What is left to do before final submission :
So far I have (mainly linked to myself)
[x] Merge branches refactor and beluga_dev into master
[x] Modify runner.py so it no longer takes num_runs arg and directly runs each experiment with the appropriate number of runs
[x] Write section on Grid Search (with subtask of correctly organizing the figures)
[x] Word and coherence iteration over all the paper
[x] Add ablation study and grid search results on the additonnal languages
[x] Make gitlab repo (and make it public)
[ ] Mock run of results collection and plotting (We can do as if the results were all collected by dropping them all in the default mlflow_uri for the plotting and simulate results collection via toy datasets)
[x] Have a beer (or 12)
[x] We won't have the results for the Vocab CutOff, we will talk about this.
[x] Remove all Beluga related stuff (including the supercomputer arg from runner and get_sbatch_rgs from Experiment)
What is left to do before final submission : So far I have (mainly linked to myself)
and grid search resultson the additonnal languages