ValueError("[E203] If the tok2vec embedding layer is not updated during training, make sure to include it in 'annotating components'")

[NeoKun004]

after the second batch of my data , i m getting this error :
Aborting and saving the final best model. Encountered exception:
ValueError("[E203] If the tok2vec embedding layer is not updated during
training, make sure to include it in 'annotating components'")

my output :
Error: /content/model - No such file or directory.
index batch : 0
epoch : 0
epoch : 1
epoch : 2
epoch : 3
epoch : 4
epoch : 5
epoch : 6
epoch : 7
epoch : 8
epoch : 9
f_score : 0.6586826347305389
✔ Created output directory: /content/model/output
ℹ Saving to output directory: /content/model/output
ℹ Using CPU
ℹ To switch to GPU 0, use the option: --gpu-id 0

=========================== Initializing pipeline ===========================
[2023-07-13 09:28:30,226] [WARNING] [W112] The model specified to use for initial vectors (en_core_web_sm) has no vectors. This is almost certainly a mistake.
WARNING:spacy:[W112] The model specified to use for initial vectors (en_core_web_sm) has no vectors. This is almost certainly a mistake.
✔ Initialized pipeline

============================= Training pipeline =============================
ℹ Pipeline: ['tok2vec', 'ner']
ℹ Initial learn rate: 0.001
E # LOSS TOK2VEC LOSS NER ENTS_F ENTS_P ENTS_R SCORE

---

0 0 0.00 69.53 10.17 7.89 14.29 0.10
16 50 53.74 2167.21 54.55 75.00 42.86 0.55
33 100 10.56 192.28 65.00 68.42 61.90 0.65
50 150 0.00 0.01 70.00 73.68 66.67 0.70
✔ Saved pipeline to output directory
/content/model/output/model-last
index_selected_k_most_informative_documents : {202: 0.1586983323890392, 88: 0.18991682159109857, 84: 0.19565444310182115, 172: 0.20040347767343344, 78: 0.2049782126110353, 236: 0.20923753382027174, 111: 0.21172295024438473, 54: 0.21987526527066636, 114: 0.22812226993688478, 212: 0.2429483873771859, 165: 0.25431208587777293, 55: 0.25711627274050625, 186: 0.2578904935852005, 118: 0.2733757439610342, 83: 0.27877870089441625, 228: 0.2866323902366378, 124: 0.2882972092524785, 218: 0.2888313589428746, 196: 0.2892895183606026, 160: 0.2929971692417957, 36: 0.3039209393249116, 45: 0.317380786937083, 204: 0.3210188284673752, 149: 0.32142007621087004, 32: 0.32399186786160844}
length predicted docs : 25
most informative docs : 25
update unlabeled docs :
before update : 221
index_selected_k_most_informative_documents : [202, 88, 84, 172, 78, 236, 111, 54, 114, 212, 165, 55, 186, 118, 83, 228, 124, 218, 196, 160, 36, 45, 204, 149, 32]
length index_selected_k_most_informative_documents : 25
length index_selected_k_most_informative_documents : 25
after update : 196
index batch : 1
epoch : 0
epoch : 1
epoch : 2
epoch : 3
epoch : 4
epoch : 5
epoch : 6
epoch : 7
epoch : 8
epoch : 9
f_score : 0.6815415821501014
ℹ Saving to output directory: /content/model/output
ℹ Using CPU
ℹ To switch to GPU 0, use the option: --gpu-id 0

=========================== Initializing pipeline ===========================
[2023-07-13 09:28:55,197] [WARNING] [W112] The model specified to use for initial vectors (en_core_web_sm) has no vectors. This is almost certainly a mistake.
WARNING:spacy:[W112] The model specified to use for initial vectors (en_core_web_sm) has no vectors. This is almost certainly a mistake.
✔ Initialized pipeline

============================= Training pipeline =============================
ℹ Pipeline: ['tok2vec', 'ner']
ℹ Initial learn rate: 0.001
E # LOSS TOK2VEC LOSS NER ENTS_F ENTS_P ENTS_R SCORE

