Parallelism2

[timmens]

What we did:

• Differentiate between search radius factor and sample radius factor in speculative sampling

What happened:

• Benchmark results got slightly worse (especially for 4 and 8 cores)

What we want to do:

General idea: Make speculative sampling resemble sampling step after acceptance decision

• Tune search radius factor for parallel optimizer
• Use filter on existing points in speculative sampling

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (1cacab1) 78.46% compared to head (ee26d56) 78.47%.
Report is 1 commits behind head on parallelism.

Additional details and impacted files

@@             Coverage Diff              @@
##           parallelism      #10   +/-   ##
============================================
  Coverage        78.46%   78.47%           
============================================
  Files               68       68           
  Lines             4928     4929    +1     
============================================
+ Hits              3867     3868    +1     
  Misses            1061     1061           

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