Thank you so much for taking the time to create and share this tutorial. I've been struggling to understand BERT tokens and using them for text classification and your video has helped me a lot. Thank you very much.
This was so neatly laid out even at double speed. Nice work, keep it going. You are doing a great service and I will definitely keep your brand in mind when looking for consulting. Cheers
Thank you very much for these amazing series on BERT (Data Processing and Classification)!!! 😊 Explanations were crystal clear. Great job!!!! Hope you keep posting more NLP stuff👌🏻
I'm so thankful for this video. I learned the basics of transformers by theory before, but had no idea how to apply it through pytorch. I'm looking forward to watch more of your tutorials. Genuine thanks to you again.
Great video. I have followed almost all your videos in pytorch and I have one word for you - "You are the BEST ". 1 will be glad if you will make a video on reinforcement learning. Thank you.
Thank you for the really great Video! (28:42) : on the graph, the length of the sequences is indicated on the X axis. But what do the values on the Y axis stand for?
Thanks for your awesome presentation ! Much Love and respect to what you're doing . I have a little question : Are you fine tuning the model for all the bert model layers or for the last layers (dropout and output layer ) only ? And what do you advise us to do . Cheers
Amazing Video. However, when I run the code and get to "data = next(iter(train_data_loader))" i get an error "BrokenPipeError: [Errno 32] Broken pipe". Do you have any suggestions?
def create_data_loader(df, tokenizer, max_len, batch_size): ds = GPReviewDataset( reviews=df.content.to_numpy(), targets=df.sentiment.to_numpy(), tokenizer=tokenizer, max_len=max_len ) return DataLoader( ds, batch_size=batch_size, num_workers=4 # make this 0 ) num_workers have to be 0. If you are using windows you should not define num_workers, because PyTorch dataloader does not support multi-processing on Windows. By default num_workers is 0 and works on Windows.
how to do that trained huggingface model on my own dataset? how i can start ? i don't know the structure of the dataset? help.. very help how I store voice and how to link with its text how to organize that I an looking for any one help me in this planet
Very Nice Explanation Venelin. As you said you do some preprocess and use for BERT, but in some cases, one should know how to preprocess RAW text data so please upload that preprocessing code on Github. (Not video)
Hello, when I open jupyter from the conda environment i create it doesnt have any of the packages such as matplotlib, numpy etc is there any way to avoid having to manually pip install these before launching jupyter?
Are you opening from the Anaconda GUI or are you installing Anaconda on your machine, and then creating a virtual env in your terminal, then running "conda activate virtual_env && jupyter notebook" ? Try doing the latter and you should only have to install your packages once if at all
Thanks for wonderful...videos.. Can u please make video on stakeholders management.... In data science or ML projects.... Communication how u maintain n keep them informed by case study.. ..
Hi Venelin, Nice content as usual in the series and the work really helps the researchers and followers. Can you share the notebooks related to this tutorial or already linked. Thanks
Really appreciate your kindness to make this video.
Found this channel today, incredible videos and I love that there are timestamps for all the different subtopics.
Thank you so much for taking the time to create and share this tutorial. I've been struggling to understand BERT tokens and using them for text classification and your video has helped me a lot. Thank you very much.
This was so neatly laid out even at double speed. Nice work, keep it going. You are doing a great service and I will definitely keep your brand in mind when looking for consulting. Cheers
Excellent video tutorial on BERT pre processing. Will wait for the next video series.
Thanks for watching 😊
Thank you very much for these amazing series on BERT (Data Processing and Classification)!!! 😊 Explanations were crystal clear. Great job!!!! Hope you keep posting more NLP stuff👌🏻
The explanation is great and the content itself is of very high quality. Thanks, Venelin.
The best tutorial about BERT I have ever watched! It would be better if there are subtitles. 😉
I'm so thankful for this video. I learned the basics of transformers by theory before, but had no idea how to apply it through pytorch. I'm looking forward to watch more of your tutorials. Genuine thanks to you again.
Very helpful vedio tutorial for learning BERT. It's a saviour when I found it. Thanks a ton
Venelin Valkov
26:33 it is not that 101 is for [CLS] is a token that will be used to predict whether or not Part B is a sentence that directly follows Part A.
A nice and brief explanation of each concept, waiting for the next part(s)
The next part is available. Thanks for watching!
@@venelin_valkov Where is the next part?
Love your explanations man! Clear-cut, to the point and very easy to follow. Pleease keep making videos!
Great video. I have followed almost all your videos in pytorch and I have one word for you - "You are the BEST ". 1 will be glad if you will make a video on reinforcement learning. Thank you.
