Investigate Github Workflow RunTime

[NimaSarajpoor]

See: #982 (comment)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.41%. Comparing base (e7bb2b8) to head (a0aee8e).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1004   +/-   ##
=======================================
  Coverage   97.41%   97.41%           
=======================================
  Files          89       89           
  Lines       14922    14922           
=======================================
  Hits        14537    14537           
  Misses        385      385           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[seanlaw]

@NimaSarajpoor The changes look good to me. Can you describe what is actually taking a long time or causing the spike in time?

If this is sufficient, is there a reason why we don't change the tests in test_ostinato and test_aamp_ostinato?

[NimaSarajpoor]

@seanlaw
The spike is coming from calling gpu_ostinato.

I think the only difference between test_ostinato and test_gpu_ostinato is the part that we compute the consensus. The former uses ostinato and the latter uses gpu_ostinato. These two functions themselves are similar and the only difference is that the former calls stump and the latter calls gpu_stump. When we turn off CUDA and JIT, stump algo should be faster. For example, on my Mac, for the window size 20 and the input list Ts that contains time series of lengths 512 and 1024, the gpu_ostinato takes 60 sec while ostinato takes 3 sec. I did not dig further.

[seanlaw]

Shall I merge this now?

[NimaSarajpoor]

@seanlaw

Shall I merge this now?

sure 👍