Daiki Higurashi

I think the most important parameter for normalization is momentum. `eps` has little effect in my experience.

I'm trying to use the trainer with batches as `dataclass` but I found it only supports `tupple` which has attribute `_fields`, `dict`, `list`, `tupple`, and `torch.Tensor`. https://github.com/pfnet/pytorch-pfn-extras/blob/0c16c95ed97d7e76edd1db6ba25ddda2042d4440/pytorch_pfn_extras/runtime/_runtime.py#L72-L84 I guess it...

`BestValueTrigger` does not work well in mpi run like multi node setting

3

## What happened `_DistributedSnapshot` with `BestValueTrigger` gets stuck. ### code https://gist.github.com/dhgrs/56424106e00bafee9617b0a15a028c2c ### command `CUDA_VISIBLE_DEVICES=0,1 mpiexec -N 2 python3 mnist.py` ## Why it causes Reporter works in all mpi processes without...

WaveNet example's test is not implemented

1

Now WaveNet example is not tested at all. I'm working on adding the test into https://github.com/chainer/chainer-test/blob/master/test_example.sh .