PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
Commits on main
WNA16 does not apply optimized RTN for moe layers by default (#1245)
9588bf9 Wenhua Cheng committed 10 months ago
Organize test directory structure with logical categorization (#1243)
0f6dc76 Copilot committed 10 months ago
set disable_opt_rtn to optional bool and change default value to None (#1231)
bef28c9 Weiwei committed 10 months ago
gguf add support for Mistral-Magistral-Devstral and Granite (#1226)
6f2f9b9 Heng Guo committed 10 months ago
fix bug of new version gptqmodel and fix few cuda ut (#1204)
6889793 Heng Guo committed 10 months ago
Revert "set disable_opt_rtn as default for int8 RTN" (#1203)
b3ec445 Wenhua Cheng committed 10 months ago
add online quant & offline sgl api generation cuda UT (#1198)
3424d91 Weiwei committed 10 months ago
[Important Change] change mxfp4 to OCP definition (#1200)
af20202 Wenhua Cheng committed 10 months ago