No agnostic line search

[gdalle]

Testing the waters to see if #40 has been solved by the latest release of FrankWolfe mentioned in https://github.com/ZIB-IOL/FrankWolfe.jl/issues/387

[gdalle]

@matbesancon this still errors

[codecov-commenter]

Codecov Report

Patch coverage has no change and project coverage change: -0.26% :warning:

Comparison is base (0493719) 89.78% compared to head (fc9992e) 89.52%.

:exclamation: Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #98 +/- ## ========================================== - Coverage 89.78% 89.52% -0.26% ========================================== Files 24 23 -1 Lines 695 678 -17 ========================================== - Hits 624 607 -17 Misses 71 71 ``` [see 5 files with indirect coverage changes](https://app.codecov.io/gh/axelparmentier/InferOpt.jl/pull/98/indirect-changes?src=pr&el=tree-more&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=None)

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[matbesancon]

Ah my bad I hadn't released the better error printing in case of NaNs. But this will improve the error message only, the underlying problem here seems to be with NaN elements in the gradient no?

[gdalle]

It's interesting that it works on 1.7 though

[matbesancon]

maybe you are not reaching the point where the gradient evaluates to NaN? The RNG in Julia changed from 1.7 onwards if I recall

[gdalle]

Possibly. Not sure how to debug this though, the MWE is pretty involved

[matbesancon]

one ugly way would be to print the gradient at every iteration, or print the direction in the LMO itself

[matbesancon]

let's debug that between main dish and dessert next week