The Hugging Face Transformers Library | Example Code + Chatbot UI with Gradio

This is the 3rd video in a series on using large language models (LLMs) in practice. Here, I give a beginner-friendly guide to the Hugging Face Transformers library, which provides an easy and cost-free way to work with various open-source language models.
Series Playlist: • Large Language Models ...
📰 Read more: towardsdatascience.com/cracki...
💻 Example code: github.com/ShawhinT/KZbin-B...
More resources
pipeline() doc: huggingface.co/docs/transform...
pip installation: huggingface.co/docs/transform...
Vanilla Chatbot demo: huggingface.co/spaces/shawhin...
--
Book a call: calendly.com/shawhintalebi
Homepage: shawhintalebi.com/
Socials
/ shawhin
/ shawhintalebi
/ shawhint
/ shawhintalebi
The Data Entrepreneurs
🎥 KZbin: / @thedataentrepreneurs
👉 Discord: / discord
📰 Medium: / the-data
📅 Events: lu.ma/tde
🗞️ Newsletter: the-data-entrepreneurs.ck.pag...
Support ❤️
www.buymeacoffee.com/shawhint
Intro - 0:00
What is Hugging Face? - 1:05
🤗Transformers - 2:05
Models - 4:40
Installation - 8:24
Sentiment Analysis - 9:49
Summarization - 11:48
Conversational - 13:04
Chatbot UI with Gradio - 14:42
Web app with Spaces - 17:38
What's next? - 20:46

[@ShawhinTalebi]

More in this series 👇 - Introduction: kzbin.info/www/bejne/qnerloiJf6aMmKc
 - OpenAI API Intro: kzbin.info/www/bejne/mavZh5yYd5efiKM 📰Blog link: towardsdatascience.com/cracking-open-the-hugging-face-transformers-library-350aa0ef0161?sk=225a4171af141daeb3869dcf614201dd 💻Example code: github.com/ShawhinT/KZbin-Blog/tree/main/LLMs/hugging-face More resources pipeline() doc: huggingface.co/docs/transformers/main_classes/pipelines#transformers.pipeline.task pip installation: huggingface.co/docs/transformers/installation Vanilla Chatbot demo: huggingface.co/spaces/shawhin/vanilla-chatbot

[@helrod6131]

Appreciate you walking through this in an easy-to-understand style.

[@ShawhinTalebi]

Happy to help! I’m glad it was clear 😁

[@GBangalore]

Great introduction. Thanks for putting this together.

[@user-xj1pr3pn1p]

I think your video about LLM is the best one in the youtube, thanks very much!!!

[@jakewastaken]

Thanks for this fantastic playlist.

[@mohammadalibazyar5079]

Thanks for compiling the stuff....amazing!

[@ShawhinTalebi]

Happy to help!

[@jmitterhofer]

Great work Shaw!!!

[@ShawhinTalebi]

Thanks! :)

[@muhammadroshanmughees1586]

I didn't know of the huggingface spaces. It is amazing. Thank you for this

[@ShawhinTalebi]

You’re welcome, I’m glad it was helpful :)

[@jakewastaken]

I’ve been living under a rock and just learned about both Hugging Face and Jupyter. Consider my mind blown 😂

[@ifycadeau]

Great video as usual! SO helpful 🙌🏾

[@ShawhinTalebi]

Thanks, glad it was helpful!

[@user-ms3nf6tv6h]

Thank you Shawhin for this really helpful and informative video.

[@ShawhinTalebi]

Thanks! Glad it helped :)

[@matthcw8745]

Extremely useful, I haven't used hugging face, so this is useful!

[@ShawhinTalebi]

Glad it was helpful!

[@spotnuru83]

Really useful for beginners like me, thank you for this

[@ShawhinTalebi]

Glad it was helpful :)

[@lalpremi]

Thank you for sharing, have a great day :-)

[@60pluscrazy]

Thanks, excellent 👍

[@AniHunsaker]

This is so helpful! can you record a video guiding us how data scientists can work in Transformers library if they work with low-resource languages?

[@alemayehudendir3890]

Amazing! Thank you so much.

[@ShawhinTalebi]

Happy to help!

