Yulong Wang

The feature is merged in main branch and it's expected to be available in the next dev build or 1.20 release.

> What's the need for writing a wrapper for ostringstream? ostringstream is already general enough. Using a function call for appending to ostream is helpful when using with macros or...

The file is already excluded in the library (NPM package). Keeping it in the code source helps to make it consistency between all dev environment, which avoid strange errors on...

try to fix #8488

> I'm wondering if the problem here is that we aren't making our libraries private or something? Maybe they should be in a subfolder? It seems like this kind of...

@gonnet please help to take a look again

Technically the axes should always be a 1D tensor. However, in reality, the CPU code has loosen the limit: https://github.com/microsoft/onnxruntime/blob/main/onnxruntime/core/providers/cpu/tensor/unsqueeze.cc#L60-L62 perhaps webgpu should have same behavior to CPU. #22054

Creating a custom device is supported in onnxruntime-web, but there are also several things to be careful with: - onnxruntime-web calls [`adapter.requestDevice(params)`](https://developer.mozilla.org/en-US/docs/Web/API/GPUAdapter/requestDevice) with specific `requiredFeatures` and `requiredLimits` in `params`. Creating...

Are all GPU CIs will be merged into one in future? Specifically, I would like to know whether WebGPU CIs should be put into this CI or saperatedly.

no. currently WebGPU EP is only enabled in Windows and macOS in Node.js binding. There are multiple reasons: - the current release pipeline does not yet support CUDA + WebGPU...