensored @ 128K 极速响应版%RESET% %PURPLE%^|%RESET%
echo %PURPLE% ^|%RESET% %CYAN%Q4_K_M ^| no-mmap ^| MoE CPU ^| KV D:/slots%RESET% %PURPLE%^|%RESET%
echo %PURPLE% +---------------------------------------------------+%RESET%
echo.
echo %CYAN% --no-mmap: 模型走虚拟内存 ^| 物理内存由OS动态管理%RESET%
echo %CYAN% --no-warmup ^| 跳过预热 省启动内存%RESET%
echo %CYAN% --n-cpu-moe 999: 全CPU计算 ^| 冷专家在.gguf文件中按需加载%RESET%
echo %CYAN% --cache-ram 0: KV不走内存 ^| 直接写入D盘(同盘免跨盘)%RESET%
echo %CYAN% --host 127.0.0.1: 仅本地访问 ^| 减少网络延迟%RESET%
echo %YELLOW% --ctx-size 200000 ^| 200K上下文%RESET%
echo.

"%SERVER%" ^
 -m "%MODEL%" ^
 --mmproj "%MMPROJ%" ^
 --host 127.0.0.1 --port 8080 ^
 --n-gpu-layers 999   ^
 --n-cpu-moe 999     ^
 --no-warmup          ^
 --no-mmap            ^
 -t 8                 ^
 --ctx-size 200000    ^
 -np 2                ^
 --batch-size 512     ^
 --ubatch-size 256    ^
 --cache-type-k q4_0  ^
 --cache-type-v q4_0  ^
 --cache-ram 0         ^
 --slot-save-path "%SLOTS%" ^
 --rope-scale 2.5     ^
 --flash-attn auto    ^
 --temp 0.28 --top-k 35 --top-p 0.72 --min-p 0.18 --repeat-penalty 1.18 ^
 --repeat-last-n 256  ^
 --alias "Gemma4-26B-A4B-Uncensored" ^
 --timeout 1800       ^
 --ui

pause

Logo

openEuler 是由开放原子开源基金会孵化的全场景开源操作系统项目,面向数字基础设施四大核心场景(服务器、云计算、边缘计算、嵌入式),全面支持 ARM、x86、RISC-V、loongArch、PowerPC、SW-64 等多样性计算架构