Add Auto Quantize in ModelOpt quantize example - #4821

Merged

jenchen13 merged 13 commits into

NVIDIA:mainfrom

jenchen13:jennifchen/auto_quant

Jul 2, 2026

Commits on May 15, 2026

auto quant modelopt example in quantize.py
jenchen13
committed
b24527e

Commits on May 17, 2026

Mask padding in auto_quantize loss

realAsma
and
claude
committed
bb63a19

Commits on May 18, 2026

Simplify auto_quantize loss comments
realAsma
committed
3c2e036

Commits on May 19, 2026

Support forward-backward calibration for modelopt PTQ
realAsma
committed
6ebfbe3

Commits on Jun 25, 2026

use forward only in ptq
jenchen13
committed
02af4e2
Merge branch 'main' into jennifchen/auto_quant
jenchen13
committed
1152231
docs: add Auto Quantize section and fix typo in modelopt README

jenchen13
and
claude
committed
f30dffe
Remove redundant calibration helpers from auto_quantize path
jenchen13
committed
fe0f4ff
remove outdated code
jenchen13
committed
07b20fb
Simplify auto_quantize calibration path, require modelopt 0.46+
jenchen13
committed
84cb37b
wire --generate-output-len to generation function
jenchen13
committed
ca6a0d7
restore tokenizer handling
jenchen13
committed
21bb887

Commits on Jun 26, 2026

move lm batch utils out of finetune.py and fix forward backward in auto quant
jenchen13
committed
0429f9f