Troubles encountered to make it work properly, some updates needed

[djowalker]

Due to version mismatch, it seems like tock-flair is no more working.

I use python 3.8 on venv.

Some changes made to progress in installation :

#before
torch==1.3.1
torchvision==0.4.2
numpy==1.18.1
#now
numpy==1.20.0
torch==1.4.0
torchvision==0.5.0

I had to get the models from now from :

curl https://flair.informatik.hu-berlin.de/resources/models/ner-fast/en-ner-fast-conll03-v0.4.pt -o models/en-ner-fast-conll03-v0.4.pt
curl https://flair.informatik.hu-berlin.de/resources/models/fr-ner/fr-ner-wikiner-0.4.pt -o models/fr-ner-wikiner-0.4.pt

When i launch the server i have the following error wich seems to come from the torch package that evolved since 1.3.1 (no more existing and minimum 1.4.0) :

Traceback (most recent call last):
  File "application.py", line 2, in <module>
    from ner import get_flair, evaluate
  File "/home/djowalker/tock/tock-flair/ner.py", line 18, in <module>
    flair_fr = load_flair_fr()
  File "/home/djowalker/tock/tock-flair/ner.py", line 11, in load_flair_fr
    return SequenceTagger.load(path.join(Path.cwd(), 'models/fr-ner-wikiner-0.4.pt'))
  File "/home/djowalker/Arkea/tock/tock-flair/venv/lib/python3.8/site-packages/flair/nn.py", line 86, in load
    model = cls._init_model_with_state_dict(state)
  File "/home/djowalker/tock/tock-flair/venv/lib/python3.8/site-packages/flair/models/sequence_tagger_model.py", line 238, in _init_model_with_state_dict
    model = SequenceTagger(
  File "/home/djowalker/tock/tock-flair/venv/lib/python3.8/site-packages/flair/models/sequence_tagger_model.py", line 201, in __init__
    self.to(flair.device)
  File "/home/djowalker/tock/tock-flair/venv/lib/python3.8/site-packages/torch/nn/modules/module.py", line 443, in to
    return self._apply(convert)
  File "/home/djowalker/tock/tock-flair/venv/lib/python3.8/site-packages/torch/nn/modules/module.py", line 203, in _apply
    module._apply(fn)
  File "/home/djowalker/tock/tock-flair/venv/lib/python3.8/site-packages/torch/nn/modules/module.py", line 203, in _apply
    module._apply(fn)
  File "/home/djowalker/tock/tock-flair/venv/lib/python3.8/site-packages/torch/nn/modules/module.py", line 203, in _apply
    module._apply(fn)
  [Previous line repeated 1 more time]
  File "/home/djowalker/tock/tock-flair/venv/lib/python3.8/site-packages/torch/nn/modules/rnn.py", line 149, in _apply
    self._flat_weights = [(lambda wn: getattr(self, wn) if hasattr(self, wn) else None)(wn) for wn in self._flat_weights_names]
  File "/home/djowalker/tock/tock-flair/venv/lib/python3.8/site-packages/torch/nn/modules/module.py", line 593, in __getattr__
    raise AttributeError("'{}' object has no attribute '{}'".format(
AttributeError: 'LSTM' object has no attribute '_flat_weights_names'

So the parsing is not working properly.

[djowalker]

Sems to be ok after upgrading the following :

update download_modules.sh

curl https://flair.informatik.hu-berlin.de/resources/models/ner-fast/en-ner-fast-conll03-v0.4.pt -o models/en-ner-fast-conll03-v0.4.pt
curl https://flair.informatik.hu-berlin.de/resources/models/fr-ner/fr-ner-wikiner-0.4.pt -o models/fr-ner-wikiner-0.4.pt

update requirements/commont.txt

before

bpemb==0.3.0
flair==0.4.4
importlib-metadata==1.3.0
torch==1.3.1
torchvision==0.4.2
numpy==1.18.1
packaging==20.0
requests==2.22.0
sentencepiece==0.1.85
torch==1.3.1
torchvision==0.4.2
transformers==2.3.0

after

bpemb==0.3.2
flair==0.10
importlib-metadata==3.7.0
numpy==1.20.0
packaging==20.9
torch==1.4.0
torchvision==0.5.0
requests==2.25.1
sentencepiece==0.1.95
torch==1.5.0
torchvision==0.6.0
transformers==4.0.0

[djowalker]

Updated with #15

The models are available directly in the code, no need to download them like before.

[djowalker]

@auroredea
@vsct-jburet
Any chances to update and release the recent PR update to recent versions ?