this model was trained on 95% of my dataset menutp/hate_speech-fr_mini and validated on the last 5% it reached 65% accuracy. labels are to be interpreted as follow :

{
        0: "neutral",
        1: "toxic",
        2: "severe toxic",
        3: "obscene",
        4: "threat",
        5: "insult",
        6: "identity hate",
        7: "generally_offensive"
}

due to the lack of training data for the label between 1 and 6 only label 0 and 7 are to be trusted

Downloads last month: 6

Safetensors

Model size

0.1B params

Tensor type

F32