Google Colab not working, default is now Python 3.8

[akristoffersen]

Describe the bug
The tinycudann install is failing because Colab recently changed to python 3.8, not 3.7

To fix, we should either update the wheel or add an if-case depending on what python version colab is using right now. This would be in case the roll-out isn't immediate for everyone.

[elloza]

In the meantime, you can use the runtime fallback version.

This is available from the Command Palette via the "Use fallback runtime version"

Ref: googlecolab/colabtools#3246

[vvdb101]

Actually, I also had to comment out the condacolab installation, simply reverting to the fallback runtime version did not work for me.

See my fixed version of the demo notebook.

The condacolab installation restarts the runtime, which falls back to Python 3.8 per default, hence the tinycudann installation still fails.

Indeed, it seems a more stable solution would be to update the tinycudann wheel. Reverting to the fallback runtime is only temporarily offered by Google Colab apparently (until mid-December from what I read), so the demo will become unusable once the transition period to Python 3.8 ends.

[akristoffersen]

@PuzzleFish I just made a PR for this. I spend a good while trying to get a new wheel to work, but with no dice. But luckily the default pip install works fine. Let me know if it works for you!

[vvdb101]

@akristoffersen The tinycudann install works in your fix, thanks!

The only thing is that it takes considerably longer (maybe 13min longer), but given its a temporary fix I guess that's not really an issue.

Unfortunately, the notebook as a whole still fails at the data download part, though due to the unrelated issue #1062, which is, in my opinion unjustly, marked as closed. See my comment there.

[akristoffersen]

The only thing is that it takes considerably longer (maybe 13min longer), but given its a temporary fix I guess that's not really an issue.

@PuzzleFish just created a PR to fix this. Had to bring the tinycuda version down to v1.5, but now its a wheel which should be much faster.

On the data downloading, we are looking into possible solutions to this outside of just reposting the data on google drive. Sorry for the delay on this!