llama.cpp
quantize: be able to override metadata by key
#6321
Merged

quantize: be able to override metadata by key #6321

ggerganov merged 2 commits into master from ik/quantize_with_kv_overrides
ikawrakow
ikawrakow1 year ago

This can be useful, e.g., for low-bit quantization where the experience has shown that one can improve the model by changing the f_rms_norm parameter. Instead of having to specify the metadata override each time the model is used, with this PR one can encode the override during quantization using

./quantize --override-kv KEY=TYPE:VALUE other_commandline_arguments

The --override-kv argument can be repeated multiple times.

quantize: be able to override metadata by key
fc4c2a6f
ggerganov minor : spacing
9c5fd6be
ggerganov
ggerganov approved these changes on 2024-03-26
ggerganov ggerganov merged d25b1c31 into master 1 year ago
sakura-umi
sakura-umi1 year ago

Can this also be able to override string type metadata? This has not been supported yet.

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone