[Bug]: cannot load 'de-pos-fine-grained'

[lykoerber]

Describe the bug

Hi! I tried to use the fine-grained PoS-Tagger for German Twitter data described here, but unfortunately, I ran into an error trying to load the model. Is this model still included in flair? Otherwise, can you recommend another fine-grained flair model for PoS-tagging for German (apart from basic de-pos)? Many thanks in advance!

To Reproduce

import flair
model = flair.models.SequenceTagger.load('de-pos-fine-grained')

Expected behavior

no error?

Logs and Stack traces

Repository Not Found for url: https://huggingface.co/de-pos-fine-grained/resolve/main/pytorch_model.bin.
Please make sure you specified the correct `repo_id` and `repo_type`.
If you are trying to access a private or gated repo, make sure you are authenticated.
Invalid username or password.

Screenshots

No response

Additional Context

No response

Environment

[helpmefindaname]

Hello @LydiaKoerber , Looking at the commit history, I think it was only added for version 4 but never got merged into master (and therefore isn't there for any future versions).

The model is in a old version and I will talk to @alanakbik to provide an uptodate version.

In the meantime you can download the model from https://nlp.informatik.hu-berlin.de/resources/models/de-pos-tweets/de-pos-twitter-v0.1.pt when loading the model, you need to add some simple fix:

tagger = SequenceTagger.load("de-pos-twitter.pt")  # the downloaded model
tagger.embeddings.embeddings[1].chars_per_chunk=512  # add a missing value not stored in the model
tagger.embeddings.embeddings[2].chars_per_chunk=512

[alanakbik]

If I remember correctly, this model was actually trained by @stefan-it.

[stefan-it]

Ah, so we could upload the model to the Model Hub?

[stefan-it]

Hi @LydiaKoerber ,

after conversation with @alanakbik we decided that it would be better to have a re-trained version on the model hub, because lot of internal modules in the model are pretty depracted and outdated (For example CharLMEmbeddings, whereas FlairEmbeddings should be used now).

So I re-trained the model and documented it here.

I also uploaded the new model to the Hugging Face Model Hub - demo usage with latest Flair version:

from flair.data import Sentence
from flair.models import SequenceTagger

model = SequenceTagger.load('flair/de-pos-fine-grained')
sent = Sentence("@Sneeekas Ich nicht \o/", use_tokenizer=False)
model.predict(sent)

print(sent)

this should correctly return:

Sentence[4]: "@Sneeekas Ich nicht \o/" → ["@Sneeekas"/ADDRESS, "Ich"/PPER, "nicht"/PTKNEG, "\o/"/EMO]

I will prepare PR for Flair (documentation update) and model card on the Model Hub soon!

[lykoerber]

Great, thanks a lot!

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.