Justin Chu

I haven't found a clean way to annotate complex functions. Currently staying still

@titaiwangms I added tests, PTAL. I will create a separate PR for better docs if that works with you

@pytorchbot merge

@xadupre could you approve? Thanks!

cc @shubhambhokare1 this is a similar issue

Could you fix https://github.com/onnx/onnx/pull/6994/checks?check_run_id=42794836632

You can set external_data=False. It was defaulted to true mainly for the LLMs with large weights. Additionally the exported model will contain metadata_props that include source information for debugging. This...

cc @cyyever @andife

Just passing by and saw this: this usage is perfect; as fyi there is also an `optimize=` option in `export()` that does the same thing, and in PyTorch 2.7 it...

cc @gramalingam