BART: Denoising Sequence-to-Sequence Pre-training for NLG (Research Paper Walkthrough)

[@TechVizTheDataScienceGuy]

Full playlist of Research Paper Walkthroughs: kzbin.info/aero/PLsAqq9lZFOtWUz1WEoJ3GXw197LD7BxMc

[@abdulkarimabdulmunaff8783]

Great video friend. Please keep releasing more.

[@TechVizTheDataScienceGuy]

Thanks Abdul. Sure :)

[@LokeshSharma-me5pg]

wow man... you just cleared so many doubts in just one video I'm learning about AI and have all things mixed up but this video gave me a clear view thanks for this keep creating content like this❤

[@TechVizTheDataScienceGuy]

:) :)

[@pradyumnagupta3989]

Beautiful !! You could have talked about the results/performance at the last but overall it was fucking awesome !

[@TechVizTheDataScienceGuy]

Thanks :)

[@ashuagrawal8398]

Crisp and clear. Thanks! If you can maybe provide a high-level comparison on all the models out there.. would be interesting and useful.

[@TechVizTheDataScienceGuy]

Thanks Ashu. Sure will try to put it out as a blog. Will share it over as a community post.

[@ipluvzfanheart]

Need more videos like this :)

[@nilayaishwarya]

Nice one! Now Master Yoda need not worry!

[@motunrayoakinsete]

Thank you for this video. Do you video for Bart for text classification?

[@TechVizTheDataScienceGuy]

I don’t have a video for classification using BART. But you can pick the embedding from here put it in MLP.

[@abc123634]

Thanks for sharing!

[@TechVizTheDataScienceGuy]

Glad you liked it!

[@parsikarudhirbabu9462]

Thanks brother 👍 could you please explain how Bart summarization of text prioritize the important word patterns and outputs them in the desired number of tokens we like to show 🙏.

[@TechVizTheDataScienceGuy]

Hi Parsika, thanks for appreciating. BART tries to maximize the likelihood of the sequence based on the data on which it is trained on. So it will generate candidates having correct patterns most of the times. Post this based on the sampling strategy you can tweak this a bit based on application. Hope that answers your query. thanks

[@kevinzhou7296]

Not just liked it. LOVED IT!

[@TechVizTheDataScienceGuy]

Thank you so much :)

[@yashumahajan7]

Please add something on Marian MT transformers and pretraining objective

[@TechVizTheDataScienceGuy]

Sure. Thanks for the suggestion.

[@yashumahajan7]

@@TechVizTheDataScienceGuy your videos are very good thanks for the content

[@TechVizTheDataScienceGuy]

You’re welcome :)

[@ratikagarg1494]

💡💡💡