Lesson 3 - Official Topic

[ilovescience (Tanishq)]

Actually I think if you use the experimental instructions here you can get it to work:

Platform: Colab (Free; $10/month Pro)

To use Colab for the course, follow the direction here: For advanced users: EXPERIMENTAL: @imrandude and @deeplearner have made a script for using the Colab environment as a full Jupyter instance. To do so, please follow these steps: Big Warning: This is experimental, not outspokenly recommended, and could be patched at a later date by Google, please use at your own discretion Create an account and generate an authorization code here Add the two following cells into your Colab notebook o…

(see here)

[mario_carrillo]

So viola runs locally it seems?

[sgugger]

It’s in the fastai2 repo.

[gamino (Ariel Gamiño)]

Can you host voila it in github?

[vijaysai]

Streamlit is also really cool in getting things done fast.

[0tist (Jayesh)]

how can we make voila work on the paperspace?
As in, it wont really work by changing the URL of the paperspace hosted notebook.

[sgugger]

We will see the deployment in a minute.

[maya (ravisha)]

Anything about security issues using iPyWidgets / Voila on production ?

[geoffpidcock]

I have, and can testify that it’s easy and quick. This article offers a walk through for a simple dashboard hosted using heroku. https://towardsdatascience.com/quickly-build-and-deploy-an-application-with-streamlit-988ca08c7e83

[jcatanza (Joseph Catanzarite)]

Sure you could, but how many images do you think it would take to capture “not-bearness”? Interesting experiment to try. My bet would be that it would be a deal-breakingly large number…

[ilovescience (Tanishq)]

You will have to set up Colab differently

[Albertotono]

@sgugger

rk anywhere you can host a notebook

This is awesome, I also didn’t know the connection with vue.js and jupyter

[JPKab (JP Kabler)]

Definitely. Streamlit, so far, is pretty solid.

[username_not_found (unknown)]

If I am having difficulty running one of the .nbs, should I start a new post or post to an existing thread? I previously could run fastai last year on this machine, but it says now it doesn’t have enough memory on the GPU for 01_intro this year. Basically just firefox and the docker are running. nvidia-smi shows basically nothing else running.

Running in LXD docker natively with GPU passthrough to a GTX 1080 with ubuntu 18 lts host.

[nchukaobah (nkiruka chuka-obah)]

Thanks

[ilovescience (Tanishq)]

Yep I am not sure either…

[sgugger]

Use a new thread please. You should adjust your batch size as necessary if you have memory issues.

[JPKab (JP Kabler)]

Additional note on CPU vs. GPU for deployment:

You might want to build and deploy a desktop application in an enterprise environment. Basically a big use case believe it or not, especially for my customers.

Dilbert and friends don’t have GPUs on their PCs.

[ilovescience (Tanishq)]

If you do need to deploy directly on a phone, you can use PyTorch Mobile as well.

[maya (ravisha)]

Yes, this is not easy and it’s the drawback with Classification in DL today. Hence my question.