Dockerize and fix installation on linux

[3ripleM]

Based on this issue on spacy repository I had to add typing_extensions to our libraries in order to setup the project

Issue Page: Pydantic issubclass error for python 3.8 and 3.9

You should add this line in the requirement.txt typing_extensions==4.4.0

[hosseinfani]

@3ripleM @farinamhz no problem. but also put a quick comment in the requirement.txt. Also, update the environment.yml, if needed. Tnx.

[3ripleM]

I also dockerize the setup process so its now easy to reproduce on any machine :) and also we can publish the docker image to a private container registry to make it available for anyone who wants to run the codebase (with all the requirements already installed)

something like this:

docker run fani_lab/lady --volume ./output:/app/output \
    python main.py \
    -naspects 5 \
    -am rnd \
    -data ../data/raw/semeval/toy.2016SB5/ABSA16_Restaurants_Train_SB1_v2.xml \
    -output ../output/toy.2016SB5/

[farinamhz]

Thank you very much @3ripleM and welcome to LADy :)

[hosseinfani]

@3ripleM how can we have a container registry? why not a public container registry?

[3ripleM]

@hosseinfani Yeah, we can have a public one. It was just only a suggestion. but in either case it would be great to have one.

For setting it up I think we can use github registry (packages) for that. we have two options:

1. Manually push to the registry (which means that we have to make the image on a local computer)
2. Automatic build process (which relies on github action to do image building and pushing for us)

for the second option we can decide to do it automatically for us when we push to the master branch or we can do it manually trigger the pipeline via a button

[hosseinfani]

@3ripleM can we go with the second option and automatic build, but also running a quick test on a toy dataset? if so, would you please handle that?

[3ripleM]

@hosseinfani Yes, I believe we can build on multi stage, means when the build is finished we can test our image by running it on the cloud to test that every thing is working. After that, we can push the image to the registry.

I'll keep you updated

[3ripleM]

@hosseinfani Hossein, our current method of processing toy datasets consumes a significant amount of time and resources.

I would like to propose a solution to address this issue. We could consider releasing two versions of our model:

1. Full Version: This version would require the use of a separate system equipped with a GPU.
2. Lite Version

Furthermore, I am of the opinion that we could explore the possibility of setting up a GitHub Actions runner on our servers that have GPUs. This approach could potentially streamline and automate the process. However, further investigation is necessary to determine the feasibility of this solution.

[hosseinfani]

@3ripleM

• really? it's a toy dataset. it shouldn't take a lot of time. let's discuss it in person.
• we don't have any internal GPU server, nor we should think of having that. We need to provide a unit test on any pcs.

[hosseinfani]

@3ripleM Please prepare two guideline as a md file for our library: 1- From producer's perspective to create dockerization 2- From the consumer perspective how to use the docker

[3ripleM]

@hosseinfani @farinamhz

Dockerization is done and the readme is updated.

[farinamhz]

Great, thank you very much, @3ripleM!

[hosseinfani]

@3ripleM thank you. pls close the issue then.