ValueError: [E143] Labels for component 'trainable_lemmatizer' not initialized in Spacy 3.6

[Norky101]

Hi,

I am using a config created from the Spacy page (selected my component preferences, config init fill command, debug config )
Everything works great.

When I run train spacy like so, (including without the extra --code parameters):

 python -m spacy train 'config.cfg' --output 'model\' --code 'sentence.py' --code 'matcher.py'

I am getting the error:

ValueError: [E143] Labels for component 'trainable_lemmatizer' not initialized. This can be fixed by calling add_label, or by providing a representative batch of examples to the component's `initialize` method.

I found this discussion and have followed the steps (run train), but it does not seem to have worked. Perhaps because the post is on an older version of Spacy.

From my understanding, spacy should fill these needed data files once it has looped through the training data, the same way ner labels are filled once it has looped over the training data annotations.

In addition, my 'sentence.py' and 'matcher.py' work fine as I have used them recently in another config file.

By setting changing this:

[components.trainable_lemmatizer]
factory = "trainable_lemmatizer"
backoff = "orth"
min_tree_freq = 3
overwrite = false
scorer = {"@scorers":"spacy.lemmatizer_scorer.v1"}
top_k = 1

to this:

[components.trainable_lemmatizer]
factory = "lemmatizer"
overwrite = false
scorer = {"@scorers":"spacy.lemmatizer_scorer.v1"}

It does remove the original warning, then once running train, it gives this warning:

Token indices sequence length is longer than the specified maximum sequence length for this model (757 > 512). Running this sequence through the model will result in indexing errors

then this error:

ValueError: [E143] Labels for component 'tagger' not initialized. This can be fixed by calling add_label, or by providing a representative batch of examples to the component's `initialize` method.

• I cant see what other factories I could use for 'tagger'

This updated error relates to this huggingface/transformers#1791 but seems to be separate from spacy:

Any help is much appreciated! Thank you!

config.cfg:

[paths]
train = "output.spacy"
dev = "output.spacy"
vectors = null
init_tok2vec = null

[system]
gpu_allocator = "pytorch"
seed = 0

[nlp]
lang = "en"
pipeline = ["transformer","ner","parser","trainable_lemmatizer"]
batch_size = 128
disabled = []
before_creation = null
after_creation = null
after_pipeline_creation = null
tokenizer = {"@tokenizers":"spacy.Tokenizer.v1"}

[components]

[components.ner]
factory = "ner"
incorrect_spans_key = null
moves = null
scorer = {"@scorers":"spacy.ner_scorer.v1"}
update_with_oracle_cut_size = 100

[components.ner.model]
@architectures = "spacy.TransitionBasedParser.v2"
state_type = "ner"
extra_state_tokens = false
hidden_width = 64
maxout_pieces = 2
use_upper = false
nO = null

[components.ner.model.tok2vec]
@architectures = "spacy-transformers.TransformerListener.v1"
grad_factor = 1.0
pooling = {"@layers":"reduce_mean.v1"}
upstream = "*"

[components.parser]
factory = "parser"
learn_tokens = false
min_action_freq = 30
moves = null
scorer = {"@scorers":"spacy.parser_scorer.v1"}
update_with_oracle_cut_size = 100

[components.parser.model]
@architectures = "spacy.TransitionBasedParser.v2"
state_type = "parser"
extra_state_tokens = false
hidden_width = 128
maxout_pieces = 3
use_upper = false
nO = null

[components.parser.model.tok2vec]
@architectures = "spacy-transformers.TransformerListener.v1"
grad_factor = 1.0
pooling = {"@layers":"reduce_mean.v1"}
upstream = "*"

[components.trainable_lemmatizer]
factory = "trainable_lemmatizer"
backoff = "orth"
min_tree_freq = 3
overwrite = false
scorer = {"@scorers":"spacy.lemmatizer_scorer.v1"}
top_k = 1

