77 Star 596 Fork 1.2K

Ascend/pytorch

报错RuntimeError: call aclnnMm failed, detail:EZ9999: Inner Error! EZ9999 Cannot parse json for config file

DONE
需求
创建于  
2023-10-17 18:22

代码:

import torch
import torch_npu

x = torch.randn(2, 2).npu()
y = torch.randn(2, 2).npu()
z = x.mm(y)

print(z)

报错:

RuntimeError: call aclnnMm failed, detail:EZ9999: Inner Error!
EZ9999  Cannot parse json for config file [/usr/local/Ascend/ascend-toolkit/latest/opp/built-in/op_impl/ai_core/tbe//kernel/config/ascend910b/mat_mul.json].
        TraceBack (most recent call last):
        Failed to parse kernel in mat_mul.json.
        AclOpKernelInit failed opType
        Op MatMulV2 does not has any binary.
        Kernel Run failed. opType: 134, MatMulV2
        launch failed for MatMulV2, errno:561000.

评论 (18)

codethedream 创建了需求 2年前

装一下kernel包

具体解释一下可以么,新手很难用啊

Destiny 任务状态TODO 修改为WIP 2年前

我也遇到了。不过安装kernel包后并没有解决。
芯片:910b。
下载kernel网站:https://www.hiascend.com/software/cann/community-history
下载kernel版本:7.0.RC1.alpha005分支下的【Ascend-cann-kernels-910_7.0.RC1_linux.run】
下载toolkit版本:7.0.RC1.alpha005分支下的【Ascend-cann-toolkit_7.0.RC1_linux-aarch64.run】
python:3.10
torch:2.1.0
torch_npu:2.1.0rc1
torch和torch_npu都是pip install安装的,跟readme教程一样。

仍然出现上述报错。

你是910B的芯片,应该装910b的kernel包

另外,我在执行.npu()的时候,会出现大量警告,这些警告上网搜不到原因。

>>> import torch
>>> import torch_npu
>>> x=torch.rand(2,2).npu()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:671: ImportWarning: TBEMetaPathLoader.exec_module() not found; falling back to load_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:914: ImportWarning: TEMetaPathFinder.find_spec() not found; falling back to find_module()
<frozen importlib._bootstrap>:671: ImportWarning: TBEMetaPathLoader.exec_module() not found; falling back to load_module()

回答一下啊,大家都等着用呢,一个矩阵乘法都出这么多bug。只管开源,然后等别人修吗

有结果了吗?

你好,这个不是bug,是因为你没有安装kernel包导致的

https://www.hiascend.com/developer/download/community/result?module=cann
可以根据类似这个社区链接安装,先安装cann的toolkit包,再安装对应你的芯片的kernel包

[E OpParamMaker.cpp:273] call aclnnGroupNorm failed, detail:EZ9999: Inner Error!
910bpro遇到相似的错误,cann toolkit跟kernal包均已安装成功

我也出现了这个错误

请问有解决的方案吗?一样的报错,按照教程尝试安装了8.0.RC1 8.0.RC2 8.0.RC3 的所有toolikit和kernal包,都报这个错误

现在有人解决了吗?

Hello! I also installed kernel packages, but still encounter an error.

Here is my version.cfg contents:

runtime_running_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
compiler_running_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
hccl_running_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
opp_running_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
toolkit_running_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
aoe_running_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
ncs_running_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
opp_kernel_running_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
runtime_upgrade_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
compiler_upgrade_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
hccl_upgrade_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
opp_upgrade_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
toolkit_upgrade_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
aoe_upgrade_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
ncs_upgrade_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
opp_kernel_upgrade_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
runtime_installed_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
compiler_installed_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
hccl_installed_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
opp_installed_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
toolkit_installed_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
aoe_installed_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
ncs_installed_version=[7.5.T11.0.B081:8.0.RC3.alpha003]
opp_kernel_installed_version=[7.5.T11.0.B081:8.0.RC3.alpha003]

我也遇到这个问题,重装cann-toolkit 和 kernel 包 就好了。

类似问题,请安装配套的toolkit和kernel包

huangyunlong 任务状态WIP 修改为DONE 3个月前

登录 后才可以发表评论

状态
负责人
项目
里程碑
Pull Requests
关联的 Pull Requests 被合并后可能会关闭此 issue
分支
优先级
预计工期 (小时)
开始日期   -   截止日期
-
置顶选项
参与者(13)
Destiny-wx1103340 preordi-preordi kaka-kaka4832 11472078 jiangyanshan 1708417110 4760248 wyg1997 1591166559 kli017-kli017 Leo77-nancy_zhouting Tao_gede-tao-gede zsj-geek-bigdata rusanovivan-rusanovivan Road2open-road2open huangyunlong-huangyunlong2022
加载更多
Python
1
https://gitee.com/ascend/pytorch.git
git@gitee.com:ascend/pytorch.git
ascend
pytorch
pytorch

搜索帮助