Finalizing v0 evaluation script

[ronentk]

@ShaRefOh could you please add the items here? Having a confusion matrix would also be great!

[ShaRefOh]

I am sure there could be many more improvements. This is the first version and a place to start optimizing.

Important changes:

• [x] move from 'testing' project. -> evaluation
• [x] Customize path to input dataset artifacts
• [x] Add accuracy

Future improvements:

• [x] Confusion matrix
• [ ] Add accuracy per prediction to the evaluation artifact table output
• [x] Separate the evaluation artifact creation into a function in the script
• [x] Import the following functions from a different module when it exists: load_config, init_model. Right now, it is imported from the streamlit
• [x] Re-think zero-divisors
• [x] use tqdm for predictions

[ronentk]

Addition from our discussion -

• accuracy can be split into instance-level accuracy (one instance may have multiple tags) and treating each tag as a separate instance

[ronentk]

@ShaRefOh see this commit https://github.com/csensemakers/desci-sense/commit/a5652568849c9ea9bb4b3f53f8d662f7c7f5414e - moved the code as discussed