鲲鹏服务器+昇腾卡(Atlas 300I pro)搭建DeepSeek-R1-Distill-Qwen-7B
操作如下:
1、下载mindie:1.0.0-300I-Duo-py311-openeuler24.03-lts镜像,进入镜像后npu-smi info 能正常显示NPU卡,torch,torch_npu,都安装成功
2、下载模型DeepSeek-R1-Distill-Qwen-7B,修改模型权重config.json中torch_dtype字段为float16,修改/usr/local/Ascend/mindie/latest/mindie-service/conf/config.json配置,
npuDeviceIds:[0,1]
truncation:false
modelName:qwen2.5
modelWeightPath:/data/qpqw7b
worldSize:2
httpsEnabled" : false
进入/usr/local/Ascend/mindie/latest/mindie-service/bin
3、运行./mindieservice_daemon出现错误提示:
LLMInferEngine failed to init LLMInferModels
ERR: Failed to init endpoint! Please check the service log or console output.
4、请问怎么能够运行成功,atlas300i pro 卡都支持哪些大模型和镜像,请说明一下,或者提供详细的操作文档。
5、出现问题页面的整屏截图或者日志信息;

vim /data/dpqw7b/config.json

vim /usr/local/Ascend/mindie/latest/mindie-service/conf/config.json


运行./mindieservice_daemon 出现错误提示

查看日志vim /usr/local/Ascend/mindie/latest/mindie-service/logs/mindservice.log

鲲鹏服务器+昇腾卡(Atlas 300I pro)搭建DeepSeek-R1-Distill-Qwen-7B
操作如下:
1、下载mindie:1.0.0-300I-Duo-py311-openeuler24.03-lts镜像,进入镜像后npu-smi info 能正常显示NPU卡,torch,torch_npu,都安装成功
2、下载模型DeepSeek-R1-Distill-Qwen-7B,修改模型权重config.json中torch_dtype字段为float16,修改/usr/local/Ascend/mindie/latest/mindie-service/conf/config.json配置,
npuDeviceIds:[0,1]
truncation:false
modelName:qwen2.5
modelWeightPath:/data/qpqw7b
worldSize:2
httpsEnabled" : false
进入/usr/local/Ascend/mindie/latest/mindie-service/bin
3、运行./mindieservice_daemon出现错误提示:
LLMInferEngine failed to init LLMInferModels
ERR: Failed to init endpoint! Please check the service log or console output.
4、请问怎么能够运行成功,atlas300i pro 卡都支持哪些大模型和镜像,请说明一下,或者提供详细的操作文档。
5、出现问题页面的整屏截图或者日志信息;
vim /data/dpqw7b/config.json
vim /usr/local/Ascend/mindie/latest/mindie-service/conf/config.json
运行./mindieservice_daemon 出现错误提示
查看日志vim /usr/local/Ascend/mindie/latest/mindie-service/logs/mindservice.log