Breakdown extras

[aCampello]

Description

Fixes #366.

In particular, this reduces the "vanilla" environment to 109MB and the all environment to about 2.1GB.

It makes a bit more complicated to add new functionality. When you import something you need to check which extra it is part of, and throw the correct error message. Maybe in the future we can automate that, but for now it is fine.

The idea is that this will encourage people to use io functionalities, but also potentially EPMC api wrapper, dataset downloaders etc.

How to review:

Check setup.py and README.md and then really just skim through the rest.

Checklist

• Added link to Github issue or Notion card
• Added tests

[codecov-commenter]

Codecov Report

Merging #372 (dfcbf91) into main (2f2eb6c) will decrease coverage by 56.52%.
The diff coverage is 78.44%.

❗ Current head dfcbf91 differs from pull request most recent head 03b32bd. Consider uploading reports for the commit 03b32bd to get more accurate results

@@             Coverage Diff             @@
##             main     #372       +/-   ##
===========================================
- Coverage   84.78%   28.25%   -56.53%     
===========================================
  Files          42       42               
  Lines        2576     2651       +75     
===========================================
- Hits         2184      749     -1435     
- Misses        392     1902     +1510     

Impacted Files	Coverage Δ
wellcomeml/ml/attention.py	37.50% <0.00%> (-25.00%)	⬇️
wellcomeml/utils.py	89.65% <33.33%> (-6.90%)	⬇️
wellcomeml/ml/spacy_entity_linking.py	20.22% <66.66%> (-60.24%)	⬇️
wellcomeml/ml/spacy_knowledge_base.py	21.33% <66.66%> (-48.53%)	⬇️
wellcomeml/spacy/spacy_doc_to_prodigy.py	36.36% <66.66%> (-53.64%)	⬇️
wellcomeml/metrics/ner_classification_report.py	20.68% <71.42%> (-79.32%)	⬇️
wellcomeml/ml/sent2vec_vectorizer.py	36.00% <71.42%> (-39.00%)	⬇️
wellcomeml/ml/vectorizer.py	33.33% <71.42%> (-57.58%)	⬇️
wellcomeml/metrics/f1.py	38.09% <75.00%> (-55.24%)	⬇️
wellcomeml/ml/keras_utils.py	24.41% <75.00%> (-44.63%)	⬇️
... and 46 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2f2eb6c...03b32bd. Read the comment docs.

[ivyleavedtoadflax]

This looks great. I see what you mean about the errors, but i think what you have done looks good 👍

[nsorros]

@aCampello this is super, awesome contributions, much kudos 👍 👍 🚀

[ivyleavedtoadflax]

Codecov Report

Merging #372 (dfcbf91) into main (2f2eb6c) will decrease coverage by 56.52%.
The diff coverage is 78.44%.

exclamation Current head dfcbf91 differs from pull request most recent head 03b32bd. Consider uploading reports for the commit 03b32bd to get more accurate results

@@             Coverage Diff             @@
##             main     #372       +/-   ##
===========================================
- Coverage   84.78%   28.25%   -56.53%     
===========================================
  Files          42       42               
  Lines        2576     2651       +75     
===========================================
- Hits         2184      749     -1435     
- Misses        392     1902     +1510     

Impacted Files Coverage Δ
wellcomeml/ml/attention.py 37.50% <0.00%> (-25.00%)
wellcomeml/utils.py 89.65% <33.33%> (-6.90%)
wellcomeml/ml/spacy_entity_linking.py 20.22% <66.66%> (-60.24%)
wellcomeml/ml/spacy_knowledge_base.py 21.33% <66.66%> (-48.53%)
wellcomeml/spacy/spacy_doc_to_prodigy.py 36.36% <66.66%> (-53.64%)
wellcomeml/metrics/ner_classification_report.py 20.68% <71.42%> (-79.32%)
wellcomeml/ml/sent2vec_vectorizer.py 36.00% <71.42%> (-39.00%)
wellcomeml/ml/vectorizer.py 33.33% <71.42%> (-57.58%)
wellcomeml/metrics/f1.py 38.09% <75.00%> (-55.24%)
wellcomeml/ml/keras_utils.py 24.41% <75.00%> (-44.63%)
... and 46 more
Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2f2eb6c...03b32bd. Read the comment docs.

hang on a minute, what happened here? This looked good a minute ago

[aCampello]

Some usual codecov weirdness with the GitHub action that we get from time to time. If we go to the actual link, it tells that it doesn't change much

https://app.codecov.io/gh/wellcometrust/WellcomeML/pulls?page=1&state=open&order=-pullid

[nsorros]

No excuse not to use WellcomeML @ivyleavedtoadflax, let's use it more over at Mantis as well :)