Is there a quick start mainly elaborating the NSDI paper - CherryPick?

[xiandong79]

It is a little hard to figure out which part of the code is:

1. how to select initial 3 quasirandom training points
2. the import of Bayesian optimization python lib3
3. the definition of acquisition function

Is it possible to provide a guidance like Ernest-github which is nicely organized and illustrated?

Thanks a lot.

[iSultan]

@xiandong79 Were you able to run the code?

[SiGe]

@iSultan look into https://github.com/harvard-cns/cherrypick/tree/master/spearmint/examples/cherrypick

You should pass Cherrypick as the example to execute to spearmint.

[iSultan]

@SiGe I actually did
python2.7 spearmint/main.py --driver=local --method=GPEIOptChooser --method-args=noiseless=1 ../spearmint/examples/cherrypick/cherrypick.pb

and this is what I got

Current best: No results returned yet.
19975 candidates   0 pending   0 complete   0 executed
Choosing next candidate... 
Expected improvement: 1000.000000
>>>>>>> 0 1000
selected job 25 from the grid.
Submitted job as process: 34683
submitted - pid = 34683

[SiGe]

It seems like there is a lot going on with the Spearmint dependencies. I'll have to freeze the pip recipes we used during the submission. Is your work critical? If not, I will take a look into this next week.

[iSultan]

It turns out that I had to install sqlobject. However, I still can't run the code. The current error is:

Traceback (most recent call last):
   File "spearmint/main.py", line 373, in <module>
    main()
  File "spearmint/main.py", line 187, in main
    while attempt_dispatch(experiment_config, expt_dir, chooser, driver, options):
  File "spearmint/main.py", line 300, in attempt_dispatch
    params = job_params(job)
  File "/cherrypick-master/spearmint/spearmint/runner.py", line 108, in job_params
    dbl_vals = param.dbl_val._values
AttributeError: 'google.protobuf.pyext._message.RepeatedScalarConta' object has no attribute '_values'

We are at the stage where we compare our work with different baselines.

[SiGe]

Yes, that's the error that I get, too. Maybe try installing older versions of protobuf and see if that fixes it?

[iSultan]

I tried protobuf-2.6.1 and protobuf-3.1.0 and I got the same error.
Btw, the braninpy example works with me.

[jatinarora2409]

@xiandong79 Were you able to figure out 'how to select initial 3 quasirandom training points' ?