single mention · inference & serving
Mixed-FP8 layers in an NVFP4 recipe
An NVFP4 operating point that uses higher-precision FP8 for selected layers.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See inference quantization for the methods filed beside it.
Also called mixed-FP8 layers.
Filed under inference & serving › inference quantization