Exponential Notation is not allowed anymore

[immonomono]

Describe the issue:
After when I update version to 2.4, exponential notation is not working.

When I use v2.1, exponential notation was no issue
{
"weightDecay": {"_type":"uniform", "_value":[1e-7, 1e-5]}
}

from v2.4, following error was occured
ERROR (nni.common.hpo_utils/Thread-1) search space "weightDecay" (uniform) must have float values : OrderedDict([('_type', 'uniform'), ('_value', ['1e-7', '1e-5'])])

But when I convert exponential notation to float, it is working.
Should I not use exponential notation?

Environment:

• NNI version: 2.4
• Training service (local|remote|pai|aml|etc): local
• Client OS: Red Hat
• Server OS (for remote mode only):
• Python version: 3.8
• PyTorch/TensorFlow version: PyTorch 1.8.1
• Is conda/virtualenv/venv used?: virtualenv
• Is running in Docker?: No

Configuration:

• Search space:
  {
  "weightDecay": {"_type":"uniform", "_value":[1e-7, 1e-5]}
  }

Log message:

• dispatcher.log:
  ERROR (nni.common.hpo_utils/Thread-1) search space "weightDecay" (uniform) must have float values : OrderedDict([('_type', 'uniform'), ('_value', ['1e-7', '1e-5'])])

[QuanluZhang]

@immonomono , thanks for reporting this issue! we have reproduced this issue, will fix it soon.

[QuanluZhang]

@immonomono , upgrading pyyaml to >= 5.4 can resolve this issue. we will update the dependency in NNI v2.5

[immonomono]

@QuanluZhang, Thanks for the kind explanation.

[liuzhe-lz]

Fixed in v2.5.