|
|
|
@@ -62,10 +62,12 @@ class ExponentialDecayLR(LearningRateSchedule): |
|
|
|
decayed\_learning\_rate[i] = learning\_rate * decay\_rate^{p} |
|
|
|
|
|
|
|
Where : |
|
|
|
|
|
|
|
.. math:: |
|
|
|
p = \frac{current\_step}{decay\_steps} |
|
|
|
|
|
|
|
If `is_stair` is True, the formula is : |
|
|
|
|
|
|
|
.. math:: |
|
|
|
p = floor(\frac{current\_step}{decay\_steps}) |
|
|
|
|
|
|
|
@@ -116,10 +118,12 @@ class NaturalExpDecayLR(LearningRateSchedule): |
|
|
|
decayed\_learning\_rate[i] = learning\_rate * e^{-decay\_rate * p} |
|
|
|
|
|
|
|
Where : |
|
|
|
|
|
|
|
.. math:: |
|
|
|
p = \frac{current\_step}{decay\_steps} |
|
|
|
|
|
|
|
If `is_stair` is True, the formula is : |
|
|
|
|
|
|
|
.. math:: |
|
|
|
p = floor(\frac{current\_step}{decay\_steps}) |
|
|
|
|
|
|
|
@@ -171,10 +175,12 @@ class InverseDecayLR(LearningRateSchedule): |
|
|
|
decayed\_learning\_rate[i] = learning\_rate / (1 + decay\_rate * p) |
|
|
|
|
|
|
|
Where : |
|
|
|
|
|
|
|
.. math:: |
|
|
|
p = \frac{current\_step}{decay\_steps} |
|
|
|
|
|
|
|
If `is_stair` is True, The formula is : |
|
|
|
|
|
|
|
.. math:: |
|
|
|
p = floor(\frac{current\_step}{decay\_steps}) |
|
|
|
|
|
|
|
|