llama.cpp
Make IQ4_NL quantization be the same on CPU/CUDA/Metal when quantizing K-cache
#6196
Merged
Merged
Make IQ4_NL quantization be the same on CPU/CUDA/Metal when quantizing K-cache #6196
ggerganov
merged
3 commits into
master
from
ik/fix_k_cache_backend_tests
Make quantize_row_iq4_nl do the same thing is quantization on CUDA
cd4a7c4c
Make quantize_row_iq4_nl do the same thing is quantization on CUDA
30eef31b
Now fix test-quantize-fns
68e4fed4
Vaibhavs10
approved these changes
on 2024-03-21
ggerganov
approved these changes
on 2024-03-21
Login to write a write a comment.
Login via GitHub