shigapov/

paraly_camembert_large_multilabel

from flair.models import TextClassifier
from flair.data import Sentence

# Load the classifier model
classifier = TextClassifier.load('paraly_camembert_large_multilabel.pt')

# Define the prediction function
def predict(input_text):
    sentence = Sentence(input_text)
    classifier.predict(sentence)

    # Extract the label and confidence into a dictionary format
    labels = {label.value: label.score for label in sentence.labels}

    return labels

Results:
- F-score (micro) 0.9349
- F-score (macro) 0.9352
- Accuracy 0.8831

By class:
              precision    recall  f1-score   support

   figuratif     0.9340    0.9340    0.9340       106
     concret     0.9672    0.9077    0.9365        65

   micro avg     0.9461    0.9240    0.9349       171
   macro avg     0.9506    0.9208    0.9352       171
weighted avg     0.9466    0.9240    0.9349       171
 samples avg     0.9578    0.9448    0.9437       171