SOURCE CODE: Mid-Level API

[barnacl (Rekil Prashanth)]

@arora_aman

You might like this!

This is great.
Can you edit medium posts?
Wanted to mention the changes in names: Databunch and DataSource to Dataloaders and datasets will be helpful

[muellerzr (Zachary Mueller)]

@akashpalrecha

Or maybe, it’s just a stupid idea. I don’t know. . I can use some opinions here.

It sounds like a great idea With how complex the mid-level can be, you’d want to be sure to define just what type of problems you’d want people to do (which hopefully by the end of this we’ll all have a great understanding of it) but it seems like a fantastic idea to me!

[barnacl (Rekil Prashanth)]

Yes that sounds cool (Not sure how to do it though )

[arora_aman (Aman Arora)]

Thank you for pointing it out. I will do it now.

[arora_aman (Aman Arora)]

@akashpalrecha

where a user can select a few options from dropdown menus like the number of different items in a dataset, each item’s type, transforms, validation pct, etc. in a very “English” way, and then out pops a fastai2 snippet that gives them a dataloader after substituting in the path.

Wow! This is such an excellent idea!

[akashpalrecha]

I think I can start working on the webapp after I’m generally comfortable with the datablock API. Streamlit is a great library to help out in making simple webapps when you don’t want to deal with things outside of python and markdown.

[muellerzr (Zachary Mueller)]

So as the first lesson we’re going to mostly be discussing PETS (as it’s all we’ve covered so far), would it be a good idea for all of us to write our own first blogs on it and then we can expand outwards? This way each of us can put our own twists on the findings, etc (and so everyone that wants to do something has something to do!)

[akashpalrecha]

sounds good!

[Brainkite (Antonin Sumner)]

@akashpalrecha

I think I can start working on the webapp after I’m generally comfortable with the datablock API. Streamlit is a great library to help out in making simple webapps when you don’t want to deal with things outside of python and markdown.

Sounds awsome

[Raymond-Wu]

Streamlit looks awesome! I’d be down for making an MVP

[ganesh.bhat]

So where do we write our blogs?

[muellerzr (Zachary Mueller)]

I’d recommend using fastpages as a template and write them to GitHub! I personally have a Jekyll-based blog myself but I think I’m going to convert it over. You can write the blogs in a Jupyter notebook with fastpages.

[Brainkite (Antonin Sumner)]

I guess that would be the most fastai way of doing it

[ganesh.bhat]

While streamlit is great, streamlit can’t be used or deployed as a REST API similar to Flask and Django. Same is the case with Voila. Streamlit makes sense if you are not deploying with enterprise BI tools like Tableau, Microsoft power BI or even open source apache superset

[ganesh.bhat]

Cool. Will look up how to setup fastpages

[muellerzr (Zachary Mueller)]

For those wanting rest, etc, I’m also working on deployment to Starlette for all types of problems (should be done in the next few weeks)

[ganesh.bhat]

Do lookup at fastapi as well

[muellerzr (Zachary Mueller)]

I’m focusing on Starlette as it’s the style to what my school is focusing on (I’ll leave fastapi to others )

[ganesh.bhat]

Ok.

[akashpalrecha]

@ganesh.bhat I think you might have made a mistake/typo with the emoji there. Be sure to edit this out.