Create Custom Dataset for Question Answering with T5 using HuggingFace, Pytorch Lightning & PyTorch

Learn how to create a dataset for Question Answering with T5 using questions from the BioASQ challenge. Learn the basics of the T5Tokenizer and prepare a data module for fine-tuning on Question Answering tasks.

#T5 #QuestionAnswering #HuggingFace #Transformers #PyTorch #PyTorchLightning #MachineLearning #DeepLearning #Python

[awmawam]

I'm making my final year project and it involves training a model on custom data, thanks for this video.

[sayedathar]

Thanks for making such videos, this will be most useful resources for my next job most probably

[dv]

Amazing guide Venelin❤️. it' 'll very helpful if Provide the Notebook.

[deabsoluteschwarz]

Hi, your videos are very helpful! I would love to see an example of extractive document summarization

[pandya]

Very interesting and detailed description. Thanks a lot for this video !!
Here, I have a question for you :

For Zero-shot learning, do we need to modify embedding or not?

Detailing of Question:

I am working on a multilingual-BERT model for the Question-Answering task. The model is already pretrained on the English dataset. Now I want to check it's performance on another language ('Hindi') in Zero-shot setting.

So, to do so by zeros-shot learning which of the following is the correct approach:
1) Give evaluation data (dev-set) of Hindi to model and check the result.
2) Using training data of Hindi train tokenizer and use that new tokenizer with your previous model (do not train m-Bert on training set of hindi) to predict the answer

Which of these is the correct interpretation of Zero-shot learning.

[gprasadk]

Thank you for the walkthrough. Appreciate your effort!

[mahimanzum]

great to know you are a office fan too.

[ashishbhatnagar]

simply amazing tutorial. Thanks a lot for sharing.

[davidsimmonds]

Great tutorial. I'm just missing one piece. I'm not understanding the step between the encoding and the batching of the inputs. I would love some help with that, please.

[iamrxn]

I have a very important question. For the dataset containing only question and answer features. How should I approach? For eg. If the user input question, model must generate answer.

[shivammarathe]

Thanks a lot for sharing. What can be the research gaps/improvements that we can do in this type of tasks ?

[feravladimirovna]

Hi! thanks! Can you please provide us with the notebook to experiment with it?

[JJetinder]

Hi, could you please provide a link to the notebook? I am not understanding it properly without applying it myself.

[jiajundeng]

Hi, I recently make my own datasets, others are prepared but only answer_start is a matter, how could I give the answer_start to raw data?It is I need to find the position in the context.and label it manually?

[fatemeh]

in the BioQADataset in getitem function, why don't we use self.tokenizer? (45:41) I added self. but when I run trainer.fit() I get this error: 
target_encoding = self.tokenizer(
data_row['answer'],
max_length=self.target_max_token_len,

TypeError: 'tuple' object is not callable

without self. it just doesn't recognize the tokenizer (which makes sense). Any idea why I get this error?

[arpitshah]

Hi Venelin, I got RecursionError: Max recursion depth exceeded while calling python object error...can you please provide me the solution. Thank you

[preethiseshadri]

Hi Venelin! Thanks for these videos. Is the code for this available anywhere?

[datareactor]

can you share the link to the dataset as I;m ont able to down that?

[flowerboy_]

i get errors when i install pytorch lighting

[saurrav]

Bro how to put ml algorithms in a dataset