GLM-5.2
Z.ai (Zhipu AI)substantially increase throughput via mixed-precision training/rollouts together with MTP and Prefill-Decode (PD) disaggregation
general family · filed under optimization
Use of mixed precision in training or rollouts without an evidence-supported specification of the underlying formats or policy.
The sources on this page name it without explaining how it works. The line above is the extraction's one-line gloss. See training precision for the methods filed beside it.
Also called mixed precision, mixed-precision execution, mixed-precision training/rollouts.
3 quotes from 2 sources, one card per model, each opening with its strongest.
substantially increase throughput via mixed-precision training/rollouts together with MTP and Prefill-Decode (PD) disaggregation
The training loop supports the standard levers for scaling distributed training – FSDP (with optional CPU offloading), mixed precision, gradient checkpointing, torch.compile, and gradient accumulation.
The loop composes cleanly with the distributed training primitives users expect: FSDP with optional CPU offloading, mixed-precision execution