Packaging relation_extractor - ValueError: Existing factory.

[biagiodistefano]

Hi!

I have successfully trained my custom ner + relation_extractor starting from this great tutorial.

I tried to package it using

python -m spacy package training/model-best packaged --code scripts/rel_pipe.py,scripts/rel_model.py

So far, so good.

But when I pip install the resulting *.tar.gz and import it with `spacy.load("my_model"), first I get the following error:

...
from . import rel_pipe
  File "/Users/biagio/repos/my-project/venv/lib/python3.10/site-packages/it_my_model/rel_pipe.py", line 17, in <module>
    Doc.set_extension("rel", default={})
  File "spacy/tokens/doc.pyx", line 138, in spacy.tokens.doc.Doc.set_extension
ValueError: [E090] Extension 'rel' already exists on Doc. To overwrite the existing extension, set `force=True` on `Doc.set_extension`.

So first I tried with adding force=True as suggested by the stack. However, then this follows:

...
from . import rel_pipe
  File "/Users/biagio/repos/my_project/venv/lib/python3.10/site-packages/it_my_model/rel_pipe.py", line 39, in <module>
    def make_relation_extractor(
  File "/Users/biagio/repos/my_project/venv/lib/python3.10/site-packages/spacy/language.py", line 486, in add_factory
    raise ValueError(err)
ValueError: [E004] Can't set up pipeline component: a factory for 'relation_extractor' already exists. Existing factory: <function make_relation_extractor at 0x1164c0310>. New factory: <function make_relation_extractor at 0x1164c2c20>

Any advice on how to fix this? Thanks.

[thomashacker]

Hello,
if you installed your model via pip, why are you still importing it locally? from . import rel_pipe
I think this might be the culprit here 😄

[biagiodistefano]

Oh my, thanks! I just forgot to delete the import from when I was developing the thing 😅 I spent so many hours in this for nothing. Thanks!