Add basic support for MXFP6_MOE quantization#16777

Open

horasal wants to merge 14 commits intoggml-org:masterfrom

horasal:pr-mxfp6

Commits on Oct 25, 2025

Fix build error and disable AVX2
horasal
committed

Commits on Oct 26, 2025

Experimental support for MXFP6_E3M2
horasal
committed
Change layout to keep consistency
horasal
committed
FIX quantization
horasal
committed
Fix scaler
horasal
committed
add MXFP6-E2M3
horasal
committed
add basic avx2 support
horasal
committed
remove unused variables
horasal
committed
variable name typo in gguf-py
horasal
committed

Commits on Oct 27, 2025

restore format of gguf/constants.py and quants.py
horasal
committed
minor format cleanup
horasal
committed
solve flake8 and pyright error
horasal
committed

Commits on Oct 28, 2025

implement E8M0_TO_FP32_ANY for better e
horasal
committed
python cleanup
horasal
committed