llama.cpp
ggml : use 8-bit precision for Q4_1 intermediate results
#1047

Merged

ggml : use 8-bit precision for Q4_1 intermediate results #1047

ggerganov merged 4 commits into master from q4_1xq8_0

ggerganov added high priority

ggerganov added generation quality

ggml : use 8-bit precision for Q4_1 intermediate results (ARM)

e9c07f72

ggml : optimize ggml_vec_dot_q4_1_q8_0() via vmalq_n_f32

42623052

ggml : AVX2 implementation of ggml_vec_dot_q4_1_q8_0 (#1051)

ad7007aa

gitignore : ignore ppl-*.txt files

e582f2ad

ggerganov force pushed to e582f2ad 2 years ago

ggerganov merged 884e7d7a into master 2 years ago

ggerganov deleted the q4_1xq8_0 branch 2 years ago

ggerganov assigned

ggerganov 2 years ago

ggerganov assigned

slaren 2 years ago

Reviewers

No reviews

Assignees

ggerganov

slaren

Labels

high priority generation quality

Milestone

No milestone