llama.cpp
metal: minor q4 optimization and reduce code size
#2248

Merged

metal: minor q4 optimization and reduce code size #2248

ikawrakow merged 3 commits into ggml-org:master from lshzh-ww:mps-q4-optimize

metal: use uint16_t instead of uint8_t.

bbce3928

metal: update rms_norm kernel

4088df14

lshzh-ww requested a review from

ikawrakow 3 years ago

lshzh-ww force pushed 3 years ago

ikawrakow approved these changes on 2023-07-18

lshzh-ww force pushed 3 years ago

metal: use template to reduce size

f3f2e8ee

lshzh-ww force pushed to f3f2e8ee 3 years ago

ikawrakow merged 417a85a0 into master 3 years ago

lshzh-ww deleted the mps-q4-optimize branch 3 years ago

Reviewers

ikawrakow

ggerganov

Assignees

No one assigned

Labels

None yet

Milestone

No milestone