llama.cpp
ggml : PoC for normalizing weights for better quantization packing
#2434
Open
Open
ggml : PoC for normalizing weights for better quantization packing #2434
ggerganov
wants to merge
3 commits into
master
from
norm-quants
ggml : poc for normalizing weights for better quantization (metal)
253eab8a
ggml : use less ggml_mul tasks when src0 rows are few
df54d2f1
cebtenzzre
commented
on 2023-08-30
cuda : poc for norm quants (only -b 1 works)
8c2b8812
Login to write a write a comment.
Login via GitHub