Edward Chen

Results 64 comments of Edward Chen

We don't generally support forwards compatibility. v6 was introduced in ORT 1.16.0 so the minimum version you would be able to use a v6 model in is 1.16.0.

This might be where the error message is coming from: https://github.com/onnx/onnx/blob/990217f043af7222348ca8f0301e17fa7b841781/onnx/defs/tensor/defs.cc#L1116-L1128

Many of the log calls that occur even after a session has been created do not go through the session logger. E.g., you can find many uses of `LOGS_DEFAULT` in...

>> Many of the log calls that occur even after a session has been created do not go through the session logger. E.g., you can find many uses of `LOGS_DEFAULT`...

> Do you know how to make Squeeze and Unsqueeze compatible? The NNAPI EP doesn't currently support QDQ Squeeze and QDQ Unsqueeze. > Or if there are other methods to...

> Moving nonessential functions to the onnxruntime/tools/python/util package. Only left function that are meaningful to the build process. Can you clarify what "nonessential" means? I gather it's different from "unused",...

Sorry, we don't support building for armeabi now. It's not supported by recent NDK versions either. https://developer.android.com/ndk/guides/abis#sa "Note: Historically the NDK supported ARMv5 (armeabi), and 32-bit and 64-bit MIPS, but...

> Are dynamic input shapes supported on these two EPs? Because [this page](https://onnxruntime.ai/docs/tutorials/mobile/helpers/make-dynamic-shape-fixed.html) states that they're not, but https://github.com/microsoft/onnxruntime/issues/17654#issuecomment-1730371866 states that support has already been added. The CoreML EP has...

> Is it planned for NNAPI at some point? Unfortunately, no. NNAPI will be deprecated (see [here](https://developer.android.com/ndk/guides/neuralnetworks/migration-guide)). This limits our future plans for this EP. > Also, is there any...

> Is there already documentation I can read about implementing the following points: > > * Add Android QNN support, including a pre-build package, performance improvements, and Phi-3 model support....