Training emotions data with bert model

I am creating a emotions detection model on distilbert_base_uncased model, using my own dataset, that contains the data as format text, label where label=emotions, like happy, sad, angry

I tried different training method, but did not succeeded, can any one tell how to train the model.

Also I checked this notebook from @bhadresh-savani : UnderstandingNLP/go_emotion_of_transformers_multilabel_text_classification_v2.ipynb at master 路 bhadreshpsavani/UnderstandingNLP 路 GitHub

but the problem is that the data it has contains the labels in array format, eg: [0,0,0,1,0鈥
but my labels are direct: eg: happy, sad, angry
so it is having conflicted and I can not able to train the model.

@bhadresh-savani if you have any idea and can give me what change I would do to make it work my data that will be really helpful!