Tom Benson

Allow Catch2 version 3 to be used in addition to version 2.

1

This will hold us over for a transition period. I'd like to get to Catch2 v3 in general, though, as it's better factored. (I can do that now, but I...

Pinging @timmoon10 @tnat410 @samadejacobs.

Remove the LTFB callback

1

The LTFB training algorithm has been featured for a while now, and projects have adopted it. The training algorithm is the driving force of new development. The callback is now...

(LOW priority) Disable LBANN warning output during testing

2

I'm still not convinced I see the value of `LBANN_WARNING` at all, but we should be able to silence the warning. Actual: ``` > ./unit_test/seq-catch-tests "[data_reader][synthetic]" Filters: [data_reader][synthetic] LBANN warning...

(Slightly less low priority) MPI Catch test output

2

There's a lot of output from the Catch2 MPI tests -- too much to open individual issues. It should be cleaned up so a clean test generates no output. Actual:...

@benson31 should fix that.

Single-node CPU-only performance is not good.

2

There have been reports of slow performance with default spack builds of LBANN in CPU-only mode. @denfromufa reported it in #1443 and I was able to verify in my local...

Apparently there are counters that `lbann_comm` maintains. These are not marked `mutable` (like they should be) so it looks like collectives, etc, are non-`const` functions (they are logically `const`). This...

Setting thread affinity has errors on Lassen

3

I'm seeing ```error in pthread_setaffinity_np, error=22``` when I run Resnet-50 on the latest "develop"on Lassen. @ndryden points out, I believe correctly, that we should use HWLOC for affinity (at least...

`lbann_comm` objects are not cleanly handled, see `weights`, etc. #1247