Closed aremazeilles closed 4 years ago
Seems to be solved in octave branch. I would need to understand if you intent maintaining the octave branch, or want to merge it at a moment to master and or dev. Can you indicate the policy you would like to follow here?
Could you confirm that we would focus on the octave-version branch to be the one prepared for Eurobench? Then I think we could close that issue
Yes, I confirm, the branch octave-version
will be entirely for Eurobench.
There are some changes in the last commits:
1) Modified calculation of two PI's (we will add information about this to the Excel).
2) The default NMF replication number was changed to 50 (this is for better approximation). This makes the calculation five times longer. To facilitate testing, you can change the parameter value in line 27 of run_pepato
.
3) We also resolved the reproducibility issue. This was due to the initialization of the NMF algorithm. Now the results are the same from run to run.
OK, I will check point 3, and get back to you
With respect to 2, I would say that at the level of the code, I would need to have the proper configuration (as stated in the readme file) that enable to get the result that is to be also stored in the folder for consistency checking. We will talk back of this later, but for now, the main issue mentioned here is solved.
We envision to launch the code through a docker image, and we need to have the process completely in batch (i.e no interaction with the user).
When I launch the command indicated (pepato_basic), there is an interaction hapening related to the interaction path:
We would need to get rid of that for the eurobench automation. We can foresee three options: