add taskmodules and models from pie-modules

[ArneBinder]

This PR implements #459, i.e., it adds the models and taskmodules implemented originally in pie-modules (except for QA and span-pair based RE, see potential follow-ups below).

• added models:
  • SequenceClassificationModelWithPooler
  • SequencePairSimilarityModelWithPooler
  • SimpleTokenClassificationModel
  • SimpleGenerativeModel
  • SimpleSequenceClassificationModel
  • TokenClassificationModelWithSeq2SeqEncoderAndCrf
• added taskmodules:
  • RETextClassificationWithIndicesTaskModule
  • TextToTextTaskModule
  • LabeledSpanExtractionByTokenClassificationTaskModule
  • PointerNetworkTaskModuleForEnd2EndRE
  • CrossTextBinaryCorefTaskModule

IMPORTANT: This restricts the version of transformers to >=4.35.0,<4.37.0! So, this is breaking.

requires:

• switch to python 3.10 #482
• fix: use documents and annotations directly from pie_documents #499

Additional changes:

• add tabulate, and pytorch-crf to dev dependencies
• set dependence torchmetrics[text] >=1.5, <2 to solve conflicts with nltk (text loads the required additional dependencies and >=1.5 ensures that no deprecated nltk models are loaded. Note that we already use the modern nltk models in pie_documents.document.processing.NltkSentenceSplitter)
• add SpanNotAlignedWithTokenException and get_aligned_token_span to utils.document
• add RequiresMaxInputLength and RequiresTaskmoduleConfig to models.interface

follow-ups:

• make AutoAnnotationPipeline work #502
• [OPTIONAL] add remaining models (SimpleExtractiveQuestionAnsweringModel and SpanTupleClassificationModel)
• [OPTIONAL] add remaining taskmodules (ExtractiveQuestionAnsweringTaskModule, and RESpanPairClassificationTaskModule)

[ArneBinder]

regarding: "investigate why tests fail with missing nltk model error (only on CI, not local!)"

@RainbowRivey did some investigation:

In test run it installs 'torchmetrics (1.4.0.post0)' which is an old version using 'punkt', but we manually install latest NLTK which now supports only 'punkt_tab'
This should be fixed by setting minimal torchmetrics version higher or using older nltk (in torchmetrics deps it is nltk >=3.6, <=3.8.1)
I wonder if torchmetrics could install the proper nltk version itself without manually searching its dependencies...

[codecov]

Codecov Report

❌ Patch coverage is 88.92947% with 394 lines in your changes missing coverage. Please review.
✅ Project coverage is 82.68%. Comparing base (a6bb91d) to head (cf64392).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
...taskmodules/re_text_classification_with_indices.py	88.23%	47 Missing and 31 partials ⚠️
...h_ie/taskmodules/pointer_network_for_end2end_re.py	88.00%	22 Missing and 26 partials ⚠️
...dels/base_models/bart_with_decoder_position_ids.py	83.11%	13 Missing and 13 partials ⚠️
src/pytorch_ie/taskmodules/text_to_text.py	86.17%	11 Missing and 15 partials ⚠️
...ules/pointer_network/annotation_encoder_decoder.py	89.47%	12 Missing and 12 partials ⚠️
...es/metrics/wrapped_metric_with_prepare_function.py	76.31%	9 Missing and 9 partials ⚠️
...h_ie/models/base_models/bart_as_pointer_network.py	89.17%	11 Missing and 6 partials ⚠️
...odels/common/model_with_metrics_from_taskmodule.py	75.00%	9 Missing and 8 partials ⚠️
...h_ie/models/sequence_classification_with_pooler.py	89.44%	9 Missing and 8 partials ⚠️
src/pytorch_ie/taskmodules/common/mixins.py	86.50%	10 Missing and 7 partials ⚠️
... and 16 more

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #498      +/-   ##
==========================================
+ Coverage   76.98%   82.68%   +5.70%     
==========================================
  Files          31       64      +33     
  Lines        1803     5361    +3558     
  Branches      350     1129     +779     
==========================================
+ Hits         1388     4433    +3045     
- Misses        337      675     +338     
- Partials       78      253     +175     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.