PyTorch ONNX Export Support
summary
PyTorch ONNX exporter converts traced torch models by translating torch IR to an ONNX IR and adds concrete features: missing operator export and custom operator support, plus exports models with variable-length inputs/outputs. The exporter adds optimizations such as constant folding, broadens backend compatibility by supporting opset 7–10 (work toward opset 11), and reduces required user code changes for tensor indexing and custom loops. These changes enable production models (speech, CLIP, image tagging/content moderation) to run faster on ONNX Runtime.
AI-written summary · about
chapters (6)
related talks
11mWhat's New in Torch.onnx
10mLightning Talk: Streamlining Model Export with the New ONNX Exporter
30mThe Evolution of ONNX: Driving Interoperability in AI...
33mWhat on earth is ONNX and ONNXRUNTIME
45mLeverage Power of Machine Learning with ONNX
05mONNX Runtime
06mTorchVision
51mMaking neural networks run in browser with ONNX