Deploying BERT Question Answering System as web application using Streamlit

#datascience #machinelearning #webapplication
Video of fine tuning BERT Q&A system - • BERT Question Answerin...
Code of fine tuning BERT system - github.com/sri...
In this video we will deploy BERT question answering application using streamlit web application

[@sriadityab4794]

Can you teach us how to deploy a multivariate time series model for predicting the future trend using any web framework or cloud?

[@AIEngineeringLife]

I have that in my plans. will do it Aditya

[@jithinreddy5932]

Hi is there any way to ask questions without giving context each and every time, insted simply uploading dataset of contexts

[@prasadseptember]

Hi, Thank you so much for sharing valuable information about variety of topics. I m having a query. Can we use topic modeling(TM) outcome in Q&A? Can you share any example If we can use TM? Thank you

[@AIEngineeringLife]

You mean run a topic model on text and use it in Q&A right?. If yes possible and you can have a simple elastic search based Q&A system

[@prasadseptember]

Thank you so much for clarifying my query .

[@YouMe-cl9cf]

It is showing a error that I can't find the solution : AttributeError: module '__main__' has no attribute 'Model'

[@archanadurgam3711]

Please do a video of fake news detection in social media using bert please

[@priyabratasahoo9990]

I am not able to run this . could you help me what all file I need to download to do this

[@sangeethamr917]

Can we run this on google colab?

[@AIEngineeringLife]

Yes by tunneling but sometimes colab blocks it as well

[@amruthak8762]

CDQA is now not supported. How do we go about this?

[@LearningModel-l3t]

Hey! Did you solve this error? Kindly let me know

[@abhishekprakash9803]

sorry to say......whatever u build a models thats supports only factoid answers.........i need a answers more than one sentences .............atleast tells us about models ,,,,,,,,,,,,,,,,,this models supports and generate short answers

[@YouMe-cl9cf]

If you want to generate then try to use GPT

[@cefeida7]

Hello. The video is very interesting, thanks for sharing it. I have a question, the command: "st.spinner ('Loading Model Into Memory ...)" prints that message in the app, but also another message is printed by default: "Running load_model ()." (both messages are displayed). How could the message by default be prevented from being printed and only keep the message that was specified with the "spinner" command? Thanks a lot.

[@prabhup84]

Hi, how to get multiple answers from multiple questions in a single try? for example, "How much revence has been increased and when it was? " here, we need answer like revenue and year/date. can we get it from bert question & answering system? the second question refering the context of the previous question

[@amruthak8762]

how did you resolve cdqa error: ERROR: Could not find a version that satisfies the requirement cdqa (from versions: none) ERROR: No matching distribution found for cdqa

[@LearningModel-l3t]

@@amruthak8762 Hey! Did you solve that error? I have been struggling with it since ages

[@akanksha7474]

Is there anyway that we input the file from the user and then provide answers?

[@AIEngineeringLife]

Yes you can have the code where these files were indexed as well part of streamlit. This will allow documents to be fitted on upload by users

[@41keswick]

seems to be a problem pip installing streamlit as pyarrow is not able to download. i even tried switching to python 3.7 in the shell image and creating a virtual environment inside the editor to no avail

[@41keswick]

nvm i had to update to python 3 in cloud shell editor via the instructions in the below link and it pip installed fine. stackoverflow.com/questions/53468831/how-do-i-install-python-3-7-in-google-cloud-shell

[@41keswick]

i reloaded the customer bert model after converting it to cpu instead of running on gpu bc of the cloud shell, but after previewing im getting the message of None type has no attribute axes. seems the predict function is turning up no output

[@AIEngineeringLife]

@@41keswick .. I have used streamlit on cloud shell as here - github.com/srivatsan88/google_cloud_AI_ML/tree/main/gke_autopilot Infact have a video on that as for streamlit colab is not reliable I found with changing colab restrictions