diff options
| author | Iwan Kawrakow <iwan.kawrakow@gmail.com> | 2024-07-30 12:33:48 +0300 |
|---|---|---|
| committer | Kawrakow <48489457+ikawrakow@users.noreply.github.com> | 2024-08-01 09:38:06 +0200 |
| commit | ab4f9e1fdb7441f8250364b248b5e709ec66771f (patch) | |
| tree | 003a54ba333fb25ad5cf764610c72f1b6c0a8946 /ggml/src/ggml-rpc.cpp | |
| parent | 69842c6ad805c7de8f0416e52a1f12d3357023d9 (diff) | |
iq2_k: CUDA dot product finally works
Performance is pathetic: 140 t/s for LLaMA-3.1-8B vs
172 t/s for iq2_xs.
Diffstat (limited to 'ggml/src/ggml-rpc.cpp')
0 files changed, 0 insertions, 0 deletions