[@vipulbawankar3980]

Really great content!!

[@ShawhinTalebi]

Glad you enjoyed it :)

[@hanimahdi7244]

Great video, Thank you!

[@ShawhinTalebi]

Happy to help!

[@KumR]

Great Intro Shaw

[@ShawhinTalebi]

Thanks, glad it helped!

[@zahrahameed4098]

Your teaching method is exceptional, providing a balanced blend of theory and practical implementation. I would greatly appreciate it if you could create a video tutorial on training InsightFace using a custom dataset. I'm undertaking my thesis on this topic but struggling to find adequate instructional resources.

[@ShawhinTalebi]

Thanks for the suggestion. I added it to my list :)

[@emanuelokoth1973]

This solved me big time, I am working on a client project and this app I'm surprised it's general purpose and answers fitness related questions that is exactly what I'm working on. Big Thanks! Is there a way to fine-tune the model more for my fitness AI?? Big Thanks!

[@ShawhinTalebi]

That’s awesome, glad it helped! I talk about fine-tuning in another video: kzbin.info/www/bejne/m3SZeZdnnauppdU

[@emanuelokoth1973]

@@ShawhinTalebi cool let me check that out! was actually looking on how to fine tune this, since the first is a bitt simple!

[@Techbro-ho3pp]

Thank you for pulling all actionable code in one place and the appropriate amount of technical details. Many of YT videos lab notebooks are stale considering version changes in library. Yours is the first one that worked! In a span of 12 mins, I could complete two of your exercise. While trying the chatbot example, when I import gradio in colab, it causes import of PyTorch and memory spikes and sometimes colab free vm collapses. BTW I want to implement a RAG example to query my pdf with tables. Since my laptop is under resources, was looking for some links or videos on how to push a RAG app to H.F. Would appreciate if you point me to one if you have it, or a better link.

[@ShawhinTalebi]

Great to hear! I'm not a Collab user so not sure what's going on with that Gradio import. I found this video helpful on RAG with docs: kzbin.info/www/bejne/jX2ah2yAit5lqac

[@Techbro-ho3pp]

Thank you Shaw! Do you host or attend live AI related meetups in Dallas area? @@ShawhinTalebi

[@ShawhinTalebi]

Yes, I host at least 1 meet-up a quarter via The Data Entrepreneurs community Events: lu.ma/tde

[@souravbarua3991]

Thank you for sharing. Please make same video using hf api.

[@ShawhinTalebi]

Thanks for the rec! Anything specific you'd like to see?

[@souravbarua3991]

@@ShawhinTalebi when we are using transformers pipeline, everytime the model 1st downloads and then it works. I want to know is there any way so that we can call the above models with hf api/token?

[@ShawhinTalebi]

HF's endpoints might help: huggingface.co/inference-endpoints/dedicated

[@TheCloudShepherd]

Where did you get the Conversation class?

[@ShawhinTalebi]

Good question. I forgot to show the imports in the slides, but Conversation() comes from the Transformers library. Here's the full notebook: github.com/ShawhinT/KZbin-Blog/blob/main/LLMs/hugging-face/hf-sandbox.ipynb

[@amparoconsuelo9451]

Please display the full code and mention the Python version and system configuration, folders, files, etc.

[@ShawhinTalebi]

Code and env files are available here: github.com/ShawhinT/KZbin-Blog/tree/main/LLMs/hugging-face

[@himanshupatidar8040]

it seems vanilla chat-bot is showing error as of current date.

[@ShawhinTalebi]

Thanks for pointing that out! It seems like hugging face updated the lib. I fixed the code on the GitHub repo.

[@forheuristiclifeksh7836]

6:41

[@friedrichdergroe9664]

The sentiment analysis is strange. I put in: "It is fun to hurt people" and it came back with a .99 POSITIVE. I guess it has no clue about morality and values. Also, I have not used Python in a while, and I suppose it has trouble dealing with dependencies, because I had to install tensorflow and tf-keras separately. But it works.

[@ShawhinTalebi]

Thanks for raising this, it's a super important point. While I suspect larger and more recent models will handle that example better, accounting for these edge cases is a major challenge when working with language models.