Alan Akbik

icon indicating a website link http://alanakbik.github.io/

icon location Berlin icon location Creator of the Flair NLP framework

Results 110 comments of Alan Akbik

Add OneClassClassifier model

@jeffpicard thanks for the PR! @elenamer can you take a look?

Add OneClassClassifier model

Hello @jeffpicard also from my side sorry it took so long to reply! Thanks for the many ideas and input! - ThresholdPlugin -> yes, I think that's a good idea....

feat: Add chunking function for sequence tagger training on sentences exceeding token limit

We are getting a `System.IO.IOException: No space left on device` error for the unit tests as they seem to be taking up too much space. I tried removing some of...

Adding a Deep Nearest Class Means Classifier model to Flair

Thanks a lot for adding this @MattGPT-ai and for reviewing @plonerma!

GH-3488: Support for writing a ColumnCorpus instance to files

hello @chelseagzr very sorry it took so long! I updated your branch for our current master, but one of the imports got lost. Could you add it back in?

[Question]: tarsclassifier prediction is very slow for more then 100 classes

Hello @sainiudit, unfortunately TARS is slow in scenarios with many classes. The reason is that TARS makes an independent forward pass through the model for each class. So if you...

Log regression task metrics in multitask model

@ntravis22 @MattGPT-ai Could you paste a script to test this PR?

[Bug]: Exception raised when creating Sentence from text with apostrophes when using SegtokTokenizer

Hello @dropther thanks for reporting this. It seems the error is caused by one of the functions in segtok, the library we use for tokenization: ```python from segtok.tokenizer import word_tokenizer,...

Feat: Multi-GPU Evaluation

@MattGPT-ai this is due to a new mypy version and affected a deprecated class. I just fixed it in #3613. If you update this branch to current master the error...

[Question]: stack of multi-forward,multi-backward , and multi gives very low cos similarity

Hello @wildfluss thats a good question. There's likely two factors at play here: very important is the amount of multilingual text data used to train these models. You could try...