search

google / ml-compiler-opt

Infrastructure for Machine Learning Guided Optimization (MLGO) in LLVM.
Apache License 2.0
612 stars 92 forks source link

How to know the effect of model inlining when training the model？ #329

Closed 18liumin closed 8 months ago

18liumin commented 9 months ago

model/reward_stat_map file contains two parameters, what do these two parameters mean? default_reward and moving_average_reward

mtrofin commented 9 months ago

1 - moving_average_reward / default_reward would give you the percent improvement. "Reward" is a misnomer. It's rather a score (i.e. you need to compare that score with another one to get what really is a reward - a measure of improvement / regression)