MULTI-CLASS TEXT CLASSIFICATION USING 🤗 BERT AND TENSORFLOW 

Thank you bro! Yes i will try to be consistent.

Why i used bert-base-cased? As you can see in the name 'cased', the text is same as input text therefore no changes, but in uncased, the text has been lowercased before the tokenization step. Yes i think you can create a normal function to calculate those metrics and we have an "on_end_epoch" function in which you can feed the metrics function, you will need to do some research on it. (In the future I shall create a video on this) By the way if you liked the video please do subscribe the channel and share among your friends!

@@theartificialguy6898 i have subscribed and will share. Along with calculating above metric at the end of each epoch, if you could show how to hypertune parameters such as class weights dictionary using a functionality similar to gridsearchcv then it would make a complete tutorial. Any idea when you could post the next update? thanks!

@@nikhilgjog maybe in the upcoming weeks & thanks for the idea!

Thank you! Don't forget to like the video and subscribe the channel!

Thank you! Don't forget to subscribe the channel & like the video!

@@theartificialguy6898 kindly create a video and teach us how to use imbalanced dataset in pytorch (smote or any over sampling technique)

@@ganeshsuresh4723 i have already created a video on handling imbalanced dataset, here ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-ubxfWPg2dJ0.html

if you get confusion matrix then please give me the code..

if you get confusion matrix then please give me the code..

i got the same error as am working with different datset than the tutorial, did you know how to solve it? thanks inadvance

@@nesmaabdelaziz7268 arr = df['rating'].values labels = np.zeros((num_samples, int(arr.max()))) arr = arr.astype(int) labels[np.arange(num_samples), arr-1] = 1 Did a little bit of typecasting as my environment was reading 'arr' as float. I hope this helps.

In my case, I used a dataset where the labels was from 1-10

in your dataframe for that particular column change the type to integer. as i can guess your rating attribute is having float or string values.