华为910b4-1显卡启动报错：Call aclInit(nullptr) failed : 507008 at file /work/PaddleCustomDevice/backends/npu/runtime/runtime.cc line 403

[JackMacs]

bug描述 Describe the Bug

docker run -itd --name intention_embedding
--hostname base-model-intention-embedding-1
--privileged --network=qll-net --shm-size=128G
-v /usr/local/Ascend/driver:/usr/local/Ascend/driver
-v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi
-v /usr/local/dcmi:/usr/local/dcmi
-v /root/xiaoe_model/intention_embedding/deploy:/home/qll_chat_intention_embedding/deploy
-v /root/xiaoe_model/license:/home/qll_ai/license
-e ASCEND_RT_VISIBLE_DEVICES="0"
swr.cn-east-3.myhuaweicloud.com/weaver-qianliling/deploy:hw_intention_embedding_private
bash

python -c "import paddle_custom_device; paddle_custom_device.npu.version()
报错

ENV [CUSTOM_DEVICE_ROOT]=/usr/local/lib/python3.9/dist-packages/paddle_custom_device
I0411 10:12:08.927596 650 init.cc:146] Try loading custom device libs from: [/usr/local/lib/python3.9/dist-packages/paddle_custom_device]
Call aclInit(nullptr) failed : 507008 at file /work/PaddleCustomDevice/backends/npu/runtime/runtime.cc line 403
EE1001: 2025-04-11-10:12:09.580.593 The argument is invalid.Reason: rtGetDevMsg execute failed, reason=[context pointer null]
Solution: 1.Check the input parameter range of the function. 2.Check the function invocation relationship.
TraceBack (most recent call last):
faultVersion(0x50a) from driver[FUNC:InitSocTypeFrom910BVersion][FILE:runtime.cc][LINE:1327]
[Init][Version]init soc version failed, ret = 507008[FUNC:ReportInnerError][FILE:log_inner.cpp][LINE:145]
ctx is NULL![FUNC:GetDevErrMsg][FILE:api_impl.cc][LINE:4692]
The argument is invalid.Reason: rtGetDevMsg execute failed, reason=[context pointer null]

其他补充信息 Additional Supplementary Information

根据paddle链接（https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/hardware_support/npu/install_cn.html）拉取了镜像，安装了包。启动服务或者import paddle包报错如上。在容器内部执行npu-smi info是正常的。

[xiaoguoguo626807]

这是偶现的问题吧，换个卡试试呢

[xiaoguoguo626807]

之前这个问题升级驱动之后发生率低了，可以试下升级驱动