Flair: Question About Using Load_From_File For TextClassifier

[darien-schettler]

I have trained a model and when I attempt to load it I get the following error:

AttributeError: type object 'TextClassifier' has no attribute 'load_from_file'

NOTE: The model file exists in the current directory and is called _best-model.pt_

Code Snippet:

from flair.models import TextClassifier

def make_prediction(model_location="./best-model.pt"):
    # Load the classifier that was created
    classifier = TextClassifier.load_from_file('resources/taggers/ag_news/final-model.pt')

    sent = ["FREE entry into our £250 weekly comp just send the word WIN to 80086 NOW. 18 T&C www.txttowin.co.uk"]

    # Convert into format classifier understands
    sentence = Sentence(sent)

    classifier.predict(sentence)

    # Format the outputs labels to be easier to read
    print(sentence.labels)
    print()
    label = str(sentence.labels[0]).split()[0]
    print(f"{label}\t{sentence}")
make_prediction()

I spent a long time attempting to troubleshoot this on my own. I hope raising this issue here is appropriate. If not please let me know where I should have posted it and I will move it or delete and reupload accordingly.

Thank you!

[amitbcp]

Hi @darien-schettler ,

I believe you are using Flair v0.4.2 . In the new release the function was refactored. Please use TextClassifier.load('resources/taggers/ag_news/final-model.pt')

This can be seen from the release notes https://github.com/zalandoresearch/flair/releases
I hope that helps and answers your question

[darien-schettler]

Thank you!

[alanakbik]

Yes sorry about that, just updated the tutorial to reflect this change!

[senthilmk]

 :ok_hand: