llama.cpp
fix(quantize): add NVFP4 default type mapping and scale tensors
#22897

Open

fix(quantize): add NVFP4 default type mapping and scale tensors #22897

t-timms wants to merge 2 commits into ggml-org:master from t-timms:fix/nvfp4-quantizer-scales

fix(quantize): add NVFP4 default type mapping and scale tensors

56f62c48

perf(quantize): compute optimal NVFP4 .scale via MSE minimization

f5df5181

t-timms requested a review from

ggerganov 39 days ago

github-actions added examples

Reviewers

ggerganov

Assignees

No one assigned

Labels

examples

Milestone

No milestone