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
Reviewing the sweep syntax, the numpy distribution based sampling looks useful. How would one set a seed to ensure the generated hyperparameter sweep combinations are deterministic/repeatable? An optional --seed arg would perhaps be useful for users to specify a value for numpy's np.random.default_rng, which grid can then use to draw reproducible samples from specified distributions.
The text was updated successfully, but these errors were encountered:
Reviewing the sweep syntax, the
numpy
distribution based sampling looks useful. How would one set a seed to ensure the generated hyperparameter sweep combinations are deterministic/repeatable? An optional--seed
arg would perhaps be useful for users to specify a value for numpy'snp.random.default_rng
, which grid can then use to draw reproducible samples from specified distributions.The text was updated successfully, but these errors were encountered: