Juan M. Cruz-Martinez

If you could update this as well https://github.com/NNPDF/nnpdf/blob/master/validphys2/examples/API_extension_Pineappl.ipynb before merging that would be great. Otherwise I can have a go tomorrow or the day after and merge it then.

In the log below, the first two runs are just uniform sampling and the second two the quantum device. The first just a plain monte carlo and the second going...

> Is there an advantage? The replicas and members are not expected to be aligned anyway. The only advantage / need for it is when replicas need to be ordered...

GPU.

I'm having the same problem but I'm unable to find applications > OAuth applications in the Zenodo webpage. Maybe this has changed?

@kamillaurent please have a look at the implementation of the dataset (check that all correlations are ok / bin widths taken into account, etc)

I wonder where the difference is coming from? How are the cuts applied differently (the training/validation cuts?) what would happen if you do the closure fit without tr/vl split like...

> could you be more specific? At the beginning you did a test with trvl: 1.0 and you found small numerical differences. My suggestion to check whether these grow enough...

I thought you were applying the mask _after_ diagonalizing since you apply it to `eig_vals` (so, in a fit without MHOU, it s applied at the level of the experiment,...

Nice, thanks! Then the next step I'd say is to do a large fit (many replicas) to see how the choice of diagonal vs diagonal might affect the results. Btw,...