Add Q4_1_O quantization format that preserves outliers in weights and does dot in FP32 #19
build.yml
on: pull_request
Matrix: windows-latest-cmake
ubuntu-latest-make
12s
ubuntu-latest-cmake
20s
macOS-latest-make
36s
macOS-latest-cmake
46s
release
0s
Annotations
1 error
windows-latest-cmake (avx512, -DRWKV_AVX512=ON)
Process completed with exit code 1.
|