Update requirements for tensorflow==2.12.0

[David-davidlxl]

JetPack is compatible with TensorFlow 2.12.0 now, so I updated "requirements_pc.txt" and "README.md" for this new version. (No code needs to be changed, though.)

@davidmallasen you might want to update "requirements.txt" at some point and check whether the installation instructions you wrote for Jetson work exactly the same way for TensorFlow 2.12.0.

[davidmallasen]

Hello @David-davidlxl .

As you correctly point out, JetPack 5.x has been getting upgrades of tensorflow. However, the Jetson Nano as far as I know is stuck with JetPack 4.6.1 which only supports tensorflow 2.7. I've been thinking what to do about this because it doesn't seem they are going to give support to the Nano for the JetPack 5.x and in this case we would end up with two set of versions here: one for the Nano and one for a "standard" desktop. In fact, the chess python package only supports python >= 3.7 and the version in the nano is 3.6.9.

Also, to make sure that we're not breaking any setup when we make these changes, we'd need two things:

• Some testsuite to check that the code runs correctly.
• Some continuous integration (CI) setup that tests the commits/pr both for the "embedded" setup and the "pc" setup.

What are your thoughts on this? I'll merge this PR shortly, but I wanted to discuss all of this beforehand.

[David-davidlxl]

Ahh, I see. That's such a bummer. I like the ideas of adding test suite and setting up continuous integration, but I've just never had any experience with them before. Is the former the prerequisite of the latter? And is the former basically pytest (#15)? I'd help you do these for the standard-desktop version in my free time if you could walk me through the steps and guide me along the way

[davidmallasen]

Great! I don't have a lot of experience using these tools, but I'd be happy to help you along the way if you want to attempt this. Testing and continuous integrations are separate but related concepts.

Testing frameworks can exist on their own, and I would say it's the first step of the process. They allow to automatically (and locally in you own machine) test if the code succeeds in runing a series of code snippets. Correct, this is basically #15, which could then be extended to some basic run of the whole lc2fen flow (to check that it finishes without errors for example).

Continuous integration brings this concept to the git repositories and a collaborative environment. It allows to run a series of tests on a "standard" machine (that could be set up using docker or any similar technology, see #28 ) each time that an action is taken in the repository (e.g. git push, pull request, etc.). For this, GitHub has the packages (to host a docker container for example) and the actions (to run these tests automatically) tools. Since this is a public open-source project, it's all free!

Let me know what you think about this. I'm going to merge this PR and we can continue the discussion in the issues or discussions tab.