aup.Proposer.EASProposer

The code is based on Efficient Architecture Search by Network Transformation, github commit 6ef7229.

See license for redistribution.

Configuration

General parameters

Name

Default value

Explanation

proposer

eas

Efficient Architecture Search

The Proposer demonstrates how to integrate Neural Architecture Search based approaches with Auptimizer.

For other purpose, users need to modify this file with the following steps. The init function is used to initialize the parameters and static values for the algorithm. The get_param function uses the setup function to get different strings representing new NAS architectures based on previous architectures and their performance. A similar paradigm can be adopted for integrating other NAS based algorithms with Auptimizer.

class EASProposer(config)[source]

Bases: aup.Proposer.AbstractProposer.AbstractProposer

failed(job)[source]

Mark job as failed in proposer history.

Parameters

job (Job) – Failed job

get_param(**kwargs)[source]

Get new proposed parameter values

reload(path)[source]

Reload Proposer state from path

Parameters

path (str) – path to reload

save(path)[source]

Save Proposer state to path.

Some proposer can not generate new parameters after saving.

Parameters

path (str) – path to save

update(score, job)[source]

Update scores in proposer history

Parameters
  • score (float) – score returned by Job

  • job (Job) – Finished job