[components.trainable_lemmatizer.model]
@architectures = "spacy.Tagger.v2"
nO = null
normalize = false

[components.trainable_lemmatizer.model.tok2vec]
@architectures = "spacy-transformers.TransformerListener.v1"
grad_factor = 1.0
pooling = {"@layers":"reduce_mean.v1"}
upstream = "*"

[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.v3"
name = "roberta-base"
mixed_precision = false

[components.transformer.model.get_spans]
@span_getters = "spacy-transformers.strided_spans.v1"
window = 128
stride = 96

[components.transformer.model.grad_scaler_config]

[components.transformer.model.tokenizer_config]
use_fast = true

[components.transformer.model.transformer_config]

[corpora]

[corpora.dev]
@readers = "spacy.Corpus.v1"
path = ${paths.dev}
max_length = 0
gold_preproc = false
limit = 0
augmenter = null

[corpora.train]
@readers = "spacy.Corpus.v1"
path = ${paths.train}
max_length = 0
gold_preproc = false
limit = 0
augmenter = null

[training]
accumulate_gradient = 3
dev_corpus = "corpora.dev"
train_corpus = "corpora.train"
seed = ${system.seed}
gpu_allocator = ${system.gpu_allocator}
dropout = 0.1
patience = 1600
max_epochs = 0
max_steps = 20000
eval_frequency = 200
frozen_components = []
annotating_components = []
before_to_disk = null
before_update = null

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

[training.logger]
@loggers = "spacy.ConsoleLogger.v1"
progress_bar = false

[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 = 0.00005

[training.score_weights]
ents_f = 0.33
ents_p = 0.0
ents_r = 0.0
ents_per_type = null
dep_uas = 0.17
dep_las = 0.17
dep_las_per_type = null
sents_p = null
sents_r = null
sents_f = 0.0
lemma_acc = 0.33

[pretraining]

[initialize]
vectors = ${paths.vectors}
init_tok2vec = ${paths.init_tok2vec}
vocab_data = null
lookups = null
before_init = null
after_init = null

[initialize.components]

[initialize.tokenizer]

Using:
Python 3.11
Spacy 3.6
Models 3.6
Pycharm

[rmitsch]

Hi @Norky101! Are lemmas available in your training data ("output.spacy")?

[Norky101]

I ended up solving error by doing:

[components.tagger]
source = "en_core_web_lg"
replace_listeners = ["model.tok2vec"]

and this seems to have worked for me so far.

However, the errors I am facing now when I call train include:

RuntimeError: [E896] There was an error using the static vectors. Ensure that the vectors of the vocab are properly initialized, or set 'include_static_vectors' to False.

and

ValueError: Shape mismatch for blis.gemm: (1674, 0), (300, 96)

The first error, I did include_static_vectors= false inside of the [components.tagger.model] but the error persists.

These seem to relate to under the hood issues potentially?

Thanks very much for reaching out!

[Norky101]

I think I solved this error by doing:

[paths]
train = "output.spacy"
dev = "output.spacy"
vectors = "en_core_web_lg"    # This being the new part 
init_tok2vec = null

When running python -m spacy train 'config.cfg' --output 'model\' again, I ran into:

Token indices sequence length is longer than the specified maximum sequence length for this model (757 > 512). Running this sequence through the model will result in indexing errors

and

 did not find POS annotation for one or more tokens. Check that your pipeline includes components that assign token.pos, typically 'tagger'+'attribute_ruler' or 'morphologizer'.

The first error seems to be relating to the under the hood huggingface model

In addition,
when I run python -m spacy debug config config.cfg --code sentence.py --code matcher.py

I get:

Token indices sequence length is longer than the specified maximum sequence length for this model (947 > 512). Running this sequence through the model will result in indexing errors

This feels like a perpetual door of errors and headaches.

[rmitsch]

Let's take a step back. Do the documents in your dataset contain lemmas?