Skip to content
This repository has been archived by the owner on Sep 18, 2024. It is now read-only.

'quniform' has different meaning beween GridSearch and other tuner. #1335

Closed
xuehui1991 opened this issue Jul 19, 2019 · 1 comment
Closed
Assignees
Labels

Comments

@xuehui1991
Copy link
Contributor

xuehui1991 commented Jul 19, 2019

Short summary about the issue/question:

The 'qunform' has different meaning between GridSearch and Other tuner.

Such as these json file:
image

in GridSearch, the '10' means count, but in other tuner, the '10' means step.

There are some code for parsing 'quniform' in GridSearch.

image

Brief what process you are following:

How to reproduce it:

nni Environment:

  • nni version: N/A
  • nni mode(local|pai|remote): N/A
  • OS:N/A
  • python version: N/A
  • is conda or virtualenv used?: N/A
  • is running in docker?: N/A

need to update document(yes/no):

Anything else we need to know:

@ultmaster
Copy link
Contributor

Just spent an hour figuring out why my learning rate are all zero, after I switched my tuner from GridSearch to TPE... Even didn't realize this when I read the docs of hyperopt...

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
Projects
None yet
Development

No branches or pull requests

4 participants