---

ERROR:main:Failed to execute task: [E203] If the tok2vec embedding layer is not updated during training, make sure to include it in 'annotating components'
Traceback (most recent call last):
File "", line 179, in handler
model_path = train_model_spacy(training_data_path=output_moodel_path,
File "", line 77, in train_model_spacy
train(config_path, output_path=model_path, overrides={"paths.train": training_data_path+"/training.spacy", "paths.dev": training_data_path+"/testing.spacy"}
File "/usr/local/lib/python3.10/dist-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 "/usr/local/lib/python3.10/dist-packages/spacy/training/loop.py", line 124, in train
raise e
File "/usr/local/lib/python3.10/dist-packages/spacy/training/loop.py", line 107, in train
for batch, info, is_best_checkpoint in training_step_iterator:
File "/usr/local/lib/python3.10/dist-packages/spacy/training/loop.py", line 209, in train_while_improving
nlp.update(
File "/usr/local/lib/python3.10/dist-packages/spacy/language.py", line 1163, in update
proc.update(examples, sgd=None, losses=losses, **component_cfg[name]) # type: ignore
File "spacy/pipeline/transition_parser.pyx", line 405, in spacy.pipeline.transition_parser.Parser.update
File "/usr/local/lib/python3.10/dist-packages/thinc/model.py", line 309, in begin_update
return self._func(self, X, is_train=True)
File "/usr/local/lib/python3.10/dist-packages/spacy/ml/tb_framework.py", line 33, in forward
step_model = ParserStepModel(
File "spacy/ml/parser_model.pyx", line 213, in spacy.ml.parser_model.ParserStepModel.init
File "/usr/local/lib/python3.10/dist-packages/thinc/model.py", line 291, in call
return self._func(self, X, is_train=is_train)
File "/usr/local/lib/python3.10/dist-packages/thinc/layers/chain.py", line 55, in forward
Y, inc_layer_grad = layer(X, is_train=is_train)
File "/usr/local/lib/python3.10/dist-packages/thinc/model.py", line 291, in call
return self._func(self, X, is_train=is_train)
File "/usr/local/lib/python3.10/dist-packages/spacy/pipeline/tok2vec.py", line 292, in forward
raise ValueError(Errors.E203.format(name="tok2vec"))
ValueError: [E203] If the tok2vec embedding layer is not updated during training, make sure to include it in 'annotating components'
⚠ Aborting and saving the final best model. Encountered exception:
ValueError("[E203] If the tok2vec embedding layer is not updated during
training, make sure to include it in 'annotating components'")

[svlandeg]

Hi!

I moved your post to the discussions forum.

From the output, there's a warning that you should pay attention to:

[W112] The model specified to use for initial vectors (en_core_web_sm) has no vectors. This is almost certainly a mistake.

And, from the exception message:

[E203] If the tok2vec embedding layer is not updated during training, make sure to include it in 'annotating components'

annotating_components is a variable in your config file, in the [training] subsection. Did you add "tok2vec" to it? You can find more information on this mechanism in the docs: https://spacy.io/usage/training#annotating-components

It that doesn't resolve things, can you post your full config file and use proper markdown formatting in your posts so they are easier to read on our end?

[parzival1l]

I am training this via a notebook and while using the update function, and it seems to run into the same error "transformer embedding layer is not updated during training, make sure to include it in 'annotating components'", although I've added it.

nlp.update([example], sgd=optimizer, drop=dropout, losses=loss_dict, annotates=["transformer"])

Any solution for the same ?

[svlandeg]

I would highly recommend switching to a command-line approach to train your models, and using the spacy train command, because that command takes care of a lot of additional details that you'll have to code for manually if you want to call nlp.update directly.

That said, the error message in your original post refers to a tok2vec component, not a transformer component. You must have a Tok2VecListener in your pipeline that requires the annotations of a tok2vec component that has not been added to annotates.

[Shreya-Goswami]

Hi, I'm getting the same problem. I've added annotating_components = ["tok2vec", "parser", "ner", "transformer"] in my config.cfg file under [training] subsection but i'm still getting the problem with nlp.rehearse - [E203] If the tok2vec embedding layer is not updated during training, make sure to include it in 'annotating components' I have tried adding just annotating_components = ["tok2vec"] too. It's still giving the same error.

[svlandeg]

Hi @Shreya-Goswami : to be able to follow up better, could you start a new discussion thread with your issue, paste your config, and the full console output of the training command?

[Shreya-Goswami]

Hey @svlandeg sure, I'll start a new discussion but I won't be able to paste the full config due to company privacy policies. I'll show the output that i'm getting.

Created a new discussion here - #13332