Open Models Almanac
Methodsinference & servinginference quantization

single mention · inference & serving

W4A16 quantization of Mamba output-projection weights

Mamba output-projection weights are quantized to 4-bit while activations remain at 16-bit precision.

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 Mamba output projection→ W4A16.

Filed under inference & serving › inference quantization