Labels for component 'relation_extractor' not initialized

[kuxall]

File "spacy/pipeline/pipe.pyx", line 119, in spacy. pipeline.pipe.Pipe._require_labels ValueError: [E143] Labels for component 'relation_extractor' not initialized. This can be fixed by calling add_label, or by providing a representative batch of examples to the component's `initialize` method. 

Having Problem with this as i am trying to extract the relation and show it between the entities.
The config for rel_trf.cfg is as:

[paths]
train = null
dev = null
raw = null
init_tok2vec = null

[system]
seed = 342
gpu_allocator = "pytorch"

[nlp]
lang = "en"
pipeline = ["transformer", "relation_extractor"]
disabled = []
before_creation = null
after_creation = null
after_pipeline_creation = null
tokenizer = {"@tokenizers":"spacy.Tokenizer.v1"}
batch_size = 1000

[components]

[components.transformer]
factory = "transformer"
max_batch_items = 4096
set_extra_annotations = {"@annotation_setters":"spacy-transformers.null_annotation_setter.v1"}

[components.transformer.model]
@architectures = "spacy-transformers.TransformerModel.v1"
name = "roberta-base"
tokenizer_config = {"use_fast": true}

[components.transformer.model.get_spans]
@span_getters = "spacy-transformers.strided_spans.v1"
window = 64
stride = 48

[components.relation_extractor]
factory = "relation_extractor"
threshold = 0.5

[components.relation_extractor.model]
@architectures = "rel_model.v1"

[components.relation_extractor.model.create_instance_tensor]
@architectures = "rel_instance_tensor.v1"

[components.relation_extractor.model.create_instance_tensor.tok2vec]
@architectures = "spacy-transformers.TransformerListener.v1"
grad_factor = 1.0

[components.relation_extractor.model.create_instance_tensor.tok2vec.pooling]
@layers = "reduce_mean.v1"

[components.relation_extractor.model.create_instance_tensor.pooling]
@layers = "reduce_mean.v1"

[components.relation_extractor.model.create_instance_tensor.get_instances]
@misc = "rel_instance_generator.v1"
max_length = 100

[components.relation_extractor.model.classification_layer]
@architectures = "rel_classification_layer.v1"
nI = null
nO = null

[initialize]

[initialize.components]

[corpora]

[corpora.dev]
@readers = "Gold_ents_Corpus.v1"
file = ${paths.dev}

[corpora.train]
@readers = "Gold_ents_Corpus.v1"
file = ${paths.train}

[training]
seed = ${system.seed}
gpu_allocator = ${system.gpu_allocator}
dropout = 0.1
accumulate_gradient = 1
patience = 1600000
max_epochs = 0
max_steps = 1000
eval_frequency = 100
frozen_components = []
dev_corpus = "corpora.dev"
train_corpus = "corpora.train"
before_to_disk = null
logger = {"@loggers":"spacy.ConsoleLogger.v1"}

[training.batcher]
@batchers = "spacy.batch_by_padded.v1"
discard_oversize = true
size = 2000
buffer = 256

[training.optimizer]
@optimizers = "Adam.v1"
beta1 = 0.9
beta2 = 0.999
L2_is_weight_decay = true
L2 = 0.01
grad_clip = 1.0
use_averages = false
eps = 0.00000001

[training.optimizer.learn_rate]
@schedules = "warmup_linear.v1"
warmup_steps = 250
total_steps = 20000
initial_rate = 5e-5

[training.score_weights]
rel_micro_p = 0.0
rel_micro_r = 0.0
rel_micro_f = 1.0

and when i try to run the command !python -m spacy project run train_gpu it shows the following error

File "spacy/pipeline/pipe.pyx", line 119, in spacy.pipeline.pipe.Pipe._require_labels ValueError: [E143] Labels for component 'relation_extractor' not initialized. This can be fixed by calling add_label, or by providing a representative batch of examples to the component's `initialize` method.

Suggestions and guides are highly appreciated. Thank You

[polm]

This usually happens when your training data doesn't include annotations. How are you creating your training data?

Also please read the guide on how to use Markdown code blocks to make your post easier to read.

