Alexandre Eichenberger

See this issue for a few suggestions on how to avoid the infinite loop. https://github.com/onnx/onnx-mlir/issues/941#issuecomment-954221677 I came to realize that V and G can have different sizes: grow G to...

yes, absolutely we should consult with them. We might still be in the grace period... Watson AI Studio control everything from their app, and I believe the release just done...

got it, so we have the Java, Python, C, not the C++ ExecutionSession (which might be ok). Somewhat unrelated, is it an issue that our Debug/lib dir has so many...

@yaochengji which operations are you referring to? There are high level ops in ONNX that can be expressed in terms of other ONNX Ops, typically RNNs but there are others...

@yaochengji Your approach appears sound. Using ONNX-ONNX lowering may reduce the amount of work you need to do as you can then rely on the ONNX-MHLO to do the rest....

Correct, we need to add timing to a few meaningful sizes. @tungld you suggested a benchmark sometimes ago, was it DeepBench? I searched my inbox but could not locate it...

@sstamenova may give you more info. From the code, on windows they build .dll whereas on linux we build .so. Have you tried using the .dll instead?

@vasanthaganeshk did you get it to work? If so, please let us know how you did it so we may add instructions to our doc

Smaller with the krnl.copy removed as well as the init/add C loops. Even the krnl.matmul can be removed (but I kept it as it does add loops and conditionals in...

Confirming that the proposed solution works