would it be possible to properly support the logging of nested dicts?
In the API documentation this features is promised already:
“Logging nested metrics is encouraged and is supported in the W&B UI. If you log with a nested dictionary like wandb.log({"train": {"acc": 0.9}, "val": {"acc": 0.8}}), the metrics will be organized into train and val sections in the W&B UI.”
As pointed out previously, right now, that does not work.
An easy provisional fix would be to concatenate the keys of nested dicts with slashes “/” instead of dots “.” or to support grouping in the GUI by dots as well.
However, it would be great if you could create “subpanels” for nested dicts! Such that users can organize their plots more cleanly.
In particular, for reinforcement learning that will be extremely helpful. I want to separate training info, timing info, reward info, success info, domain randomization info (which needs to be split up again in policy, value function, actor, object…).
