Replication for MS MARCO passage on Colab

[qguo96]

Replication on Colab is successful.

Colab Environment:
OS: Ubuntu-18.04.3 LTS
Java: 11.0.8
Python: 3.6.9

results are shown below:
Re-Ranking with monoBERT:

Re-Ranking with monoT5:

[qguo96]

I add a Colab notebook under pygaggle/notebooks. Otherwise, it takes too much space. I'm not sure if this is a good idea.

[lintool]

I was just thinking this - add a section of prose to here: https:/castorini/pygaggle/blob/master/docs/experiments-msmarco-passage.md

describing any tips/tricks to getting the replication colab, as opposed to including the entire notebook... we want to URAs that come after to actually do some work... ;) and not just click "run all cells!"

thoughts @ronakice ?

[qguo96]

Is the guide concise enough? I followed the guide and redid it again. Newcomers may need a little time to understand.

[ronakice]

Yep, I think that's why I was hesitant to share the notebooks (although I have some for MS MARCO Passage and CovidQA since that's how I initially ran the experiments).

[ronakice]

@lintool @qguo96 I also think we should move these instructions to convert Colab into a GPU instance with full SSH access, elsewhere. While it is very useful, I think it needs to be in some other place since it is useful for all instructions?

[lintool]

Created new repo for holding this information: https:/castorini/onboarding

@qguo96 sorry about changing our mind. Can send PR against this new repo for your instructions?

And I thought the replication log "Results replicated by ... (Tesla T4 on Colab)" should make it clear that this was on Colab?

[qguo96]

Get it. I would make a new PR for this new repo.