[kuxall]

To be clear, I annotated data from UBIAI tools with some of the relations in it. and I exported the data in IOB format. Does that mean I exported the data in the wrong model for relation extraction? My parse_data.py from rel_component looks like this:

import typer
from pathlib import Path

from spacy.tokens import Span, DocBin, Doc
from spacy.vocab import Vocab
from wasabi import Printer
from spacy.tokenizer import Tokenizer
from spacy.lang.en import English
from spacy.util import compile_infix_regex
import re
import spacy

nlp = spacy.blank("en")
# Create a blank Tokenizer with just the English vocab

msg = Printer()


SYMM_LABELS = ["Binds"]
MAP_LABELS = {
    "STUDIED_AT": "STUDIED_AT",
    "EXPERIENCE_IN": "EXPERIENCE_IN",
    "EXPERTISE_IN":"EXPERTISE_IN",
    "WORKED_IN":"WORKED_IN",
    "IMPLEMENTED_USING":"IMPLEMENTED_USING"
}

ann = "JointNer/train.json"
train_file='JointNer/rel_component/data/train.spacy'
dev_file='JointNer/rel_component/data/dev.spacy'


def main(json_loc: Path, train_file: Path, dev_file: Path):
    """Creating the corpus from the Prodigy annotations."""
    Doc.set_extension("rel", default={},force=True)
    vocab = Vocab()

    docs = {"train": [], "dev": [], "total": []}
    ids = {"train": set(), "dev": set(), "total":set()}
    count_all = {"train": 0, "dev": 0, "total": 0}
    count_pos = {"train": 0, "dev": 0,"total": 0}

    with open(json_loc, encoding="utf8") as jsonfile:
        file = json.load(jsonfile)
        for example in file:
            span_starts = set()
            neg = 0
            pos = 0
                    # Parse the tokens
            tokens=nlp(example["document"])    

            spaces=[]
            spaces = [True if tok.whitespace_ else False for tok in tokens]
            words = [t.text for t in tokens]
            doc = Doc(nlp.vocab, words=words, spaces=spaces)


            # Parse the entities
            spans = example["tokens"]
            entities = []
            span_end_to_start = {}
            for span in spans:
                entity = doc.char_span(
                     span["start"], span["end"], label=span["entityLabel"]
                 )


                span_end_to_start[span["token_start"]] = span["token_start"]
                #print(span_end_to_start)
                entities.append(entity)
                span_starts.add(span["token_start"])

            doc.ents = entities

            # Parse the relations
            rels = {}
            for x1 in span_starts:
                for x2 in span_starts:
                    rels[(x1, x2)] = {}
                    #print(rels)
            relations = example["relations"]
            #print(len(relations))
            for relation in relations:
                # the 'head' and 'child' annotations refer to the end token in the span
                # but we want the first token
                start = span_end_to_start[relation["head"]]
                end = span_end_to_start[relation["child"]]
                label = relation["relationLabel"]
                #print(rels[(start, end)])
                #print(label)
                #label = MAP_LABELS[label]
                if label not in rels[(start, end)]:
                    rels[(start, end)][label] = 1.0
                    pos += 1
                    #print(pos)
                    #print(rels[(start, end)])

            # The annotation is complete, so fill in zero's where the data is missing
            for x1 in span_starts:
                for x2 in span_starts:
                    for label in MAP_LABELS.values():
                        if label not in rels[(x1, x2)]:
                            neg += 1
                            rels[(x1, x2)][label] = 0.0

                            #print(rels[(x1, x2)])
            doc._.rel = rels
            #print(doc._.rel)

            # only keeping documents with at least 1 positive case
            if pos > 0:
                    docs["total"].append(doc)
                    count_pos["total"] += pos
                    count_all["total"] += pos + neg

                    
                    
    #print(len(docs["total"]))
    docbin = DocBin(docs=docs["total"], store_user_data=True)
    docbin.to_disk(train_file)
    msg.info(
        f"{len(docs['total'])} training sentences"
    )


main(ann, train_file, dev_file)

Thank You

[polm]

I'm not familiar with UBIAI, but IOB data usually just has IOB tags for NER - there's no place to put anything else. Can you give a sample of the data you exported?

UBIAI won't export data in the same format as the script in the example REL project. While you can use the conversion script there as a starting point, you will need to customize it to match the data format.

[kuxall]

Here is the sample of data of IOB:

-DOCSTART- -X- O O
Professional	O
Summary	O
:	O

Over	O
5years	B-EXP
of	O
professional	O
Software	O
Development	O
work	O
Experience	O
with	O
efficient	O
application	O
development	O
and	O
designing	O
knowledge	O
of	O
Python	B-HARD SKILL
and	O
Django	B-HARD SKILL
ability	O
in	O
development	O
of	O
web	O
based	O
applications	O
using	O
Python	O
.	O

[polm]

OK, there is definitely no relation information in that file. You'll need to export in some other format.

[kuxall]

Ok thank you so much, I will definitely be back here with another set of problems if found. Thank you for your constant guidance.