Live coding 1

I’m using Ubuntu 18.04 (Bionic Beaver) within WSL 2 on Windows 10. I tried this on my work computer so perhaps the version of Windows 10 and Ubuntu are different from what one might typically install (?).

You can open the file in whatever text editor you’re used to, select the text, and press Backspace on your keyboard.

Yeah just sounds like the software on that computer hasn’t been updated to the latest versions of stuff. sudo apt update && sudo apt upgrade in your terminal should get you updated.

1 Like

Ok I think I’ll take a stab at answering the second question: It’s because Jupyter Lab includes Jupyter Notebook and not the other way round. I think it said somewhere that Jupyter Notebook is basically a part of Jupyter Lab. They also both use the same kernel and are fully compatible with each other.

On Mac, after installing MambaForge I am getting the following message when running “jupyter notebook”:

I have looked and cannot find another jupyter_client distribution other than the MambaForge installed 7.3.2 version.

notebooks seem to work fine but error message is annoying ! :slight_smile:

Ok, it looks like the Mac version of the ManbaForge installation is missing a few elements compared to the Ubuntu version.

To get jupyter to work:

Pandas is also missing…

mamba install pandas

When running “jupyter notebook” we get the following error:

Which is fixed by install ipywidgets:

mamba install -c conda-forge ipywidgets

Then as recommended we install nbextensions:

mamba install -c conda-forge jupyter_contrib_nbextensions

When running jupyter notebook we get the following messages:

This is do the the fact that nbconvert 6 changed template_path to template_paths…
The fix is to downgrade nbconvert:

mamba install “nbconvert=5.6.1”

But wait…

Can’t downgrade to 5.6.1 with python 3.9 !

Kind of stuck here, if you want to use nbextensions with this setup, you would have to downgrade to python 3.8

Here is another annoying issue with this MambaForge setup.
mamba and conda are supposed to be interchangeable…

If you install ipywidgets with conda:

This will upgrade conda-4.12 to conda-4.13 and corrupt your mamba installation.

You can recover from that by downgrading conda back:

conda install conda=4.12.0

You can safely ignore those messages.

1 Like

Just sharing…

  • I hit a bump when I chose Full Screen in Terminal > Settings.
    The Settings menu item disappeared so I could not revert.
    Turns out,
    CTRL-comma(,) is the shortcut to bring up settings from within Terminal.
  • Another way to paste text into a Terminal is a mouse right-click.

Is that in Mac Terminal or Windows Terminal. (In Windows Terminal the shortcut I use to toggle full-screen is Alt-Enter)

It was Windows Terminal. I wasn’t aware of Alt-Enter, that would have helped - now filed in my wetware neural accumulator. Thanks for the tip.

Hi everyone,

Hope you are all well.

I’ve had a hectic couple of weeks, so I’m catching up with the fast.ai walk-thrus, etc.

I just wanted to say that this first session was really incredibly awesome!

Thanks very much Jeremy, for these really amazing sessions. Much and truly appreciated.

Thanks
Kind Regards,
Zakia

2 Likes

Welcome back! Feel free to join any of the sessions even if you’re not caught up yet, and feel free to ask questions about older sessions at any time.

1 Like

Thanks Jeremy - it really means a lot to me. Much appreciated.

I have a humble request please (unsure if it has been covered already as I just started catching-up with the walk-thrus, as mentioned above) - but could you please include the following at the next walk-thru?

  • How to tackle or strategies in general, and best practices etc for tackling machine learning hackathons (irrespective of the problem at hand, etc)? I think this will be very useful, please. How to go about it, etc?

  • What are the best practices or your advice (the best way, as you say!) for structuring machine learning projects, and also their related web project (so a website for example, based on using a machine learning model, etc)? So generally I know there would be two projects, one for the model(s) one tries, and then based on the best model(s), create a user interface (say, website) for the user - so how should both projects be structured, and their connectedness for accessibility, etc? I know in general there is the concept of design patterns in computer science, etc - so what are your thoughts about this, etc? Also, can you speak about things like unit testing - so basically design patterns and unit testing - concepts from computer science - and their applicability to machine and deep learning, projects based on these, and then websites based on those, etc, please?

  • Can you provide a checklist of like things to look out for etc when deploying machine learning projects (both the models project and the web project based on the model[s]), etc? Sort off, like a deployment plan thing?

I just have those thoughts or ideas that may be helpful for all of us - it will really be appreciated to get your advices on the ‘right way’ of doing all of the above.

Looking forward, Jeremy…

P.S. Also, I will try my best to attend the LIVE session of the next walk-thru, although I’m just catching-up presently - an added factor, remember, is that it is 3AM in South Africa during the walk-thru times… :slightly_smiling_face:

Thank you,
Zakia

2 Likes

Also add to the above please:

  • What are your thoughts on utilizing behavior-driven development and including this into machine learning projects - is it possible or applicable to machine learning projects - the model(s) parts and the user-interface parts - well, its generally applicable I would highly assume, to the latter, because that is essentially general programming stuff like building websites, etc - but would like your thoughts on these things, please.
1 Like

Great questions @Zakia ! Actually I do cover quite a bit of that in the most recent sessions, so perhaps ask again if you still have questions after you’ve watched them, if that’s OK?

Apologies that the timing is so bad for SA! We do have someone from Sweden joining at 2am, so you would have some late night company :wink:

2 Likes

Ah okay, am sorry - yeah, I’m still going through the recordings - so look forward to hearing about the stuff I asked above. Yep, sure, I’ll ask questions after I watch the recent recordings.

Actually, just curious: is there like a statistics of the number of people from around the world joining the fast.ai course in this Live cohort? Would be nice to see - am I the only one from South Africa? :slight_smile:

Just curious, that’s all.

That’s not info we’ve collected - sorry!

Oh also, be sure to read ch2 of the book if you haven’t already, since there’s a lot of discussion of deployment issues there.

2 Likes

At minute 40 of the session: For Mac user using zsh instead of bash, look for .zshrc instead of .bashrc to erase the script added by mambaforge python installer

1 Like

Even after upgrading, it didn’t recognise to --no-progress-meter option. I think that only works with Ubuntu 20 (which allows the appropriate curl version …).

1 Like