[spaCy Project: Part-of-speech Tagging & Dependency Parsing] Issue with the parser

[kanayer]

I have my own custom-made dataset for Korean with train, dev, and test sets consisting of 17373, 2415, and 2361 documents respectively. I have been using ud_parser_tagger spaCy project template to train my pipeline. The preprocessing step works just fine and outputs .spacy files. However, when I run spacy project run train it initializes the pipeline, starts Training pipeline, and then gives the error: ValueError: Could not find gold transition - see logs above. And the logs are as follows:

============================= Training pipeline =============================
_ Pipeline: ['tok2vec', 'tagger', 'morphologizer', 'parser']
_ Initial learn rate: 0.001
E    #       LOSS TOK2VEC  LOSS TAGGER  LOSS MORPH...  LOSS PARSER  TAG_ACC  POS_ACC  MORPH_ACC  DEP_UAS  DEP_LAS  SENTS_F  SCORE
---  ------  ------------  -----------  -------------  -----------  -------  -------  ---------  -------  -------  -------  ------
('S', 1, 3.0)
('L-dep', 1, 2.0)
('L-nsubj', 1, 2.0)
('L-obl', 1, 2.0)
('L-obj', 1, 1.0)
('L-acl', 1, 2.0)
('L-amod', 1, 2.0)
('L-ccomp', 1, 2.0)
('L-advmod', 1, 2.0)
('L-nmod', 1, 2.0)
('L-conj', 1, 2.0)
('L-advcl', 1, 2.0)
('L-csubj', 1, 2.0)
('R-dep', 1, 4.0)
('B-dep', 1, 1.0)
('B-obl', 1, 1.0)
('B-nsubj', 1, 1.0)
('B-acl', 1, 1.0)
('B-obj', 1, 1.0)
('B-advmod', 1, 1.0)
('B-amod', 1, 1.0)
('B-nmod', 1, 1.0)
('B-advcl', 1, 1.0)
('B-conj', 1, 1.0)
('B-ccomp', 1, 1.0)
('B-csubj', 1, 1.0)
('B-ROOT', 1, 1.0)
('Gold sent starts?', 0, 0)
_ Aborting and saving the final best model. Encountered exception:
ValueError('Could not find gold transition - see logs above.')
Traceback (most recent call last):
  File "/opt/conda/lib/python3.7/runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "/opt/conda/lib/python3.7/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/opt/conda/lib/python3.7/site-packages/spacy/__main__.py", line 4, in <module>
    setup_cli()
  File "/opt/conda/lib/python3.7/site-packages/spacy/cli/_util.py", line 71, in setup_cli
    command(prog_name=COMMAND)
  File "/opt/conda/lib/python3.7/site-packages/click/core.py", line 1128, in __call__
    return self.main(*args, **kwargs)
  File "/opt/conda/lib/python3.7/site-packages/click/core.py", line 1053, in main
    rv = self.invoke(ctx)
  File "/opt/conda/lib/python3.7/site-packages/click/core.py", line 1659, in invoke
    return _process_result(sub_ctx.command.invoke(sub_ctx))
  File "/opt/conda/lib/python3.7/site-packages/click/core.py", line 1395, in invoke
    return ctx.invoke(self.callback, **ctx.params)
  File "/opt/conda/lib/python3.7/site-packages/click/core.py", line 754, in invoke
    return __callback(*args, **kwargs)
  File "/opt/conda/lib/python3.7/site-packages/typer/main.py", line 500, in wrapper
    return callback(**use_params)  # type: ignore
  File "/opt/conda/lib/python3.7/site-packages/spacy/cli/train.py", line 45, in train_cli
    train(config_path, output_path, use_gpu=use_gpu, overrides=overrides)
  File "/opt/conda/lib/python3.7/site-packages/spacy/cli/train.py", line 75, in train
    train_nlp(nlp, output_path, use_gpu=use_gpu, stdout=sys.stdout, stderr=sys.stderr)
  File "/opt/conda/lib/python3.7/site-packages/spacy/training/loop.py", line 122, in train
    raise e
  File "/opt/conda/lib/python3.7/site-packages/spacy/training/loop.py", line 105, in train
    for batch, info, is_best_checkpoint in training_step_iterator:
  File "/opt/conda/lib/python3.7/site-packages/spacy/training/loop.py", line 209, in train_while_improving
    annotates=annotating_components,
  File "/opt/conda/lib/python3.7/site-packages/spacy/language.py", line 1153, in update
    proc.update(examples, sgd=None, losses=losses, **component_cfg[name])  # type: ignore
  File "spacy/pipeline/transition_parser.pyx", line 406, in spacy.pipeline.transition_parser.Parser.update
  File "spacy/pipeline/transition_parser.pyx", line 518, in spacy.pipeline.transition_parser.Parser.get_batch_loss
  File "spacy/pipeline/_parser_internals/arc_eager.pyx", line 827, in spacy.pipeline._parser_internals.arc_eager.ArcEager.set_costs
ValueError: Could not find gold transition - see logs above.

Could you please help me figure out what is wrong and what can I do to solve the issue?

[svlandeg]

This error usually indicates an issue with the training data. Do there happen to be dependency labels in your dataset that are very infrequent? Also, could you run spacy debug data and see whether anything pops up?