ocelot icon indicating copy to clipboard operation
ocelot copied to clipboard

Conda install issues

Open irgallard opened this issue 1 year ago • 7 comments

Hello,

I am having issues installing Ocelot via Conda. This is what I get: image

Hopefully there is a solution, since Conda is very easy to manage. Will try the repo install now.

irgallard avatar Jul 19 '23 09:07 irgallard

Hi @irgallard, it is better to downloaded Ocelot from from GitHub zip file and start using it directly (without any installation).

Conda installation is handy, however, working with the downloading folder from the GitHub zip file is equally simple. Just don't forget to do: export PYTHONPATH=/your_working_dir/ocelot-master:$PYTHONPATH

dmargarone avatar Jul 19 '23 11:07 dmargarone

You can always do pip install ~/path/to/ocelot or pip install --editable ~/path/to/ocelot after cloning the repo. This is generally easier than playing with PYTHONPATH (I have not touched PYTHONPATH in many many years). This way also takes care of any dependencies/setup as well.

I don't know why the conda recipe doesn't work. That is @sergey-tomin's domain.

st-walker avatar Jul 19 '23 19:07 st-walker

Thank you @dmargarone @st-walker. I managed to run it downloading the repo, so far it works. One last thing, any plans to release an Ocelot docker image?

irgallard avatar Jul 20 '23 07:07 irgallard

It is not hard to do but I am curious what you would have in mind for it?

st-walker avatar Jul 20 '23 09:07 st-walker

@st-walker Regarding the Docker image, I think it would be nice to have a portable, isolated and compatible version of the simulator that can be deployed anywhere. Many applications now allow you to install with Docker for its simplicity and in case the other installation methods fail.

irgallard avatar Jul 20 '23 11:07 irgallard

Hi @irgallard, I understand the benefits of Docker, I am just curious where and how you are wanting to use ocelot but can't because there's no docker image.

st-walker avatar Jul 21 '23 09:07 st-walker

@st-walker Thanks for the replies!

The end goal of our work is to study Reinforcement Learning algorithms applied to accelerator control, using Ocelot as environment and Kubeflow for managing the Machine Learning pipelines. I am not an expert on Kubeflow, but as far as I understood, each element of the pipeline is typically encapsulated as a container.

So far so good though, Ocelot works just fine on my machine and I am still learning to use it.

irgallard avatar Jul 21 '23 09:07 irgallard