Today, I got this error when using scientific (exponential) notation in a sweep.
wandb: Creating sweep from: config.yaml
wandb: ERROR Error while calling W&B API: Invalid sweep config: invalid hyperparameter configuration: lr (<Response [400]>)
wandb: ERROR Find detailed error logs at: /h/319/wlin/eigen_NGD/models/fastvit/wandb/debug-cli.wlin.log
Error: Invalid sweep config: invalid hyperparameter configuration: lr
the config.yaml is
program: train.py
method: random
metric:
name: validation_loss
goal: maximize
parameters:
lr:
min: 1e-4
max: 1e-1
distribution: log_uniform_values
The error disappears when using the following config file
method: random
metric:
name: validation_loss
goal: maximize
parameters:
lr:
min: 0.0001
max: 0.1
distribution: log_uniform_values
Is scientific (exponential) notation no longer supported in a sweep?