ambiguous · filed under optimization
Learning-rate annealing
A learning-rate schedule described as annealing, with the evidence specifying decay durations but not a particular curve.
- source
- 1
- model
- 1
- lab adopt it
- 1
- strongest
- used
How sources treat it
One count per evidence span, weakest treatment to strongest.
used 1
Documented in
Evidence
1 span quoted from the sources, strongest treatment first.
starting learning rate annealing (both a 5T and 10T decay) immediately after rolling back to the 15T token checkpoint mitigates divergence
usedoptimizationin Nemotron 3 UltraNVIDIA
Filed alongside
Other methods under optimization :: learning-rate schedule.
Cosine decayWarmup-Stable-DecayBatch-size warmupTraining without batch-size warmupBatch-size warmup with adjusted peak learning rateBatch-size warmup with constant-batch optimum peak learning rateEngram learning-rate scalingFixed elevated learning rateLinear warmup followed by cosine decayScaling-law fitScheduled batch-size growthWSD-specific scaling law