diff options
author | Iwan Kawrakow <iwan.kawrakow@gmail.com> | 2024-07-30 16:11:25 +0300 |
---|---|---|
committer | Kawrakow <48489457+ikawrakow@users.noreply.github.com> | 2024-08-01 09:38:06 +0200 |
commit | 4f237d44f6d75afbb5cef39d4d6b0b35b2a517c7 (patch) | |
tree | 097eb7ee5be6ec08d3958f747775f9225db3c7a4 /ggml/src/ggml-blas.cpp | |
parent | 36204c4ec7815f1b0f8f38c8e1cd3e6b3fef2eaa (diff) |
iq3_k: Basics
Quantize/dequantize, CUDA dequantize.
PPL of LLaMA-3.1-8B is better than iq3_s and iq3_m.
Diffstat (limited to 'ggml/src/ggml-blas.cpp')
0 files changed, 0 insertions, 0 deletions