Deploy with FastAPI | | Sentiment Analysis with BERT using huggingface, PyTorch and Python Tutorial

[@bankingitaspirant1405]

I'm glad to have found this channel.
I am new to this stuff, but I am learning quickly.
Your PyTorch book is awesome!
It brought me here. It is the reason I googled your name and found this channel.
Thank you man, you are a great help for new aspirants like me.

[@shreyanambiar5073]

I love this channel so much! keep up the good work :)

[@vivekg8667]

How to deploy intent classifier model with fast API

[@BerkayBerkman]

Venelin, you made my day with your cool reaction to the earthquake :D :D

[@MO-xi1kv]

ahah same had to have a rewatch!

[@mateusfmf994]

Thanks for the video!
how can i do this for a model with four labels in the predict data ( I'm trying but I didn't succeed)

[@arronslacey]

Hi - the pretrained model link no longer exists. Do you have a newer link?

[@clivefernandes5435]

Hey I was wondering what will be the cost of the of deploying this model ?

[@dhyanpatel4680]

How did you get the model on your google drive and make it downloadable?

[@buithanhlam3726]

Great! This is exactly what I'm looking for!

[@every_instant]

Thanks for the vids. How do we output attentions score for each of the user input word?

[@venkatesanr9455]

Waiting as usual it is going to be highly informative

[@DistortedV12]

Does this integrate with kubernetes or docker?

[@shaheerzaman620]

Fantastic! Smooth and Easy!

[@sudharsan4040]

Thanks for the awesome video. Can't we using hugging face pipelines for inference rather so much code? Could you please share your suggestions on inference with a custom NER model.

[@TOpCoder100]

Hello, can you make a tutorial on deploying this model on the cloud, plz? Azure | AWS | GCP?

[@evandrojosedossantosdiniz3046]

Thanks! Fantastic!

[@shyampatadia649]

Hello, Great explanation, I had one doubt, can you explain the variable confidence?

[@0730pleomax]

the highest probability from the predicted class

[@slimenbouras3184]

Great vidéo

[@venelin_valkov]

The forward method of BertModel can return the hidden states (or just the last hidden state): huggingface.co/transformers/model_doc/bert.html#transformers.BertModel.forward
You can apply your pooling strategy on top of that. Here are some you might want to implement:
github.com/hanxiao/bert-as-service#q-what-are-the-available-pooling-strategies
Best layer? I doubt there is such a thing (would be great if I am wrong, though). I look at this as another hyperparameter - you need to try a lot of values/strategies.

[@fuat7775]

Awesome content, thank you!

[@vivekg8667]

Can we use intent classification model for this fast API

[@venelin_valkov]

You can use FastAPI and the intent classification BERT model together, yes. Thanks for watching!

[@kabirnarayanjha]

Can I deploy this code on heroku?

[@musilicks]

nah

[@FaizanShaikh-zg9to]

19:35 earthquake

[@yofootball746]

Damn that earthquake !!!!

[@yagelardan8504]

Can you please make more tf2 videos? Ps love your content

[@venelin_valkov]

I'll do some videos on TensorFlow.js soon. Other than that, I am not making any promises. Thanks for the kind words!