You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have a project with many parameters/metrics, so keepsake ls is only showing a subset of them which is fine. However, if I want to compare the different experiments with regard to a parameter not displayed by keepsake, the only possibility is doing keepsake diff with two experiments at a time (or I missed something).
Example:
I want to see all experiments with a mean IoU greater than 0.1 and compare the actual values but the values are not displayed:
$ keepsake ls --filter "mean_iou > 0.1"
EXPERIMENT STARTED STATUS PARAMS BEST CHECKPOINT LATEST CHECKPOINT
951eca5 23 hours ago stopped ae_n_epochs=0 dff7f56 (step 5) 7d1bcec (step 7)
deep=false f1_score=0.3743 f1_score=0.25648
n_epochs=0
params_path=./models/train-hy...
train_limit=10000
8d2f62f 22 hours ago stopped ae_n_epochs=30 f6b8339 (step 3) a7a1b91 (step 7)
deep=true f1_score=0.25103 f1_score=0.003937
n_epochs=50
params_path=./models/train-de...
train_limit=1000
Since there is already sort & filter is there any plan to add some sort of selection parameter to keepsake ls ? If it is not too tricky I would be happy to contribute. This has already been mentioned on #284 (comment)
Hi!
I have a project with many parameters/metrics, so
keepsake ls
is only showing a subset of them which is fine. However, if I want to compare the different experiments with regard to a parameter not displayed by keepsake, the only possibility is doingkeepsake diff
with two experiments at a time (or I missed something).Example:
I want to see all experiments with a mean IoU greater than 0.1 and compare the actual values but the values are not displayed:
Since there is already sort & filter is there any plan to add some sort of selection parameter to
keepsake ls
? If it is not too tricky I would be happy to contribute. This has already been mentioned on #284 (comment)Anyway, thanks for making keepsake 🎊
Edit: I made my own query tool called sake.
The text was updated successfully, but these errors were encountered: