1
1
1
AI人工智能
用Singularity启用JAX镜像时import报错
2024-08-14 21:24:04
86

我在自己的服务器上build了镜像:

singularity pull --arch amd64 docker://image.sourcefind.cn:5000/dcu/admin/base/vscode-jax:0.4.23-ubuntu20.04-dtk24.04.1-py3.10

目的是想基于JAX在DCU硬件上开发。然而我scp到集群上之后,如下操作测试时遇到了报错:

$ module load singularity/3.8.0
$ singularity shell --bind /var/log/user-log:/var/log/user-log vscode-jax_0.4.23-ubuntu20.04-dtk24.04.1-py3.10.sif
Singularity> python
Python 3.10.8 (main, Nov 24 2022, 14:13:03) [GCC 11.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import jax
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/opt/conda/lib/python3.10/site-packages/jax/__init__.py", line 39, in <module>
    from jax import config as _config_module
  File "/opt/conda/lib/python3.10/site-packages/jax/config.py", line 15, in <module>
    from jax._src.config import config as _deprecated_config  # noqa: F401
  File "/opt/conda/lib/python3.10/site-packages/jax/_src/config.py", line 28, in <module>
    from jax._src import lib
  File "/opt/conda/lib/python3.10/site-packages/jax/_src/lib/__init__.py", line 82, in <module>
    import jaxlib.cpu_feature_guard as cpu_feature_guard
ImportError: libhsa-runtime64.so.1: cannot open shared object file: No such file or directory
>>>

请问如何解决呢?我自行搜索了下问题,但是看不太懂。
https://github.com/ROCm/ROCm/issues/1852
https://github.com/ROCm/ROCm/issues/1581

全部评论(4)
只看楼主
D
dcu_fae_kangle
1楼

报错原因是:未导入dtk的环境变量

DTK下载路径:https://download.sourcefind.cn:65024/1/main

执行步骤:

1、解压dtk

2、进入dtk目录

3、source env.sh

2024/8/15
土
土豆
回复 @dcu_fae_kangle

你好,我按照做了。import jax 确实成功了,但是还是识别不到GPU,考斯是缺 hydcu amdgpu module?

FYI: 如下方式打开的镜像

$ singularity shell --bind /var/log/user-log:/var/log/user-log --bind /opt/hyhal/lib:/opt/hyhal/lib vscode-jax_0.4.23-ubuntu20.04-dtk24.04.1-py3.10.sif

2024/8/16
土
土豆
回复 @dcu_fae_kangle
已经通过测试!
2024/8/16
网
网友l6jnR7
回复 @dcu_fae_kangle
请问是按照上述的方法解决的吗
2024/9/27
没有更多内容了
--
楼主
Level.4
0 帖子
楼主热帖
暂无数据