marcsun13
/

ggml-quantization

Model card Files Files and versions

ggml-quantization / README.md

Commit History

add mul_mat_id, and compile upstream's kernels as they ship

9534e0b

Marc Sun commited on Aug 28

vendor only the files the build reaches

091adac

Marc Sun commited on Aug 28

drop the cuda kernel: metal is the only backend built for now

77f577c

Marc Sun commited on Aug 28

Minimal README: ops, devices, source and how to refresh it

fd5d3aa

marcsun13 HF Staff commited on Aug 7

GGUF kernels: dequantize + fused gemv over packed blocks

f79d690
verified

marcsun13 HF Staff commited on Aug 5

GGUF kernels: dequantize + fused gemv over packed blocks, 9 CUDA variants

2c20074
verified

marcsun13 HF Staff commited on Aug 5