summaryrefslogtreecommitdiff
path: root/gguf-py/scripts/__init__.py
diff options
context:
space:
mode:
authorIwan Kawrakow <iwan.kawrakow@gmail.com>2024-07-24 08:02:56 +0200
committerIwan Kawrakow <iwan.kawrakow@gmail.com>2024-07-24 08:04:47 +0200
commit6b4167164cdde5dd21b3786bebc0688f5023f326 (patch)
tree4698a915ed6cab82f2de3e4f1bce773185dc102f /gguf-py/scripts/__init__.py
parent2e49f0172f6c11b286a410039ad87433099bc1b9 (diff)
iqk_mul_mat(NEON): special case for n not divisible by 8
Else fp16 PP performance drops by nearly a factor of 2 compared to what we had before.
Diffstat (limited to 'gguf-py/scripts/__init__.py')
0 files changed, 0 insertions, 0 deletions