Why can a TensorRT engine reject an input shape even though the ONNX model has dynamic dimensions?
GPU / CUDALevel: AdvancedTensorRT dynamic-shape optimization profiles
Why can a TensorRT engine reject an input shape even though the ONNX model has dynamic dimensions?
Hints
- Distinguish ONNX shape symbols from built-engine constraints.
- What roles do min, opt and max shapes play?
Answer
This answer is part of mentoring
Advanced questions come with a full breakdown and a mock interview with a working quant researcher. The first call is free.
Source and licence: OpenMMLab — MMDeploy FAQ · Apache-2.0 · answer: Quantoros editorial