mathurinm

It seems the original issue is addressed, so can we close this and maybe open a new one for the continuation of #473 (import from a `template_solver.py` in the `solvers/`...

@tpanum thanks to the work of @floriankozikowski, we are now able to solve problems of size `3000 x 1000` with a new solver and a smoothed approach: Using the code...

Reproduce with ``` from benchopt.benchmark import _extract_parameters _extract_parameters("data=path/to/stuff,reg=0.1") ``` This works: ``` _extract_parameters(data='path/to/stuff',reg=0.1") ``` The problem is that it's the first case which is triggered by `'data': ["path/to/stuff"],`

Just to make sure: we already have a caching mechanism with `BaseDataset._get_data` that loads the data only once per benchmark. But you're talking about once and for all caching so...

Btw, if the concern comes from the Breheny datasets, @theoguyard is adding them to libsvmdata this week : https://github.com/mathurinm/libsvmdata/pull/36

@Badr-MOUFAD no need to review yet, we realized that a barebone gramcd solver works much better than anderson working set here, we're still in the benchmarking phase. We'll ping you...

@Perceptronium the unit tests are failing: - you need to lint the code (run flake8 to see where the issues are ; customize VS code so that it shows the...

@Perceptronium you can pull upstream main to fix the tests

@wassimmazouz the tests fail on ubuntu because in `test_cli.test_import_ctx_na,me` we are creating a solver in a tempfile, that is no longer deleted, and this solver then gets matched in autocompletion...