Resume Training Through The Command Line

[markFriel]

Hi,

I am trying to finetune the spacy en_core_web_lg model with some additional examples through the command-line interface. I receive this error message when I run the training command.

OSError: [E884] The pipeline could not be initialized because the vectors could not be found at './data/processed/en_vectors/'. If your pipeline was already initialized/trained before, call 'resume_training' instead of 'initialize', or initialize only the components that are new.

I previously stored the vectors in the en_core_web_lg model by using vectors.to_disk() but I'm wondering if this was the correct way to store them when trying to then finetune from the command line.

My config file is below.

[paths]
train = "./data/processed/train_cli_data.spacy"
dev = "./data/processed/dev_cli_data.spacy"
vectors = ./data/processed/en_vectors/
raw = null
init_tok2vec = null
vocab_data = null

[system]
gpu_allocator = null
seed = 0

[nlp]
lang = "en"
pipeline = ["tok2vec","senter","attribute_ruler","ner"]
disabled = ["senter"]
before_creation = null
after_creation = null
after_pipeline_creation = null
batch_size = 256
tokenizer = {"@tokenizers":"spacy.Tokenizer.v1"}

[components]

[components.attribute_ruler]
factory = "attribute_ruler"
validate = false


[components.ner]
source=en_core_web_lg

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

[components.senter]
factory = "senter"

[components.senter.model]
@architectures = "spacy.Tagger.v1"
nO = null

[components.senter.model.tok2vec]
@architectures = "spacy.Tok2Vec.v2"

[components.senter.model.tok2vec.embed]
@architectures = "spacy.MultiHashEmbed.v2"
width = 16
attrs = ["NORM","PREFIX","SUFFIX","SHAPE"]
rows = [1000,500,500,500]
include_static_vectors = true

[components.senter.model.tok2vec.encode]
@architectures = "spacy.MaxoutWindowEncoder.v2"
width = 16
depth = 2
window_size = 1
maxout_pieces = 2

[components.tagger]
factory = "tagger"

[components.tagger.model]
@architectures = "spacy.Tagger.v1"
nO = null

[components.tagger.model.tok2vec]
@architectures = "spacy.Tok2VecListener.v1"
width = ${components.tok2vec.model.encode:width}
upstream = "tok2vec"

[components.tok2vec]
factory = "tok2vec"

[components.tok2vec.model]
@architectures = "spacy.Tok2Vec.v2"

[components.tok2vec.model.embed]
@architectures = "spacy.MultiHashEmbed.v2"
width = ${components.tok2vec.model.encode:width}
attrs = ["NORM","PREFIX","SUFFIX","SHAPE"]
rows = [5000,2500,2500,2500]
include_static_vectors = true

[components.tok2vec.model.encode]
@architectures = "spacy.MaxoutWindowEncoder.v2"
width = 96
depth = 4
window_size = 1
maxout_pieces = 3

[corpora]

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

[corpora.train]
@readers = "spacy.Corpus.v1"
path = ${paths:train}
max_length = 5000
gold_preproc = false
limit = 0


[training]
train_corpus = "corpora.train"
dev_corpus = "corpora.dev"
seed = ${system:seed}
gpu_allocator = ${system:gpu_allocator}
dropout = 0.1
accumulate_gradient = 1
patience = 5000
max_epochs = 0
max_steps = 0
eval_frequency = 1000
frozen_components = []
before_to_disk = null
annotating_components = []

[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 = true
eps = 0.00000001
learn_rate = 0.001

[training.score_weights]
tag_acc = 0.16
dep_uas = 0.0
dep_las = 0.16
dep_las_per_type = null
sents_p = null
sents_r = null
sents_f = 0.02
lemma_acc = 0.33
ents_f = 0.33
ents_p = 0.0
ents_r = 0.0
ents_per_type = null

[pretraining]

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

Any help would be much appreciated.

Thanks

[thomashacker]

Hello markFriel,
You can add the vectors directly by using "en_core_web_lg" instead -> vectors = "en_core_web_lg"

When fine-tuning a model, be aware of the catastrophic forgetting problem (https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting)