v2.17.0 — decode an int8-quantized vector buffer back to Float32. Inverse of encodeInt8Vector. Caller passes dim so we know how many bytes are int8 vs the trailing min/scale tuple.
encodeInt8Vector
dim
v2.17.0 — decode an int8-quantized vector buffer back to Float32. Inverse of
encodeInt8Vector. Caller passesdimso we know how many bytes are int8 vs the trailing min/scale tuple.