trtcheck

Static pre-flight checker for ONNX to TensorRT conversion.

trtcheck reads an ONNX file, runs five independent checkers, and tells you in seconds whether the model will convert cleanly to a TensorRT engine. If it will not, the report explains what to fix. It runs anywhere Python runs -- no TensorRT, no CUDA driver, no GPU required.

Why

The PyTorch -> ONNX -> TensorRT pipeline fails most of the time on the last hop. Errors are cryptic; the iteration loop ("export, wait, read a C++ traceback, google, try again") burns hours per fix.

trtcheck predicts the failure modes locally so you correct them before invoking trtexec.

Quick links

• Install -- pip install trtcheck
• Usage -- CLI flags, examples, CI integration
• Fixers -- what --fix rewrites and when it refuses
• Operators -- per-operator TensorRT support matrix

What it checks

Checker	Catches
operator support	Ops missing or partial in the target TRT version
precision	UINT8 / FLOAT64 / STRING inputs, INT64 weights, BF16 on older targets
dynamic shapes	Multiple symbolic dims on inputs
control flow	Loop with runtime trip count, nested Loop, If, Scan
graph structure	Empty outputs, duplicate node names, oversized constants