Skip to content

Commit

Permalink
WOQ: Optimize quantization of activation (#2584)
Browse files Browse the repository at this point in the history 
* WOQ: Optimize quantization per-tensor/per-block of activation for lowp-mode=INT8

* Refine threshold of activation size to parallelize quantization
  • Loading branch information
@Xia-Weiwen
Xia-Weiwen committed Feb 7, 2024
1 parent 444d17e commit 05d0764
Showing 1 changed file with 534 additions and 136 deletions.

0 comments on commit 05d0764

Please sign in to comment.