Initial new API

[saeliddp]

What does this PR do?

This PR creates the groundwork for the new API. Note that this is definitely still a draft, and I'm mainly making this PR to get some feedback and ideas. The main principles I followed in creating the new API are below:

1. each im2im experiment (configs/experiments/im2im) will get its own model class
2. these experiment model classes will inherit from a generic model class
3. the model classes train and predict methods will take as arguments all of the config items which are not reasonable to set a default for (e.g. checkpoint path)
4. For config items which seem to have a reasonable default (e.g. the manifest column names), expose setters and getters

So, as an example use of the new API, we would have something like:

# train
model = SegmentationPluginModel() # uses default config
model.set_split_column("split")
model.train(5, Path("example/train.csv"), Path("example/output_dir"))
model.save_config(Path("example/config.yaml"))
...
# predict
model = SegmentationPluginModel(config_path=Path("example/config.yaml"))
model.predict(Path("example/predict.csv"), Path("example/output_dir"), Path("example/ckpt/last.ckpt"))

Before submitting

• Did you make sure title is self-explanatory and the description concisely explains the PR?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you list all the breaking changes introduced by this pull request?
• Did you test your PR locally with pytest command?
• Did you run pre-commit hooks with pre-commit run -a command?

Did you have fun?

Make sure you had fun coding 🙃

[benjijamorris]

love it!

[benjijamorris]

@yrkim98 @hughes036 bumping this, once it's reviewed I can incorporate all of our recent fixed into main.

[hughes036]

should ckpt_path value be MUST_OVERRIDE ?

[saeliddp]

ckpt_path is not necessary for training, so I didn't want to use the MUST_OVERRIDE convention here

[hughes036]

oh, the comment makes sense then.