Add ONNX export, INT8/FP8 quantization, and SafeTensors inference loading#14

Merged

ivanbasov merged 19 commits into

mainfrom

igor/onnx_quantization

Mar 19, 2026

Commits on Mar 11, 2026

feat(onnx): add QUANT_FORMAT int8/fp8 PTQ via modelopt.onnx
IgorBaratta
committed
fix(onnx): re-derive engine_path from final onnx_path after quant fallback
IgorBaratta
committed
review: fix run.py, temp file cleanup, YAPF, README ONNX section

ivanbasov
and
claude
committed

Commits on Mar 13, 2026

Commits on Mar 16, 2026

fix(ci): YAPF, move nvidia-modelopt to train reqs, add prerequisite tests

ivanbasov
and
claude
committed
feat(onnx): add onnxruntime INT8 fallback for Python 3.13+

ivanbasov
and
claude
committed
fix(prereqs): document tensorrt as optional GPU dep, add fallback tests

ivanbasov
and
claude
committed
fix(onnx): use import-based dispatch for modelopt/ort; install modelopt on py3.13

ivanbasov
and
claude
committed
fix(test): pin ONNX IR version 8 in ort quantize test

ivanbasov
and
claude
committed
test(onnx): add end-to-end mq.quantize() tests for modelopt

ivanbasov
and
claude
committed
fix(test): use float32 calibration data in TestModeloptQuantize

ivanbasov
and
claude
committed
fix(trt): raise RuntimeError when tensorrt missing for ONNX_WORKFLOW=2/3

ivanbasov
and
claude
committed
fix(test): fix ORT calibration for ort quantize test

ivanbasov
and
claude
committed
fix(test): skip ort quantize output test when modelopt is installed

ivanbasov
and
claude
committed