Reputation
Badges 1
64 × Eureka!Thanks @<1523701087100473344:profile|SuccessfulKoala55> @<1523701070390366208:profile|CostlyOstrich36> .
Fixed with the RC.
Not sure I understand the purpose of this.
it meant pip will look for wheels at thus url?
AgitatedDove14
Yes, I'd like to point to a specific binary, which is in a conda environment.
(b.t.w how can I specify the python version on the Task?)
AgitatedDove14
It's still failing.
I updated clearml-agent to 1.2.0rc7 and also:docker_setup_bash_script= [ "export PATH=""/workspace/miniconda/bin:$PATH", "export LOCAL_PYTHON=/workspace/miniconda/bin/python3", "conda activate"])
But the conda activate (base env) returns:CommandNotFoundError: Your shell has not been properly configured to use 'conda activate'.
I noticed that conda ...
Anyway, when I add the binary's path to PATH, it still won't work.
I calltask.set_base_docker(docker_image='my/docker/image', docker_setup_bash_script= ["export PATH=""/workspace/miniconda/bin:$PATH"])
just after Task.init.
On execution, the agent installs all pip requirements with python binary at /usr/bin/python3.6
but eventually the task fails with, /workspace/miniconda/bin/python3: No module named clearml_agent
AgitatedDove14
I'm not sure.
In my case I'm not trying to reproduce a local environment in the agent, but to run a script inside a docker which already has the environment built in.
The environment is conda based.
It didn't work either. Still same error.
conda command was already in PATH and the conda activate
is executed properly, but it prompts to run conda init
(i.e conda wasn't configures at that shell):
` You should consider upgrading via the '/usr/bin/python3 -m pip install --upgrade pip' command.
CommandNotFoundError: Your shell has not been properly configured to use 'conda activate'.
To initialize your shell, run
$ conda init <SHELL_NAME>
Currently supported shells are:
- ba...
AgitatedDove14 .
Note that the actual error is /workspace/miniconda/bin/python3: No module named clearml_agent
since all the packages (including clearml_agent) were already installed by the agent on the default (non conda) python binary.
Absolute in my hard drive?
Oh wow AgitatedDove14 . Appreciate it!
Are you sure it’s just a matter of the python version?
The same experiment script, was working on the exact docker image in the past (with older clearml versions though…).
For example this experiment log:
CostlyOstrich36
I do get errors - failing to launch the clearml images.WARNING: The requested image's platform (linux/amd64) does not match the detected host platform (linux/arm64/v8) and no specific platform was requested standard_init_linux.go:228: exec user process caused: exec format error
Well, after diving into this, it seems like the clearml images were built usin amd64 (on top of amd64 base images...)
SuccessfulKoala55
I'm not sure I understand your suggestion
/mnt/clearml_storage
is the mount point of the NAS on one linux machine.
On macos it would be /Volumes/clearml_storage,
On windows - //NAS/clearml_storage
SuccessfulKoala55
we are using the fileserver which is configured at clearml.conf to a path on a network drive (i.e the NAS) -files_server: file:///mnt/clearml_storage
Because we want all our data to be stored on premises.
SuccessfulKoala55
Were using the community server
SuccessfulKoala55
Is it a rare use case to use a NAS as the fileserver?
What would you suggest?
UnevenDolphin73 Thanks! Ill look into and reach out if needed
Its mounted automatically.
The local .ssh folder is copied to a temp folder which is mounted by the agent in the docker run command -Executing: ['docker', 'run', '-t', ......., '-v', '/tmp/clearml_agent.ssh.fuu4r8ta:/root/.ssh', .... , '--rm', 'nvidia/cuda:10.1-cudnn7-runtime-ubuntu18.04', ....]
Yes you are right.
This is the default docker image from clearml, and I was thinking that the agent will install conda if it's not already there (like it installs pip...) Isn't it?
Yes.
Its kwcoco
This is my conda env export
:
name: clearml
channels:
- defaults
dependencies: - ca-certificates=2021.10.26=hecd8cb5_2
- certifi=2021.10.8=py39hecd8cb5_0
- libcxx=12.0.0=h2f01273_0
- libffi=3.3=hb1e8313_2
- ncurses=6.3=hca72f7f_2
- openssl=1.1.1l=h9ed2024_0
- pip=21.2.4=py39hecd8cb5_0
- python=3.9.7=h88f2d9e_1
- readline=8.1=h9ed2024_0
- setuptools=58.0.4=py39hecd8cb5_0
- sqlite=3.36.0=hce871da_0
- tk=8.6.11=h7bc2e8c_0
- tzdata=2021e=hda174b...
AgitatedDove14 , here's the log
Maybe I should understand better how this actually works.
It's not a rare use case to use conda environments, but install some packages using pip (when they're not available at any conda channel). Will that use case be supported on clearml?
No. But this isn't the only package that was installd using pip inside a conda env.
Compare it with Pandas as an example - was installed using pip nad it's actually instaled inside the worker as well (using pip).