Skip to content

Commit 32ce04a

Browse files
NexesenexNexesenex
committed
Use of vocab as difquant criteria
The pre-vocab>128k models are more sensitive to ffn_down quant than to ffn_gate and up.
1 parent 86a7e4a commit 32ce04a

File tree

1 file changed

+117
-114
lines changed

1 file changed

+117
-114
lines changed

0 commit comments

Comments
 (0)