Update examples/ner/run_ner.py

[lifefeel]

Update the example file by changing the name of AlbertForTokenClassification to AlbertForSequenceClassification.

[codecov-io]

Codecov Report

Merging #3261 into master will increase coverage by 0.12%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #3261      +/-   ##
==========================================
+ Coverage   77.82%   77.94%   +0.12%     
==========================================
  Files          98       98              
  Lines       16666    16666              
==========================================
+ Hits        12970    12990      +20     
+ Misses       3696     3676      -20     

Impacted Files	Coverage Δ
src/transformers/modeling_utils.py	93.56% <0.00%> (-0.14%)	⬇️
src/transformers/modeling_tf_utils.py	86.22% <0.00%> (+3.75%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update afea70c...547efb9. Read the comment docs.

[stefan-it]

Well, I don't think the `SequenceClassification" head is not the right one, as it is supposed for sequence classification/regression tasks (see https://huggingface.co/transformers/model_doc/albert.html#transformers.AlbertForSequenceClassification).

NER requires a per token classification implementation: https://github.com/huggingface/transformers/blob/master/src/transformers/modeling_albert.py#L791

[srush]

Will this be fixed by switching to AutoModel? I think we are doing that here for run_pl_ner.py #3290

[lifefeel]

I knew what the problem was. In version 2.5.1, there is no definition for AlbertForTokenClassification in run_ner.py
However, it is included in the master branch. I'll close this request.

[lifefeel]

@srush I think using AutoModelForTokenClassification is better than calling each model class. How about making a change to run_ner.py?

[srush]

Yes, if you can send that PR and add me as a reviewers