Logging available resources with Tune

Ray Tune
1

Hi,
Does tune log the available resources for an experiment?

Currently I use the analysis tools:

analysis = tune.Analysis(experiment_dir="/path/to/my/experiments") 
df = analysis.dataframe()

Looking at the columns of the dataframe, there does not seem to be any information of how many CPUs and Memory were available in the ray cluster (given by ray.available_resources).

If tune is already logging this information, how can I access it?

If not, what is the proper way to save this information when using the tune.run workflow?

2

ray.available_resources() can be an expensive call, once the number of nodes begins to reach >50.

If you want to log this, you could use a custom Tune Callback.