Skip to content

Conversation

ashelkovnykov
Copy link
Contributor

  • Photon team at LinkedIn has continued to work on hyperparameter tuning code
  • Working on releasing it as a separate open source project
  • Until then, removing the old version which was available but unused

Copy link
Collaborator

@cmjiang cmjiang left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM.

* @param transformMap A Map with key-value pairs of indices and names of transform functions.
* @return The transformed vector.
*/
def transformForward(
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why do we delete this function? Our internal atlas-ml does not have functions to perform parameter transformation.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's wait to merge this PR until the internal dependency changes.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@yunboouyang @cmjiang
I don't understand this comment - are these two functions not available in the internal hyperparameter tuning library? Is there a blocker to copying the implementation over from here?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The reason to leave these two functions in Photon ML is that, we wanted to do the same transformation in Photon ML. Photon ML doesn't have the dependency on the internal hyperparameter tuning library. If Photon ML doesn't need to reuse these functions, we can move it to the tuning library.

* @param transformMap A Map with key-value pairs of indices and names of transform functions.
* @return The transformed vector.
*/
def transformBackward(
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have the same concern here.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

See above please

* @throws IllegalAccessException
*/
def apply(className: String): HyperparameterTuner[GameEstimator.GameResult] =
Class.forName(className)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

How do users specify the class name? Do they need to specify the full class path "com.linkedin.atlas.tuner.AtlasTuner"?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That's right - full class path.

- Photon team at LinkedIn has continued to work on hyperparameter tuning code
- Working on releasing it as a separate open source project
- Until then, removing the old version which was available but unused
Copy link
Collaborator

@yunboouyang yunboouyang left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants