Open Models Almanac
Methodsinference & servingserving parallelism

general family · filed under inference & serving

Tensor parallelism

Partitions model computation across GPUs; the cited serving configurations specify particular parallelism degrees.

The sources on this page name it without explaining how it works. The line above is the extraction's one-line gloss. See serving parallelism for the methods filed beside it.

Also called Four-way tensor parallelism for serving, tensor-parallel-size 4, tensor-parallel serving on 8 GPUs, tensor parallel on 8 GPUs, Tensor parallelism with degree four, TP4.

sources
4
models
3
labs adopt it
2
strongest
default

Documented in

Tencent (Hunyuan)

Evidence

4 quotes from 4 sources, one card per model, each opening with its strongest.

How sources treat it
optional 2used 1default 1
↑