Alexandre Eichenberger

IBM folks are not currently working on any GPU accelerators in ONNX-MLIR and given our focus, I don't anticipate any internal needs in the near/medium timeframe, As an open source...

> Is it possible to perform inference on GPU using the PyRuntime API? OyRuntime API is simply there to invoke a model that was compiled. PyRuntime is very very lean,...

Working on compress

FYI, here are some of the benchmarks we are focusing on and that have ops that are not working yet. high priority: (from model zoo) - Roberta, - Bertsquad (onehot),...

working on one hot to work with multiple types

Status of implemented ops are listed here now: https://github.com/onnx/onnx-mlir/blob/main/docs/SupportedONNXOps-cpu.md

Can someone look into this one? I am getting a strange error ``` (onnx-mlir) alexe@pancetta mnist % onnx-mlir conv_mnist.onnx Assertion failed: (!scopes.back().contain(name) && "Tensor already exists."), function AddMapping, file /Users/alexe/Onnxcode/onnx-mlir/src/Builder/SymbolTable.hpp,...

To rephrase your situation, you want to use some of the onnx-mlir passes but are not planning to use either the onnx-mlir or onnx-mlir-opt drivers. And you are stating that...

I now realize that they were two versions for the same original issue. After reviewing the concepts behind #1363, you convinced me that that one is better. Thanks for having...

I believe we have some restrictions to run CIs on our internal machines, once its approved and ran successfully, I think you can merge it. But we can do that...