I would love to go back to RL, but I would prefer to do it with JavaScript in the browser. Stay tuned for that :) Thank you for your kind words!
In 33:50 , how could we use a stratified split in order to tackle the imbalance issue of the dataset?
love your video. Crisp instructions. Do you know why Colab would crash on Mac OS every time when running sequence length choosing (around 28:17)
same pb
Very nice explanation. I hope you will teach us more in NLP.
Thank you for making such beautiful tutorials
How to build and training model and upload it in huggingface like bert uncased model
Great work! Can you elaborate on why you flatten the output tensors in the __getitem__ in the Dataset class? Thanks
Amazing tutorial, understanding was such a breeze . Thank you very much, your way of teaching is excellent.
Thank you so much for this easy to understand tutorial. Can someone please post the link to this playlist? I can't find the next video.
hello, where is the colab for this tutorial?
Why there were no symbols ($,&,@,#, * etc) removal done?
how do we cite you ?
Thanks for your awesome video!!!
Question: Why should we create dataloaders?
Thank you for the insights and powerful illustrations of these techniques work.
Thank you, could you activate the automatic translation please ?
Is there a size limit for BERT, i am trying to work on 50k of data and it gives out of memory error
Thank you for the really great Video!
(28:42) : on the graph, the length of the sequences is indicated on the X axis. But what do the values on the Y axis stand for?
so bert doesnt need to lemmitize, remove stopwords and all that preprocessing steps?
Hi, How can use BERT to build word embedding model
Why the shape is [8, 1, 128] and not [8, 128]? Why do we need that extra 1 dimension there?
Thanks for your awesome presentation !
Much Love and respect to what you're doing .
I have a little question :
Are you fine tuning the model for all the bert model layers or for the last layers (dropout and output layer ) only ?
And what do you advise us to do .
Cheers
Great tutorial, has been following along, simple and clear!
how to get the dataset from google link....can you paste the link
and nice worf ...hope to see more video like this
Amazing Video. However, when I run the code and get to "data = next(iter(train_data_loader))" i get an error
"BrokenPipeError: [Errno 32] Broken pipe". Do you have any suggestions?
for me it taking too much time
any Suggestion @Venelin
@@mmonut I have the same problem, any suggestions?
def create_data_loader(df, tokenizer, max_len, batch_size):
ds = GPReviewDataset(
reviews=df.content.to_numpy(),
targets=df.sentiment.to_numpy(),
tokenizer=tokenizer,
max_len=max_len
)
return DataLoader(
ds,
batch_size=batch_size,
num_workers=4 # make this 0
)
num_workers have to be 0. If you are using windows you should not define num_workers, because PyTorch dataloader does not support multi-processing on Windows. By default num_workers is 0 and works on Windows.
how to do that trained huggingface model on my own dataset? how i can start ? i don't know the structure of the dataset? help.. very help
how I store voice and how to link with its text how to organize that
I an looking for any one help me in this planet
Very Nice Explanation Venelin. As you said you do some preprocess and use for BERT, but in some cases, one should know how to preprocess RAW text data so please upload that preprocessing code on Github. (Not video)
i think you can also use ktrain for bert
how do we train the model on another langage?do we first train it with MLM and NSP?and pls if so let me know
Very nice explanation, could you make a session about xlmroberta
My text contains more than 512 so how can I address this problem in case of nepali language.
Amazing job, really good !
Amazing work
Hello, when I open jupyter from the conda environment i create it doesnt have any of the packages such as matplotlib, numpy etc is there any way to avoid having to manually pip install these before launching jupyter?
Are you opening from the Anaconda GUI or are you installing Anaconda on your machine, and then creating a virtual env in your terminal, then running "conda activate virtual_env && jupyter notebook" ? Try doing the latter and you should only have to install your packages once if at all
Thanks for wonderful...videos..
Can u please make video on stakeholders management.... In data science or ML projects....
Communication how u maintain n keep them informed by case study.. ..
Very Nice Tutorials
So good, clear and understandable.. Make more videos for us ;)
Hi Venelin,
Nice content as usual in the series and the work really helps the researchers and followers. Can you share the notebooks related to this tutorial or already linked.
Thanks
Hey,
I am working on complete tutorial and notebook. Once done, they will be linked into the description (free, of course). Thanks for watching!
@@venelin_valkov Thanks a lot for your kind efforts Venelin
why cant you use vscode?
Fantastic! Thank you!!!
Awesome tutorial !
Great tutorial.
Great work keep it up!!!
Fantastic as usual
Gracious.
Subscribe your channel just after glancing the comments here.