Open Models Almanac
Methodsmodel architecturetoken mixer

single mention · model architecture

Kernel-size-7 convolution

The convolution uses kernel size 7, contrasted in the evidence with size 3 in LFM2.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See token mixer for the methods filed beside it.

Also called kernel size of 7.

Filed under model architecture › token mixer