Cannot get dimension 'nI' for model 'linear': value unset

[matteoarcangeli99]

Hello,
I'm trying to build a pipeline with the textcat_multilabel component that use Tok2VecTransformer.v3 as token2vec

This is my configuration file

[paths]
train = train.spacy
dev = dev.spacy

[system]
gpu_allocator = null
seed = 0

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

[components]

[components.textcat_multilabel]
factory = "textcat_multilabel"
scorer = {"@scorers":"spacy.textcat_multilabel_scorer.v1"}
threshold = 0.5

[components.textcat_multilabel.model]
@architectures = "spacy.TextCatCNN.v2"
exclusive_classes = false
nO = 7

[components.textcat_multilabel.model.tok2vec]
@architectures = "spacy-transformers.Tok2VecTransformer.v3"
name = "bert-base-german-cased"
tokenizer_config = {"use_fast": false}
transformer_config = {}
grad_factor = 1.0

[components.textcat_multilabel.model.tok2vec.get_spans]
@span_getters = "spacy-transformers.strided_spans.v1"
window = 128
stride = 96


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

[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]
dev_corpus = "corpora.dev"
train_corpus = "corpora.train"
seed = ${system.seed}
gpu_allocator = ${system.gpu_allocator}
dropout = 0.1
accumulate_gradient = 1
patience = 1600
max_epochs = 0
max_steps = 20000
eval_frequency = 200
frozen_components = []
annotating_components = []
before_to_disk = null

[training.batcher]
@batchers = "spacy.batch_by_words.v1"
discard_oversize = false
tolerance = 0.2
get_length = null

[training.batcher.size]
@schedules = "compounding.v1"
start = 100
stop = 1000
compound = 1.001
t = 0.0

[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
learn_rate = 0.001

[training.score_weights]
cats_score = 1.0
cats_score_desc = null
cats_micro_p = null
cats_micro_r = null
cats_micro_f = null
cats_macro_p = null
cats_macro_r = null
cats_macro_f = null
cats_macro_auc = null
cats_f_per_type = null
cats_macro_auc_per_type = null

[pretraining]

[initialize]
vectors = null
init_tok2vec = null
vocab_data = null
lookups = null
before_init = null
after_init = null

[initialize.components]

[initialize.tokenizer] 

but I obtain this error

=========================== Initializing pipeline ===========================
[2022-12-04 16:41:53,517] [INFO] Set up nlp object from config
[2022-12-04 16:41:53,521] [INFO] Pipeline: ['textcat_multilabel']
[2022-12-04 16:41:53,523] [INFO] Created vocabulary
[2022-12-04 16:41:53,524] [INFO] Finished initializing nlp object
Traceback (most recent call last):
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/runpy.py", line 197, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/spacy/__main__.py", line 4, in <module>
    setup_cli()
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/spacy/cli/_util.py", line 71, in setup_cli
    command(prog_name=COMMAND)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/click/core.py", line 1128, in call
    return self.main(*args, **kwargs)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/click/core.py", line 1053, in main
    rv = self.invoke(ctx)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/click/core.py", line 1659, in invoke
    return _process_result(sub_ctx.command.invoke(sub_ctx))
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/click/core.py", line 1395, in invoke
    return ctx.invoke(self.callback, **ctx.params)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/click/core.py", line 754, in invoke
    return __callback(*args, **kwargs)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/typer/main.py", line 500, in wrapper
    return callback(**use_params)  # type: ignore
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/spacy/cli/train.py", line 45, in train_cli
    train(config_path, output_path, use_gpu=use_gpu, overrides=overrides)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/spacy/cli/train.py", line 72, in train
    nlp = init_nlp(config, use_gpu=use_gpu)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/spacy/training/initialize.py", line 84, in init_nlp
    nlp.initialize(lambda: train_corpus(nlp), sgd=optimizer)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/spacy/language.py", line 1317, in initialize
    proc.initialize(get_examples, nlp=self, **p_settings)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/spacy/pipeline/textcat_multilabel.py", line 188, in initialize
    self.add_label(cat)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/spacy/pipeline/textcat.py", line 343, in add_label
    self.model = self.model.attrs["resize_output"](self.model, len(self.labels))
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/spacy/ml/models/textcat.py", line 68, in resize_and_set_ref
    resizable_layer = resize_model(resizable_layer, new_nO)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/thinc/layers/resizable.py", line 45, in resize_model
    new_layer = model.attrs["resize_layer"](old_layer, new_nO)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/thinc/layers/resizable.py", line 77, in resize_linear_weighted
    new_layer.initialize()

File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/thinc/model.py", line 299, in initialize
    self.init(self, X=X, Y=Y)
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/thinc/layers/linear.py", line 57, in init
    model.set_param("W", init_W(model.ops, (model.get_dim("nO"), model.get_dim("nI"))))
  File "/Users/giorgiopaoletti/miniforge3/envs/multilabel_classification_model/lib/python3.9/site-packages/thinc/model.py", line 175, in get_dim
    raise ValueError(err)
ValueError: Cannot get dimension 'nI' for model 'linear': value unset

[polm]

Sorry this wasn't working. I need to look at this in more detail, but I think this might be a bug with the TextCat architecture - if you change TextCatCNN.v2 to TextCatCNN.v1 does it work for you?

[matteoarcangeli99]

Using TextCatCNN.v1 it works but I have worst result compared to TextCatCNN.v2 without trasnfromer.

[polm]

Thanks for confirming it worked at least. I think this is a bug so we'll work on fixing it, but it may take a bit; in the meantime you'd need to stick with v1 or use another architecture.

That said, I am surprised there would be a significant difference in performance - v2 of the CNN was just supposed to add the ability to be resizable.

[polm]

Thanks for your patience with this. I have confirmed this is a bug and made #11968 to track it. I managed to get it working locally by making the modification in explosion/thinc#820, so you might try that, but test any models you train thoroughly, as we haven't tested it for any unwelcome side effects yet.