I ran experiments according to this:
wandb.init(
entity="xxx",
project="yyy",
group=str(_id_seed_independent),
name=str(_id),
job_type="train_eval",
)
Now, in the wandb UI of course I want to group by seeds – that’s why I would run multiple ones, right? – but sadly then the other variables are lost even though they are all the same within the group, see:
This way, I can not nicely analyze the groups in graphs, since all variables become
