Open Models Almanac
Methodsinference & servinginference quantization

single mention · inference & serving

Higher-precision activations for Mamba output projection

Using higher activation precision as a mitigation for an effect associated with the Mamba output projection.

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 higher activation precision.

Filed under inference & serving › inference quantization