fix(tf2): correct off-by-one error in learning rate decay calculation

rickstaa / stable-learning-control

A framework for training theoretically stable (and robust) Reinforcement Learning control algorithms.

https://rickstaa.dev/stable-learning-control

MIT License

6 stars 1 forks source link

fix(tf2): correct off-by-one error in learning rate decay calculation #415

Closed rickstaa closed 6 months ago

rickstaa commented 6 months ago

This pull request resolves an issue that led to incorrect learning rate decay. The root cause was an off-by-one error in the step count, which skewed the decay calculation. With this fix, the learning rate now decays accurately according to the specified schedule.