| feat: add w4a16 quant
Co-authored-by: xusiyang<xusiyang2@huawei.com>
# message auto-generated for no-merge-commit merge:
!3334 merge master into master
feat: add w4a16 quant
Created-by: weixin_44492126
Commit-by: xusiyang;weixin_44492126
Merged-by: ascend-robot
Description: What this PR does / why we need it?
QAT支持W4A16伪量化
Does this PR introduce any user-facing change?
详细说明见:docs/zh/features/qat_quant.md
How was this patch tested?
参数添加"--qat-scheme w4a16-mxf4"时启用伪量化
See merge request: Ascend/MindSpeed!3334 | 2 个月前 |