• Daniel Hiltgen's avatar
    MXFP4 support · 4fb47ed3
    Daniel Hiltgen authored
    This implements the Open Compute Microscaling (MX) FP4 format
    as a tensor type with backend implementations focusing
    on mulmat and mulmatid on CPU, CUDA, and Metal.
    4fb47ed3
ggml.go 35.6 KB