Add TensorRT export support and inference-models inference integration (#853)
* Add TensorRT export support and inference documentation
* Add RFDETRTensorRT class for engine inference, tensorrt export param, and tests
* Pivot TRT inference to inference-models; remove RFDETRTensorRT class
* Rebase onto develop: tensorrt export + inference-models integration
* docs: fix pip extra, document .engine artifact scope, clarify export() docstring
* fix(export): typed trtexec() signature, fix shell-arg path corruption
* test(export): cover RFDETR.export(tensorrt=True) wrapper, device-restore on failure, CLI kwarg forwarding
* fix: repair stale trtexec() doc example, tautological device-restore test
* refactor(export): fold tensorrt bool flag into format param
* refactor(export): use .trt extension for TensorRT engine output
* feat(export): accept format="trt" as alias for "tensorrt"
* docs(agents): default to small model, never base; scope preview variants
* docs: use rfdetr-small instead of deprecated rfdetr-base in examples
* refactor(export): build TensorRT engine in-process via polygraphy
* docs(agents): supersede seg-preview with released segmentation sizes
* refactor(export): replace pathlib with os for consistent path handling
---------
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: Borda <6035284+Borda@users.noreply.github.com>
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
Copilot
93c1300e0b