'Conda doesn't work in Vertex AI deeplearning images
When provisioning a VM on GCS with a single T4 GPU and Tensorflow conda doesn't work correctly. I can't create new virtual environments or install any packages.
$ conda info
active environment : base
active env location : /opt/conda
shell level : 1
user config file : /home/plateu/.condarc
populated config files : /opt/conda/.condarc
conda version : 4.11.0
conda-build version : not installed
python version : 3.7.12.final.0
virtual packages : __cuda=11.2=0
__linux=4.19.0=0
__glibc=2.28=0
__unix=0=0
__archspec=1=x86_64
base environment : /opt/conda (writable)
conda av data dir : /opt/conda/etc/conda
conda av metadata url : None
channel URLs : https://conda.anaconda.org/conda-forge/linux-64
https://conda.anaconda.org/conda-forge/noarch
https://repo.anaconda.com/pkgs/main/linux-64
https://repo.anaconda.com/pkgs/main/noarch
https://repo.anaconda.com/pkgs/r/linux-64
https://repo.anaconda.com/pkgs/r/noarch
package cache : /opt/conda/pkgs
/home/plateu/.conda/pkgs
envs directories : /opt/conda/envs
/home/plateu/.conda/envs
platform : linux-64
user-agent : conda/4.11.0 requests/2.26.0 CPython/3.7.12 Linux/4.19.0-18-cloud-amd64 debian/10 glibc/2.28
UID:GID : 1002:1003
netrc file : None
offline mode : False
Trying out the conda installation
$ conda create --name fm37 python=3.7.12
Collecting package metadata (current_repodata.json): failed
NotWritableError: The current user does not have write permissions to a required path.
path: /opt/conda/pkgs/cache/18414ddb.json
uid: 1002
gid: 1003
If you feel that permissions on this path are set incorrectly, you can manually
change them by executing
$ sudo chown 1002:1003 /opt/conda/pkgs/cache/18414ddb.json
In general, it's not advisable to use 'sudo conda'.
Following the advise from conda:
$ sudo chown 1002:1003 /opt/conda/pkgs/cache/18414ddb.json
chown: cannot access '/opt/conda/pkgs/cache/18414ddb.json': No such file or directory
I really need to be able to create virtual envs in this VM as I SSH into it using VSCode and can't access any ipykernels from vscode unless they are in a conda env.
$ sudo /opt/conda/bin/conda create --name fm37 python=3.7.12
Collecting package metadata (current_repodata.json): done
Solving environment: done
## Package Plan ##
environment location: /opt/conda/envs/fm37
added / updated specs:
- python=3.7.12
..... etc.
So that works but only rudimentary. I get major problems using VSCode and I need to reinstall Tensorflow which was in the base environment, so should also be in any other environment.
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|
