rdk_robot_dev_agent/knowledge_hub/视觉语言模型部署流程.md

视觉语言模型部署流程：RDK X5

1. 下载功能包 sudo apt update sudo apt install tros-humble-hobot-llamacpp

2. 系统配置 手动使用命令srpi-config修改ION memory大小为1.9GB，设置方法参考RDK用户手册配置工具srpi-config使用指南Performance Options章节。 重启后设置CPU最高频率为1.5GHz，以及调度模式为performance，命令如下： sudo bash -c 'echo 1 > /sys/devices/system/cpu/cpufreq/boost' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu1/cpufreq/scaling_governor' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu2/cpufreq/scaling_governor' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu3/cpufreq/scaling_governor' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu4/cpufreq/scaling_governor' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu5/cpufreq/scaling_governor' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu6/cpufreq/scaling_governor'

3. 下载模型文件到运行路径 wget https://hf-mirror.com/D-Robotics/InternVL2_5-1B-GGUF-BPU/resolve/main/Qwen2.5-0.5B-Instruct-Q4_0.gguf wget https://hf-mirror.com/D-Robotics/InternVL2_5-1B-GGUF-BPU/resolve/main/rdkx5/vit_model_int16_v2.bin

4. 启动程序 source /opt/tros/humble/setup.bash cp -r /opt/tros/${TROS_DISTRO}/lib/hobot_llamacpp/config/ . ros2 run hobot_llamacpp hobot_llamacpp --ros-args -p feed_type:=0 -p image:=config/image2.jpg -p image_type:=0 -p user_prompt:="描述一下这张图片." -p model_file_name:=vit_model_int16_v2.bin -p llm_model_name:=Qwen2.5-0.5B-Instruct-Q4_0.gguf

视觉语言模型部署流程：RDK S100

1. 下载功能包 sudo apt update sudo apt install tros-humble-hobot-llamacpp

2. 系统配置 手动使用命令srpi-config修改ION memory大小为1.9GB，设置方法参考RDK用户手册配置工具srpi-config使用指南Performance Options章节。 重启后设置CPU最高频率为1.5GHz，以及调度模式为performance，命令如下： sudo bash -c 'echo 1 > /sys/devices/system/cpu/cpufreq/boost' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu1/cpufreq/scaling_governor' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu2/cpufreq/scaling_governor' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu3/cpufreq/scaling_governor' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu4/cpufreq/scaling_governor' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu5/cpufreq/scaling_governor' sudo bash -c 'echo performance >/sys/devices/system/cpu/cpu6/cpufreq/scaling_governor'

3. 下载模型文件到运行路径 wget https://hf-mirror.com/D-Robotics/InternVL2_5-1B-GGUF-BPU/resolve/main/Qwen2.5-0.5B-Instruct-Q4_0.gguf wget https://hf-mirror.com/D-Robotics/InternVL2_5-1B-GGUF-BPU/resolve/main/rdks100/vit_model_int16.hbm

4. 启动程序 source /opt/tros/humble/setup.bash cp -r /opt/tros/${TROS_DISTRO}/lib/hobot_llamacpp/config/ . ros2 run hobot_llamacpp hobot_llamacpp --ros-args -p feed_type:=0 -p image:=config/image2.jpg -p image_type:=0 -p user_prompt:="描述一下这张图片." -p model_file_name:=vit_model_int16.hbm -p llm_model_name:=Qwen2.5-0.5B-Instruct-Q4_0.gguf