Fan Yue

The following hyper-parameters give me 64.96% test accuracy while it is 68.3% in the paper. Could you please check where I set wrongly or if there are other hyper-parameters I...

Hi, You env setting is fine. Actually the problem I think is where you add `x_attn0 = self.SA4(x) x = x*x_attn0`. @HRanWang provided an example in /ablation/Backbone_with(out)_SA1/reid/models/resnet_fusion.py. If the SA4...

> My current workaround is to catch the FileNotFoundError and recreate the data_iterator. To get back to the batch you were on, you can then run load_next_batch multiple times: >...

I just encountered the same issue in a different project. Even the toy example from the [official tutorial](https://flax.readthedocs.io/en/latest/guides/converting_and_upgrading/orbax_upgrade_guide.html#async-checkpointing) fails. I'm running the following code on a TPUv4-16. ``` import numpy...

> I think the issue is that Orbax is assuming the root directory is a global storage. If you specify `/tmp/....`, it is local to each process. Directory creation is...