diff options
| author | Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com> | 2023-06-10 01:59:17 -0600 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2023-06-10 10:59:17 +0300 |
| commit | 4f0154b0bad775ac4651bf73b5c216eb43c45cdc (patch) | |
| tree | 33a6036c589fd494af7de0cd786e395d4fd3f699 /ggml-metal.metal | |
| parent | ef3171d16241c18581d4d08374f0b9e396ade6b7 (diff) | |
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691)
* Add support for quantizing already quantized models
* Threaded dequantizing and f16 to f32 conversion
* Clean up thread blocks with spares calculation a bit
* Use std::runtime_error exceptions.
Diffstat (limited to 'ggml-metal.metal')
0 files changed, 0 insertions, 0 deletions