specific method · filed under post-training
Continual pretraining for long-context extension
Continual pretraining used to extend a model's context length, including staged context-length increases.
Also called long-context extension, long-context training.
- sources
- 3
- models
- 3
- labs adopt it
- 2
- strongest
- used
How sources treat it
One count per evidence span, weakest treatment to strongest.
used 3
Documented in
Evidence
3 spans quoted from the sources, strongest treatment first.
The first sub-stage extends the context to 32K tokens; the second extends it to 128K tokens.
usedtraining objectivein Laguna XS.2Poolside
added a long-context phase (LC-Phase) at the end of pretraining
usedpost trainingin Nemotron 3 UltraNVIDIA
we included a continued pre-training (CPT) stage at a 512k sequence length
usedotherin Nemotron 3 NanoNVIDIA
Filed alongside
Other methods under post-training :: mid-training & continual pretraining.