llm_export_utils

Modules

modelopt.onnx.llm_export_utils.export_utils	Utilities for exporting LLM models to ONNX.
modelopt.onnx.llm_export_utils.quantization_utils	Quantization utilities for LLM models.
modelopt.onnx.llm_export_utils.surgeon_utils	Utilities to surgeon ONNX graph after export.

Deprecated shim for the legacy modelopt.onnx.llm_export_utils package.

The in-repo LLM ONNX export pipeline (formerly examples/torch_onnx/llm_export.py plus this package) was removed in 0.44.0rc1 in favor of TensorRT-Edge-LLM, which provides a more complete and actively maintained pipeline.

This package is preserved only as a compatibility shim so external consumers that still import modelopt.onnx.llm_export_utils (notably TensorRT-Edge-LLM 0.6.1 and earlier) continue to work. It will be removed in a future release.

New code should migrate to:

• modelopt.onnx.export — quant exporters (FP8QuantExporter, NVFP4QuantExporter, etc.)

• modelopt.onnx.graph_surgery — graph transforms (GQA replacement, BF16 conversion, etc.)

• TensorRT-Edge-LLM — end-to-end LLM export.